CONF-781046 ^ U.S. Department of Energy October 1978Luray, Virginia

Coordinated by theLawrence Livermore Laboratory for theAssistant Secretary for Energy TechnologyDivision of Energy Storage Systems

Proceedings of the 1978Mechanical and MagneticEnergy Storage Contractors'Review Meeting

05- v

CONF-781046Dist. Category UC-94b

U.S. Department of EnergyCoordinated by theLawrence Livermore Laboratory for theAssistant Secretary for Energy TechnologyDivision of Energy Storage SystemsWashington, D.C. 20545

October 24-26,1978Luray, Virginia

Proceedings of the 1978Mechanical and MagneticEnergy Storage Contractors'Review Meeting

Edited by:G. C. ChangT. M. Barlow

- N O T I C E -

Thil report was prepared 21 i n accotnl of worksponsored by the United Stiles Government- Neither theUnited Stites nor the United Stites Depsnment ofEnergy, noi any of their employees, nor uiy of theircontractors, subcontractors, or their employees, maKesany warranty, express or Implied, or awimes any legalliability or responsibility for the accuracy, completenessor usefulness of any information, apparatus, product orprocess disclosed, or represents that its use would notinfringe privately owned rights.

TABLE OF CONTENTS

Page

PREFACE v

OPENING REMARKS - George C. Chang vi

WELCOME - James H. Swisher vii

FLYWHEELS

Thomas M. Barlow, Lawrence Livermore Laboratory, "Mechanical Energy StorageTechnology Development for Electric and Hybrid Vehicle Applications" 3

Robert F. McAlevy, Robert F. McAlevy & Associates, "The Impact ofMechanical-Energy-Storage Device Addition on the Performance ofElectric Vehicles" 13

Arthur E. Raynard, Garrett-AiResearch, "Advanced Flywheel Energy StorageUnit for a High Power Energy Source for Vehicular Use" 27

Edward L. Lustenader, General Electric Company, "Regenerative Flywheel EnergyStorage System" 33

D. W. Rabenhorst, Johns Hopkins University, "Low Cost Flywheel Demon-stration" 43

James A. Rinde, Lawrence Livermore Laboratory, "Materials Progr&m forFiber Composite Flywheel: 55

R. G. Stone, Lawrence Livermore Laboratory, "The Laminated Disk FlywheelProgram: A Rotor Development Project by LLL and G. E. Co 67

B. B. Smith, Union Carbide Corporation, "Flywheel Test Facility" 75

Robert 0. Woods, Sandia Laboratories, "Overview of Component Development... 81

E. David Reedy, Jr., Sandia Laboratories, "Sandia Composite-RimFlywheel Development" 87

A. Keith Miller, Sandia Laboratories, "Structural Modeling of Thick-RimRotor" 93

Mel Baer, Sandia Laboratories, "Aerodynamic Heating of High-Speed Fly-wheels in Low-Density Environments" 99

M. W. Eusepi, Mechanical Technology, Inc., "The Application of FluidFilm Bearings and a Passive Magnetic Suspension to EnergyStorage Flywheel s" Ill

David B. Eisenhaure, The Charles Stark Draper Laboratories, "Low-LossBall Bearings for Flywheel Applications" 123

I. Anwar, The Franklin Institute, "Seal Studies for AdvancedFlywheel Systems" 129

Flywheels (Conf d)

Francis C. Younger, William M. Brobeck and Associates, "A CompositeFlywheel for Vehicle Use" 141

P. Ward Hill, Hercules, Inc., "Progress in Composite FlywheelDevelopment" 155

Donald E. Davis, Rockwell International, "Advanced Composite Flywheelfor Vehicle Application" 163

D. L. Satchwell, Garrett-AiResearch, "High-Energy-Density Flywheel".. 171

SOLAR MECHANICAL

B. C. Caskey, Sandia Laboratories, "Solar Mechanical Energy StorageProject" 179

David G. Uliman, Union College, "The Band Type Variable Inertia Fly-wheel and Fixed Ratio Power Recirculation Applied To It" 185

Arthur G. Erdman, University of Minnesota, "Cellulosic Flywheels".... 195

John M. Vance, Texas A&M University, "A Concept for Suppression ofNon-synchronous Whirl in Flexible Flywheels" 205

Francis C. Younger, William M. Brobeck and Associates, "ConceptualDesign of a Flywheel Energy Storage System" 211

Theodore W. Place, Garrett-AiResearch, "Residential Flywheelwith Turbine Supply" 227

SUPERCONDUCTING MAGNETIC ENERGY STORAGE

R. W. Boom, University of Wisconsin, "Superconductive DiurnalEnergy Storage Studies" 237

S. van Sciver, University of Wisconsin, "Recent Component DevelopmentStudies for Superconductive Magnetic Energy Storage" 247

R. L. Cresap, Bonneville Power Corporation, "Power System StabilityUsing Superconducting Magnetic Energy Storage Dynamic Character-istics of the BPA System" 253

Paul C. Krause, Purdue University, "Hybrid Computer Study of a SMESUnit for Damping Power System Oscillations" 263

John D. Rogers, Jr., Los Alamos Scientific Laboratory, "Super-conducting Magnetic Energy Storage" 271

Carl Chowaniec, Westinghouse Electric Corp., "Superconducting Mag-netic Energy Storage for Power System Stability Applications" 283

UNDERGROUND PUMPED HYDROELECTRIC STORAGE

Shiu-Wing Tarn, Argonne National Laboratory, "Underground Pumped

Hydro Storage -- An Overview" 293

Underground Pumped Hydroelectric Storage (Cont'd)

J. Degnan, Allis-Chaimers Corp., "Evaluation of One and Two StageHigh Head Pump/Turbine Design for Underground Power Stations" 305

Alexander Gokhman, University of Miami, "Multistage Turbine-Pumpwith Controlled Flow Rate" 315

COMPRESSED AIR ENERGY STORAGE

Walter V. Loscutoff, Battelle Pacific Northwest Laboratories," Compressed Ai r Energy Storage Program Overvi ew" "331

L. E. Wiles, Battelle Pacific Northwest Laboratories, "Fluid Flowand Thermal Analysis for CAES in Porous Rock Reservoirs" 337

J. R. Friley, Battelle Pacific Northwest Laboratories, "ThermoMechanical Stress Analysis of Porous Rock Reservoirs" 349

J. A. Stottlemyre, Battelle Pacific Northwest Laboratories, "PotentialAir/Water/Rock Interactions in a Porous Media CAES Reservoir" 355

P. F. Gnirk, RE/Spec Inc., "Preliminary Design and StabilityCriteria for CAES Hard Rock Caverns" 363

R. L. Thorns, Louisiana State University, "Preliminary Long-TermStability Criteria for CAES Caverns in Salt Domes" 375

H. J. Pincus, University of Wisconsin, "Fabric Analysis of RockSubjected to Cycling with Heated, Compressed Air" 385

Shosei Serata, Serata Geomechanics Inc., "Numerical Modeling of Beha-vior of Caverns in Salt for CAES" 393

Frederick W. Ahrens, Argonne National Laboratory, "The DesignOptimization of Aquifer Reservoir-Based CAES" 403

George T. Kartsounes, Argonne National Laboratory, "Evaluation ofTurbo-Machinery for Compressed Air Energy Storage Plants" 417

George T. Kartsounes, Argonne National Laboratory, "Evaluation ofthe Use of Reciprocating Engines in Compressed Air Energy StoragePI ants" 427

Walter V. Loscutoff, Battelle Pacific Northwest Laboratories,"Advanced CAES Systems Studies" 439

Gerrard T. Flynn, Massachusetts Institute of Technology, "SolarThermal Augmentation of CAES" 449

Table of Contents (Cont'd)

APPENDICES

Alan Mi liner, Massachusetts Institute of Technology, LincolnLaboratory, "The Application of Flywheel Energy Storage Tech-nology to Solar Photovoltaic Power Systems" 457

Program... 463

List of Attendees 469

IV

PREFACE

The Mechanical and Magnetic Energy Storage Annual Contractors' ReviewMeeting was held in Luray, Virginia, on October 24-26, 1978, to review thecurrent research and development activities being carried out in theseprogrammatic areas by the Department of Energy's Division of Energy StorageSystems. Approximately 140 representatives from government, national labor-atories, universities and industry attended the three-day meeting, duringwhich 45 presentations were made by major DOE contractors and subcontractors.

Within the Division of Energy Storage systems, the responsibility forthe development of mechanical and magnetic energy storage lies within theAdvanced Physical Methods Branch. Included are five technology areas —flywheels, mechanical energy storage for solar/wind applications, compressedair energy storage, underground pumped hydroelectric energy storage, andsuperconducting magnetic energy storage. Applications for the technologyinclude ground transportation, solar energy systems, and large-scaleelectric utility systems.

The meeting was organized and conducted by the Lawrence LivermoreLaboratory under the direction of Dr. George C. Chang, Chief, AdvancedPhysical Methods Branch of the Division of Energy Storage Systems.Session chairmen were selected from the active program participantsfor the eight designated technical sessions. Specific arrangementsfor the respective sessions were coordinated through the sessionchairmen. Credit for the success of this meeting is largely due tothe active support of these session chairmen, as well as that of theindividual speakers and of the Bradford National Corporation, throughwhom the administrative arrangements for the meeting were made.

The technical papers which make up these proceedings wereindividually propared by the responsible contractors and subcontractors;the function of the "editors" in this case has been primarily that ofreviewing the individual submissions for completeness and assembling thedocument for printing. Technical editing, in the formal sense, has beenminimal.

Thomas M. Barlow, ChairmanMechanical and Magnetic Energy StorageAnnual Contractors Review Meeting-1978

OPENING REMARKS

George C. Chang

As the Branch Chief for Advanced Physical Methods, I am happy to welcome youto the 1978 Mechanical and Magnetic Energy Storage Contractors' Review Meeting. Weenvision this to be the first of a series of annual meetings for which there are twopurposes. The first is to make it possible for DOE Headquarters personnel to obtain,with a minimum of travel, a comprehensive overview of our programs and an acquaintancewith the participants. The second purpose is to stimulate the exchange of informationand ideas between different organizations and disciplines. To help accomplish the firstpurpose we have located this meeting in a pleasant place with easy access to Washington.To accomplish the second purpose we want to encourage all of you to be open and unin-hibited in making your comments and constructive criticism during the presentationsto follow. Me also plan to publish the proceedings of this meeting so that the generalpublic can have ready access to the information presented here.

During the next few days you will hear reports from a wide variety of disci-plines, ranging from compressed air for utility peaking plants to flywheels forregenerative braking. You will hear about such esoteric subjects as superconductingmagnets and more basic concepts such as underground pumped hydro. Each of thesestorage technologies has something special to offer. Perhaps the only quality theyhave in common is that all of them are advanced physical methods as opposed to chemical,thermal, thermo-chemical, or electro-chemical methods. Some of the technologies are fordispersed systems. The best examples are flywheels to store solar energy captured byphotovoltaic collectors. Other technologies apply exclusively to centralized systems.Compressed air storage to be used in utility peaking plants is the outstanding example.Some of our programs such as magnetic energy storage have a long-term pay-cff. Others,such as compressed air, are almost ready for near-term exploitation. Flywheels forautomotive application are a mid-term technology which we expect to be ready forcommercialization in the mid-to-late 80's.

We are pleased both with our mix of long-term and short-term projects and withour mix of dispersed and centralized technologies. We are also pleased with the progressthat has taken place since 1975 when these programs were initiated. We hope that you,too, will be impressed as the details of our programs unfold in front of you during thenext three days.

VI

WELCOMING REMARKSI-

>i James H. Swishert

ii I, too, would like to welcome you to the conference. For the past two yearswe have had annual contractors' review meetings for both the Thermal Storage Programand the Hydrogen Program, but until this year we have never had one for the Mechanicaland Magnetic Programs. It is our intent to make this an annual event.

In the past, we have found that we had a great deal of difficulty insending all the Headquarters staff to annual meetings away from Washington. Hence, we

\ would like to continue to have the meeting within a 150 mile radius of Washington.Also, we thought that we were missing out by not encouraging and enticing othergovernment people to come to some of our reviews to learn more about our programs. Wehave made an attempt this year to see if some Congressional staff aides might liketo come. We have also invited representatives from both DOE policy groups and othergovernment agencies. By looking around the room I can see that we have beensuccessful in obtaining the wider audience which we have been seeking. I amparticularly pleased to be able to offer a special welcome to the new participants inour meetings. I hope that you will favor us with your comments and suggestionsduring the days to come.

vii

SESSION I: FLYWHEELS

PROJECT SUMMARY

Project T i t le : Mechanical Energy Storage Technology Development for Electricand Hybrid Vehicle Applications

Principal Investigator: Thomas M. Barlow

Organization: Lawrence Livermore LaboratoryMail Stop L-209P. 0. Box 808L-svermore, CA 94550(415) 422-6434

Project Goals: The goals of the E&HV-MEST project are to develop and demonstratemechanical energy storage technology for effective application toelectric and hybrid vehicles in accord with the provisions of theElectric and Hybrid Vehicle Research, Development and DemonstrationAct and to maximize the commercialization potential of this tech-nology.

Project Status: Work was init iated in the final quarter of FY1977; three majorsubcontracts for the development of MES technology have been placedand are scheduled for completion in FY1979. Two of these, withGarrett AiResearch and General Electric CR&D, respectively,involve fully-contained flywheel energy storage systems withelectric input/output machines. The third contract, with theEaton Corporation, is for the design and evaluation of anelastomeric energy storage subsystem.

Contract Number: W-7405-ENG-48

Contract Period: Sept. 1977 - Sept. 1979

Funding Level: $1,400,000

Funding Source: U. S. Department of Energy

MECHANICAL ENERGY STORAGE TECHNOLOGY DEVELOPMENTFOR

ELECTRIC AND HYBRID VEHICLE APPLICATIONS*

Thomas M. BarlowLawrence Livermore Laboratory

P.O. Box 808 (L-209)Livermore, California 94550

ABSTRACT

The Department of Energy authorized t'.a E&HV-MEST Project in September 1977 toprovide a focus for its efforts to develop mechanical energy storage technology forapplication to electric and hybrid vehicles. Technical management responsibility forthe project was assigned to the Lawrence Livermore Laboratory. LLL has, in the pastyear, contracted with industry for the development of two advanced flywheel conceptsand one elastomeiic energy storage concept, all applicable to regenerative brakingand designed to improve the performance and fuel economy of electric vehicles.Additional efforts include an experimental study of the effect of load leveling onbattery life and analytical evaluations of mechanical energy storage technology.These activities are integrated in an overall plan and management structure designedto enhance the commercialization of electric vehicles.

INTRODUCTION

The Electrical and Hybrid VehicleMechanical Energy Storage Technology(E&HV-MEST) Project was created in Sep-tember 1977 by the U.S. Department ofEnergy for the purpose of developing anddemonstrating the technology of flywheelsand other mechanical energy storagesubsystems for application to electric andhybrid vehicles. The responsibility forconducting the project was delegated tothe Lawrence Livennore Laboratory (LLL).

This paper describes the backgroundwhich led to the establishment of the E&HV-MEST Project and identifies the authorityunder which it is conducted. It alsopresents the project plan and the organizationby which the plan is carried out. Asummary of the progress made in FY 1978 ineach of the task areas describes the specificefforts undertaken, both within the Labora-tory and by subcontract. Other presentationsat this meeting will provide additionaldetail concerning the subcontracted efforts.

BACKGROUND

FLYWHEEL TECHNOLOGY DEVELOPMENT

The U.S. Department of Energy'sflywheel technology program has shownsteady growth and progress since itsinception in 1975 . As a result of anearly technical and economicfeasibility study,2 the main focus ofthe program has been on the developmentof composite rotors which have highenergy density and can be economicallyproduced. Application of thetechnology to the transportationsector has been emphasized.

In the technical and economicfeasibility study noted above, severalfindings were reported which haveguided the DOE Program:

o The cost of the flywheel is themajor determining factor inthe application of flywheeltechnology.

o The development of high energydensity is the dominant factorin reducing flywheel cost.

"Work performed under the auspices of theU.S. Department of Energy by the LawrenceLivcrmore Laboratory under contract numberW-7405-ENG-48."

• The development of composite ro-tors with high energy densitieswill broaden the applicability offlywheels in the transportationsector and could provide oil sav-ings in the range of 100 millionbarrels of oil (0.56 quads) peryear by 1995.

• Flywheel safety need not be regard-ed as an unusually severe problem.

• Flywheel systems with metallic ro-tors are applicable to certainnear-term vehicle systems and wouldresult in moderate oil savings.

hire recent studies3"5 have reinfor-ced these findings and have further de-fined the potential of flywheels (andthat of other mechanical energy storageconcepts) to provide electric and hybridvehicles with improved performance, in-creased range, and extended battery life.

In addition to the technology studiesconducted under the DOE Program, technol-ogy development has also been addressedin the laboratory on both the componentand subsystem levels. These efforts havebeen reported in the flywheel technologysymposia6'7 and elsewhere and include thedevelopment of advanced composite rotorsand other components, as well as the ad-vancement of complete systems for auto-motive application.

THE ELECTRIC AND HYBRID VEHICLE ACT OFT97o"

The enactment of the Electric andHybrid Vehicle Act of 1976 (Public Law94-413) gave added impetus to the DOEflywheel technology program. Within theAct, Congress authorized the Departmentof Energy to conduct a program designedto promote electric vehicle technologiesand to demonstrate the commercial feasi-bility of electric and hybrid vehicles.8

The Act cites recognition of the factsthat the Nation's dependence on foreignoil must be reduced and that the intro-duction of electric and hybrid vehiclesinto the transportation fleet could havea substantial impact on the use of pet-roleum in this country.- By means of theAct, Congress declared the policy to en-courage and support research and devel-opment of electric and hybrid vehicles;to demonstrate the economic and technical

practicability of electric and hybrid ve-hicles; and to promote the introductionof electric and hybrid vehicles wherepractical and beneficial. With regardto energy storage technology, the Actprovides for research and development;regenerative braking is specificallyidentified as an area for development.

E&HV-MEST PROJECT SUMMARY

Through the cooperation of the DOEDivisions of Energy Storage Systems andTransportation Energy Conservation, theE&HV-MEST Project is carrying out themechanical energy storage technology de-v.lopment specified by the Act. The pro-ject was authorized by DOE in September

1977 and is being conducted by the Univ-ersity of California's Lawrence LivermoreLaboratory in tivermore, California. Theproject emphasizes the development andevaluation of mechanical energy storagetechnology on a subsystem level. While mostof the experimental work is carried out bysubcontract to private industry, both uni-versities and the National Laboratoriesare involved in the analytical and projectplanning efforts.

THE E&HV-MEST PROJECT

GOALS AND APPROACH

The goals of the E&HV-MEST Project areto develop and demonstrate mechanical energystorage subsystems technology for effectiveapplication to electric and hybrid vehiclesin accord with the provisions of the Elec-tric and Hybrid Vehicle Research, Develop-ment, and Demonstration Act of 1976 and tomaximize the commercialization potential ofthis technology through the continuing in-volvement of the private sector.

In carrying out the project, the po-tential for energy savings is investigatedthrough concurrent theoretical analysis andlaboratory evaluation. Emphasis is givento technology development efforts which are,for the most part, carried out by subcon-tract to private industry. Supporting an-alyses are conducted by the National Labor-atories, universities, and other organiza-tions as appropriate.

Efforts within the E&HV-MEST Projectare coordinated with complementary projectswithin the Energy Storage Technology Pro-gram and the E&HV Program. Information

Technologydevelopment

and evaluation

• Flywheels

• Other MESconcepts

E & HV-MESTproject

Advancedcomponentevaluation

• Mechanicalcomponents

• Safety andcontainment

• Controls anddata systems

Systemsapplicationassessment

• Energy storagerequirements

• Technologyassessment

Technicalmanagement

• Project planningand analysis

• Project direction

Fig. 1. E&HV-MEST Project Organization

exchange is carried out routinely in anumber of areas of mutual interest.

ORGANIZATION

The E&HV-MEST Project consists offour principal tasks, with subtasks iden-tified as new efforts are planned and in-itiated (Fig. 1). The four task areas,each of which is described in more de-tail in later sections, are:

• Technology Develoment and Evalu-ation, including both flywheelsand other mechanical energy stor-age (MES) concepts, such as hyrau-lic accumulators, elastic energystorage devices, and the like;

• Advanced Component Evaluationwhich emphasizes the line ofcommunication with the EnergyStorage Technology Program;

• Systems Application Assessment;and

• Technical Management, under whichthe project planning and execu-tion functions are carried out.

As noted earlier, the Technology Develop-ment and Evaluation Task is emphasized.

TECHNOLOGY DEVELOPMENT AND EVALUATION

The objective of this task is to ad-vance the current state-of-the-art of me-chanical energy storage technology by de-veloping and evaluating advanced MEST con-cepts on a subsystems level.

The approach followed in this taskis to establish subcontracts with industry,on a competitive basis, for the develop-ment and laboratory testing of specificenergy storage concepts. These develop-ment efforts are designed to incorporatenew technology appropriate for commer-cialization in the early 1980's and toprovide the capability, when applied toelectric and hybrid vehicles, to meet stat-ed improvements in vehicle acceleration,hill-climbing ability, and range. In ad-dition, cost, reliability, and operationalsimplicity are included as factors to beconsidered.

The individual subcontracted effortsare limited to periods of 12 to 18 months,thus allowing an updated level of technol-ogy to be included in subsequent contracts.This approach is designed to encourage max-imum participation by private industry; toallow the development of a number of prom-ising concepts; to assure that the technol-ogy is advanced at rhe maximum rate consist-ent with industry's capabilities; to permitthe development of competitive sources; and,

finally, to provide necessary flexibilitywithin the project.

Included as a part of the technologydevelopment projects are a production andeconomic evaluation and a commercializationplan for the technology should it be suc-cessfully developed and demonstrated.

The FY 1978 plan for the Technologyand Evaluation Task specified the placementof subcontracts with private industry forthe development and testing of advanced MESconcepts applicable to regenerative brakingand combined (flywh&tVbattery or flywheelbattery/internal combustion engine) powersystems. Accomplishments during the yearincluded the successful negotiation of acontract with General Electric CR&D forthe Phase II development of the inductormotor/alternator/flywheel energy storageconcept and with the Garrett AiResearchManufacturing Company of California forthe development of an advanced flywheelenergy storage concept which features arim-type composite flywheel rotor and anelectrical input-output machine* A thirdcontract was awarded to the Eaton Corpor-ation in September 1978 as a result ofa competitive proposal process. Each o£the subcontracted technology developmentactivities is designed to provide a re-generative braking capability to electricvehicles and each will provide improve-ments in performance (acceleration), fueleconomy, and range.

Figure 2 summarizes the schedule ofthese activities, and indicates their con-tinuation into FY 1979.

The General Electric and GarrettAiResearch efforts will be adequately ad-dressed in separate reviews by Lustenaderand Raynard, respectively. Because theEaton contract was quite recently nego-tiated, and because it involves a rathernovel concept, some additional detail con-cerning both the concept and the processby which it was selected is provided here.

During the third quarter of FY 1978,proposals were solicited and evaluatedfor new MES technology development efforts.Thirty-three "Invitations For Proposal"were sent to companies having expressedan interest in the effort. Thirteen pro-posals (which included a variety of energystorage concepts) were received and eval-ated; from these, the Eaton Corporationand the AVCO Corporation were selected forfurther action, with priority being directedto Eaton. Current funding limitations havelimited contract action to Eaton; furthernegotiations with AVCO will be undertakenas funding is available.

The concept proposed by the Eaton Corp-oration involves the application of an elas-tomeric energy storage concept to regener-ative braking in an urban vehicle. Theconcept is quite simple and involves a min-imal amount of stored energy (less than

Garrett AiResearchContract negotiationDesign studyHardware design and fabTest and evaluation

General Electric CR&DContract negotiationDesign studyHardware design and fabTest and evaluation

Eaton CorporationProposal and negotiationPhase I design studyPhase II fab and test

Fig . 2 . FY 1978 Technology Development and Evaluation Task Schedule

100 wh), yet the application is one inwhich significant oil savings could berealized. The first phase of this effortinvclves a design analysis and evaluation,which should be completed in January 1979.If the Phase I results are favorable,a Phase II contract will be entered. ThePhase II effort will involve a completelaboratory test and evaluation of the con-cept.

The AVCO Corporation proposal des-cribed the development and evaluation ofa flywheel energy storage concept whichfeatures a disc-type composite rotor andan infinitely variable mechanical trans-mission. This concept could be appliedeither to regenerative braking or to a"combined" power system and is particu-larly attractive because of the high-efficiency, all-mechanical power trans-mission system.

ADVANCED COMPONENT EVALUATION

The objective of this task is to pro-vide an evaluation of advanced componenttechnology from inclusion in demonstrationsubsystems. A variety of components areinvolved, including flywheel rotors, bear-ings, and seals; safety and containmentsystems; and various transmission con-cepts, both mechanical and electrical.

This task is carried out in coordin-ation with the DOE Flywheel Energy StorageTechnology Project, for which the SandiaLaboratory, Albuquerque, New Mexico, hasresponsibility.

A significant flywheel rotor burstand containment test was conducted and re-ported by the Garrett AiResearch Companyin the third quarter of FY 1978. Thetest was carried out as a part of theGarrett Company's development effort underthe Near Term Electric Vehicle Program(DOE Contract EV-76-C-03-1213). A rim-type flywheel was tested to failure at26,000 rpm within a closed containmentsystem. The purposes of the test wereto demonstrate the system's capabilityto contain the flywheel failure; to es-tablish flywheel failure data; and toobtain data pertinent to the design andanalysis of containment devices.

The failure test was conducted bybringing the flywheel rotor up to testspeed in a vacuum environment and then

slowly increasing the pressure (fromvacuum toward atmospheric) in the chamber.At about 4500 microns (4.5 Torr) the outertwo layers of the rotor failed. Thefailure was completely contained, andno yielding of the housing was noted.The maximum torque experienced was 1100ft lbs., and the time required for therotor to come to a stop was approximatelytwo seconds.

The test is significant in the factthat the non-catastrophic failure modeof the rim-type composite rotor was dem-onstrated 3nd the integrity of the con-tainment concept was proven.

SYSTEMS APPLICATIONS ASSESSMENT

The objectives of this task are todefine the applicability of MES subsystemsto electric and hybrid vehicles of var-ious types; to define the functional anddevelopmental requirements for these sub-systems; and to provide technology devel-opment criteria as required by otheractivities within the project.

As indicated by the statement of ob-jectives, this task provides guidance andcriteria for the Technology DevelopmentTask. It includes the consideration ofcomplementary energy storage technologyin order to provide perspective to the MESefforts. Specifically, the following top-ics are addressed:

• Energy storage system functionalrequirements,

• Identification of viable concepts,

• Technology development requirements,

• Economics and production evaluation,

• Environmental and safety consider-ations, and

• Technology utilization.

The operating plan for this task inFY 1978 included an evaluation of the effectof load leveling on battery life, a deter-mination of the potential impact of MEStechnology on electric vehicle performance,and an analysis of the MES performance re-quirements.

The first of these activities is being

addressed through the conduct of a coopera-tively supported battery load-leveling testat the U. S. Army's Mobility Equipment Re-search and Development Command (MERADCOM),Ft. Belvoir, Virginia. The test includestwo sets of batteries - one is cycled to a"normal" load profile experienced by abattery-powered transit vehicle, and thesecond includes a load profile which simu-lates the effect of flywheel load levelingon the battery set. This test is currentlyunderway, and an indication of the effectof load leveling is expected when it iscompleted next year.

The second and third activities arebeing addressed analytically in separateefforts by Dr. R. F. McAlevy and by theLLL Transporation Systems Group. A reportof the LLL analysis is being prepared andwill be presented in the future, and Dr.McAlevy will report his findings later inthis meeting.

As a general comment, several criteriaaffect the design of the energy storagesystem for vehicle applications:

• Specific energy is important forrange, long grades, high-speedpassing, and other conditionsrequiring considerable quantitiesof energy.

• Specific power is important foracceleration, braking, and highspeed grades and passing.

• Cost is always Important, and bothinitial and life-cycle cost mustbe considered.

• The vehicle load cycle is import-ant in that it specifies the re-tired acceleration and braking,stops per mile, and cruise speedand time - all of which impactthe design of the energy storagesystem.

The unique characteristics of fly-wheels (and, to some extent, other mech-anical energy storage concepts) can beused to advantage to extend the range ofelectric vehicles, improve their acceler-ation characteristics, and provide themwith increased battery life. These char-acteristics include the unequaled abilityto effectively recover and store, forfuture use, the energy normally lost in

vehicle braking and downgrade travel.Additional design flexibility is afforded,since the specific power of a flywheelis essentially Independent of its specificenergy. Flywheels can be designed witha high specific power capability, whichcan be utilized to provide accelerationand regenerative braking capability nototherwise achievable in electric vehicles.

TECHNICAL MANAGEMENT

Activities within this task includepreparing and maintaining the project plan;implementing the plan by developing cri-teria, preparing specifications, and estab-lishing technology development projects;directing contractor efforts and reportingprogress to DOE; and recommending companionresearch and development efforts. Each ofthese activities has been addressed inFY 1978, with the Project Plan being sub-mitted in April 1978; proposals for tech-nology development projects invited and con-tracts awarded to three companies; and reg-ular progress reports provided to the DOEDivisions of Energy Storage Systems andTransportation Energy Conservation. In ad-dition to the regular project reviews,project personnel participated in elevenworkshops, review meetings, and symposiaduring the year in order to facilitate co-ordination with complementary efforts car-ried out elsewhere.

Figure 3 summarizes the resource allo-cations for the project (in Budget Author-ity) by task and by performing institution.As the figure indicates, the majority ofthe funding supports the Technology Devel-opment and Evaluation Task, and private in-dustry is, by far, the primary "performer."Funding for the MERADCOM test is reflectedas "U. S. Army," while the university seg-ment consists primarily of faculty consul-tants to the project.

As noted earlier, the E&HV-MESTProject interacts formally with a numberof organizations which support DOE in boththe Energy Storage Technology Program pprtthe Electric and Hybrid Vehicle RD&DProgram. Figure 4 schematically indi-cates these interrelationships:

• The DOE Division of Energy StorageStorage Systems (STOR) is respon-sible for the Flywheel TechnologyProject, for which the Sandia Lab-oratory in Albuquerque (SLA) has

By Task By Performing Institution

SyitamiTechnical Application

Management,. .^Assessment Universities2%

U.S.Army5%

Fig. 3. Distribution of rY1978 Budget (Budget Authority)

Flywheeltechnologydevelopment

VehicleMEStechnology

Vehiclesystemstechnology

Propulsionsystemtechnology

Fig. 4. E&HV-MEST Project Interactions

been delegated execution responsi-bility. STOR provides the tech-nical direction to the E&HV-MESTProject.

The DOE Division of Transporta-tion Energy Conservation (TEC)is responsible for the Electricand Hybrid Vehicle Program- Sup-

porting TEC in this regard are theJet Propulsion Laboratory (JPL) ofthe California Institute of Tech-nology in the vehicle systems areaand the NASA-Lewis Research Center(LERC) in the propulsion systems .area. TEC exercises the execution(budget) authority for the E&HV-MEST Project.

10

FY 1979 PLANS

Current plans for the E&HV-MEST Pro-ject in FY 1979 include completing theGarrett AiResearch and General ElectricCR&D flywheel technology efforts; evalu-ating the Eaton Corporation's elastomericenergy storage concept, and initiatingthe Phase II development, if warranted;and completing the MERADCOM battery load-leveling test. Additional tasks, includingthe development of the AVCO flywheel energystorage concept, will be initiated as fund-ing permits. Supporting activities will becontinued in the areas of applications an-alysis and technical management.

CONCLUSIONS

The first year's progress of the E&HV-MEST Project has been significant:

• Three major technology developmentefforts are being carried out undersubcontract to private industry.

• The effect of load leveling on bat-tery life is being experimentallyinvestigated.

• Mechanical energy storage systemperformance criteria for vehicleapplications has been determined.

• An analysis of the importance ofMES to electric vehicle performancehas been completed.

Although the current state-of-the-artis adequate for some vehicular applications,further development is required, on boththe component and subsystems level, inorder to improve the performance and eco-nomics of mechanical energy storage tech-nology.

REFERENCES

1. G. C. Chang, J. H. Swisher, and G. F.Pezdirtz, "DOE's Flywheel Program,"in 1977 Flywheel Technology SymposiumProceedings, G. C. Chang and R. G.Stone, Editors, U. S. Department ofEnergy Report CONF-771053, 1978.

2. Economic and Technical FeasibilityStudy for Energy Storage Flywheels,Energy Research and Development Admin-istration Report ERDA 76-65, 1975.

3. E. Behrin, J. Bolger, C. Hudson, L.O'Connell, B. Rubin, M. Schwartz, C. -Waide, and W. Walsh, Energy StorageSystems for Automobile Propulsion.Lawrence Livermore Laboratory ReportUCRL-52303, Vols. I and II, 1977.

4. Determination of the Effectivenessand Feasibility of RegenerativeBraking Systems on Electric and OtherAutomobiles. U. S. Department ofEnergy Report UCRL/W52306, Vols. Iand II, 1978.

5. R. F. McAlevy, R. F. McAlevy and As-sociates, Hoboken, IJJ, Private com-munication, September 1, 1978.

6. Proceedings of the 1975 Flywheel Tech-nology Symposium. G. C. Chang and R.G» Stone, Editors, Energy Research andDevelopment Administration ReportERDA-76-85, 1976.

7. 1977 Flywheel Technology Symposium Pro-ceedings . G. C. Chang and R. G. Stone,Editors, U. S. Department of EnergyReport CONF-771053, 1978

8. Introduction to the ERDA Electric andHybrid Vehicle Demonstration Project.Energy and Research Development Admin-istration Report ERHQ-0008, 1977.

9. B. H. Rowlett, Burst and ContainmentTest Report for Flywheel Power Systems.Near-Term Electric Vehicle ProgramPhase II. Garrett AiResearch Manufac-turing Co. of California Report 78-15148, 1978.

NOTICE"This report was prepared as an account of worksponsored by the United States Government.Neither the United States nor the United StatesDepartment of Energy, nor any of their employees,nor any of their contractors, subcontractors, ortheir employees, makes any warranty, express orimplied, or assumes any legal liability or respon-sibility for the accuracy, completeness orusefulness of any information, apparatus, productor process disclosed, or represents that its usewould nol infringe privately-owned rights."

11

Project Title:

PROJECT SUMMARY

The Impact of Mechanical-Energy-Storage DeviceCharacteristics on Hybrid Vehicle Performance

Principal Investigator:

Organization:

D r . R o b e r t F . M c A l e v y I I I

Robert F. McAlevy I I I & Associates1204 Bloomfield StreetHoboken, NJ 07030

Project Goals:

Project Status:

Establishment of a rational basis for evaluatingthe impact on automotive vehicle performanceproduced by addition of a mechanical-energy-storage-device.

Phase I work was in i t i a ted on May 25, 1973 andcompleted on September 1 , 1978. Phase I I e f for tswere funded on September 14, 1978, with an expectedcompletion date of December 1 , 1978.

Contract Number: University of Cal i fornia Purchase Order 2122409

Contract Period: May 25, 1978 - Dec. 1, 1978

Funding Level: $22,410

Funding Source: Lawrence Livermore Laboratory

13

THE IMPACT OF MECHANICAL-ENERGY-STORAGE-DEVICE ADDITIONON THE PERFCIUVIANCE OF ELECTRIC VEHICLES

Robert F. McAlevy IIIRobert F. McAlevy III & Assocs.

1204 Bloomfield St.Hoboken, New Jersey 07030

ABSTRACT

Electric vehicle, EV, power and energy balances were used to estab-lish battery requirements as a function of vehicle mission specification.Normal freeway/urban commuter car driving patterns were found to requirebattery power capabilities in excess of those forecast for this century.However, Ni/Zn and other batteries should ptrmit EV's to follow normalurban commuter car driving patterns. A criterion for the optimum batteryis evolved and used to delineate between power-determined (battery massin excess of that needed to meet range specification) and range-deter-mined EV designs. The impact of adding a mechanical-energy-storage, MES,device to an EV depends strongly on the "baseline" EV design. For range-determined EV's, analytical relationships are developed for the maximumMES device mass permitted as a function of energy-efficiency incrementproduced, so that there will be no change in: (a) vehicle mass and (b)vehicle energy consumption,compared to the baseline EV. The limitingcase of MES device mass approaching zero (e.g., infinite power-density andenergy-density) while the increment of energy-efficiency remains is alsoexamined. Therefore, the boundaries of beneficial MES device applicationto range-limited EV's has been established. Presently, the analysis isbeing extended to EV's of power determined design, where the payoff forMES device addition can be much greater, and to ICE vehicles.

BACKGROUND AND INTRODUCTION

By incorporating mechanical-energy-storage (MES) devices intoautomotive vehicles it is possibleto improve the performance of thevehicles. The performance incre-ment depends on both the performanceof the original, or "baseline" ve-hicle, and the characteristics ofthe MES device employed.

Electric vehicles, EV's,arethe baseline vehicles considered inthis document. In order to qualifyfor participation in Federally sup-ported programs, they must exhibitcertain minimum performance levels .Recently, the performance of extantEV's was critically reviewed^. Boththe Federal Standards and actualperformance are substantially in-ferior to that exhibited by presentday automobiles of common experi-ence. This shortfall is a directresult of the shortcomings of pres-

ent day batteries compared to a"tank" of liquid fuel. For example,the best lead/acid oattery has agravimetric energy content lessthan 1/100 that of petroleum-basedfuels. Therefore, in order toachieve a range of 60km or so, it isnecessary to allocate about 1/3 ofthe EV mass to batteries (coiflparedto about 1/30 for a fuel gas tankin petroleum-fueled automobiles).In order to carry such battery mas-ses in vehicles with reasonablepayload-fractions, it is necessaryto compromise their structural in-tegrity. For example, while usingthe same materials and fabricationtechniques as automobiles, EV'swere found to have body/chassismass-fractions that are, typically,only 0.45^ compared to 0.65 or sofor present day automobile.

Battery research and develop-ment activities should result inimproved batteries in the future.

A panel of experts has forecast thecharacteristics of future batteriesthrough the end of the century^. Acomputer model was used to calcu-late the performance of EV's thatuse such batteries^.

The subject program involvesan analytical model of baseline EVperformance. It is based on amethodology developed and describedelsewhere* and used previously forcomparisons amongst different kindsof alternate energy vehicles^. Hereit is used to establish the crite-rion for the optimum EV battery asa function of EV mission requirementand to evaluate the potential offuture batteries-^ to power EV'sthrough different missions — spe-cifically, the urban commuter carand freeway/urban commuter car mis-sions. It is' used to establish thethreshold MES device characteris-tics required for "breakeven" per-formance when the devices are addedto baseline EV's of range deter-mined design.

This work is continuing andfuture reports will deal with thebreakeven condition for baselineEV's of power determined design, aswell as the performance incrementspossible by incorporation of MESdevices of known characteristics.It is expected that this work willlead to a rational basis for evalu-ating the impact of MES device ad-dition on the performance of EV's.The benefits that accrue will varyfrom situation to situation, butthey include making it possible forthe MES/electric vehicle to performmissions from which the baseline EVis precluded due to inadequate bat-tery power, improving vehiclestructural integrity, and reducingthe ownership costs of electricallyenergized vehicles — which arecurrently about 5 times greater thanpetroleum-fueled automobiles*". Ap-plication of the methodology topetroleum-fueled baseline vehiclesis planned as well.

ANALYTICAL APPROACH

The framework for the analysiswill be established here. It de-pends on the vehicle power and en-ergy balances and specification of

15

the vehicle mass-fractions alloca-ted for the several vehicle func-tions that must be fulfilled.

The Vehicle Drive Cycle. Drive-cycles are used to model vehicledriving patterns. A drive-cycle ismerely the specification of the ve-hicle's velocity-history over atime interval. The vehicle is pre-sumed to repeat the pattern overits entire range.

Integration of the velocity-history over the drive-cycle inter-val yields the distance traveledduring the drive-cycle, d. Thenumber of drive-cycles accomplishedbefore depleting the battery belowsome reasonable level (say 80% dis-charge) , n, times d yields the ve-hicle1 s practical range, R. That

R = nd (1)

Knowledge of vehicle charac-teristics — aerodynamic-drag-co-efficient, frontal-area, and coef-ficient of rolling-resistance —coupled with knowledge of the in-stantaneous velocity, permits cal-culation of the instantaneous road-load imposed at the vehicle'swheels. Multiplication of the in-stantaneous road-load by the in-stantaneous velocity yields theinstantaneous power. Inspectionof the instantaneous power at eachpoint in the cycle reveals the max-imum vehicle power demand, Pjnax-Integration of the instantaneouspower over the drive-cycle yieldsthe energy requirement at. the ve-hicle's wheels in order to achieveone drive-cycle, Ev.

These simple exercises havebeen performed assuming typical EVcharacteristics? and a variety ofdrive-cycles or a level road. Theresults of these and other calcula-tions** are shown in Table 1.* VJ<j>is the total vehicle mass.

The Scott-cycle representsthe average of experimentally-de-termined driving patterns of auto-mobiles observed in different are-as of this country. About 1/3 ofthe total travel was on freewaysso that Pjnax/ T is sufficient for

*Table follows References.

entrance up freeway ramps and high-speed passing maneuvers, etc. TheEPA cycle is also derived from ob-servation of automobile driving-patterns. They were made in LosAngeles during rush hour. The ab-sence of a freeway component relax-es the maximum power demand from32 to 23 w/kg and energy demandfrom 0.12 to 0.09 whr/kgkm.

The SAE cycles are artificialin that they are not based on ex-tensive observation of actual auto-mobile driving patterns. Rather,they have been promulgated as con-venient driving schedules for thetesting of EV's^. The "D" cyclewas designed to test EV automobilesused as commuter cars. Note thatthe energy requirements and maximumpower requirements are close tothose of the EPA urban cycle, whichis based on commuter vehicle driv-ing patterns. Unfortunately, only1 of 23 EV's tested by the Depart-ment of Energy in order to charac-terize their present "state-of-the-art" ^ could generate sufficientpower to follow the "D" cycle. Infact, it was as a direct result ofthis finding that the Federal EVdemonstration program performancestandard was set at the "C" cyclelevell. The "C" cycle was designedto test electric parcel-post deliv-ery vans, not automobiles.

Near-term objectives for EVimprovement should be to develop abattery that will permit EV com-muter vans to follow the "D" cycleor EPA Urban cycle without havingto pay too large a vehicle weightpenalty. Longer term objectivesmight be to produce a battery thatwill readily allow the vehicle tofollow a Scott-type cycle so somefreeway travel in an EV would be-come possible. By incorporatingMES devices in the EV, it shouldbe possible to relax the need forthe battery to supply all of thevehicle maximum power requirement.

Sizing the Electric Motor/Transmis-sion/Drive- System . The EV motor/transmission/drive-system maximumpower output is Pmax* T n e mass ofthe system required to deliver P^xis defined as W^c yielding a sys-tem "power-density" of (Pmax/wEc)•

The vehicle mass-fraction re-presented by the energy-conversionsystem, a, therefore can be repre-sented as:

" WT(2)

Advanced electric motor sys-tems have power densities approxi-mately equal to 1/5.5 kw/kg3. Usingthe values presented in Table 1,EV's require a= 0.07 for the SAEJ227a/C cycle, a = 0.13 for the EPAUrban cycle and a = 0.18 for theScott-cycle.

Incorporation of an MES woulddecrease the maximum power produc-tion of the electric motor and,therefore, the fraction of <x re-presented by motor mass. But thedrive-train would still have to besized to deliver Pmax at the wheelsso the fractional decrease in awill be less than the fractionaldecrease in motor mass.

Sizing the Battery System for Max-imum Power Demand. Power-traininefficiencies result in an EVbattery maximum power draw that isgreater than (Pmax/Wlp) WT- D e~fining the power-train efficiencyat the maximum power point at rip,then the maximum power-draw imposedon the battery is (Pmax/WT) WTA)P.

Defining Wg as the battery-system mass (including all compo-nents, controls, supports, etc.associated with the system) andthe battery gravimetric peak-power-density as (pVd.), then the maxi-mum power output of the battery is(pTd.) WS.

Equating maximum power demandto maximum power output, yields,

s = <Pmax/wT)wT/np (3)w

or !(Pmax/WT>

np(pTd.)(4)

By incorporating an MES de-vice into an EV that is power de-termined, it relieves the need forthe battery to supply all of the

16

maximum power level. Thus,could oe reduced, or a more econom-ical battery of reduced (pTd.)could be employed, etc.

Sizing the Battery System to Meetthe Range Demand. Due to losses inthe power-train, more energy mustbe supplied to it than can be de-livered to the wheels. The propul-sion efficiency, np# of the energy-conversion-device, which is strong-ly dependent on drive-cycle, ac-counts for these losses. For elec-trochemical batteries, more energymust be stored in them than can bedelivered to the energy-conversion-device. The battery energy effic-iency, rig, which is strongly de-pendent on drive-cycle and rip, ac-counts for the internal dissipa-tion within the battery. The ve-hicle "energy-efficiency" is de-fined as n = ngnp. Note that thisdefinition does not account forlosses during charging of the ener-gy-storage-device, and properly so,since the charging process hasnothing to do with vehicle perform-ance per se once the vehicle is putinto operation.

Defining the gravimetric-en-ergy-content of the energy-storage-device, following charging and be-fore power is drawn, as (es)p_Q,then the balance between the ener-gy initially stored on board,Wg (es)p=Q, and the energy requiredto execute n drive-cycles, can bewritten as:

<es>P=0 WS =

[E /W d]

Substituting Eq. 1 and rearrangingyields,

nSnP(es}P=0[Ev/WTd]

(6)

Defining Rult as ngnp(es)p_Q/[Ev/WTd] permits Eq. 6 tobe written as:

w,.ult

(7)

Note that every term, except(es)p_Q, depends on drive-cycle.[Ev/WTd] depends on vehicle aero-dynamic drag and rolling-resistanceand their interaction with thedrive-cycle, np depends on power-train characteristics and their in-teraction with the drive-cycle andnp depends on battery system char-acteristics and their interactionwith the drive-cycle and rip-

Nevertheless, despite thesomewhat complicated compositionof Rult» it n a s a simple meaning.It is the range achieved by a Uto-pian vehicle composed only of theenergy storage-device, having anenergy-efficiency of ripig and fol-lowing a drive-cycle with an en-ergy-requirement of [Ev/WTd]. Thatis, R •* Ruit when Wg •+ W^.

Real vehicles are composed ofmore than an energy-storage de-vice, so Wg < WT and therefore,R < Ruif

A s previously mentionedWg/WT for EV's are typically 10times greater than that of gaso-line-fueled vehicles, or Wg = 0.3WT. Therefore, for EV's R = 0.3Rult* T h e better EV lead-acidbatteries exhibit (es)p=o = 45whr/kg and r)S = 0.65 over the SAEJ227a/C-cycle. The better power-trainsexhibit rip = 0.7 over the cycle andsince [Ev/WTd] = 0.09 whr/kg-km,Rult s 2 0 0 kln- Therefore, therange of an advanced lead-acid EVover the SAEJ227a/C-cycle, assum-ing W S / W T = 0.3, should be R =(0.3) (200) = 60 km. Experimental-ly-measured valves for R supportthis result2 and thus tend to con-firm the validity of the underly-ing methodology.

Two positive results can ac-crue by incorporating a MES deviceinto a range determined EV. Bysupplying some of the maximum ve-hicle power requirement, the MES"load-levels" the battery powerdraw — thus, increasing rig andbattery lifetime. By capturingsome of the vehicle kinetic ener-gy that otherwise would be dissi-pated in the brake-drums duringstopping, it can later be suppliedto help accelerate the vehicle;this "regenerative braking" re-sults in an increase of rip.

17

Therefore, incorporating an MES de-vice increases n and therefore,Rulf

Distribution of Vehicle Mass-Frac-tions. The vehicle total mass, Wrp,is composed of the sum of the mas-ses of its several components.That is,

WT = WPL + WEC + WS WM + W B (8)

where: W P L is the payload mass

W,'M is the MES-device mass,including that of itscontainer, shaft,gears, etc.

W B is the body/chassismass

and the other terms have been de-fined previously in this section.

Dividing Eq. 8 by W_ yields,

WP WS(9)

where: 6 is the MES-device mass-fraction

B is the body/chassismass fraction

Re-arranging Eq. 9 yields,

w.PL 1 -

l-g-a-6(10)

In the case of 6 = 0, that is,an EV, Eq. 10 reduces to.

W

w.T (1-B-a) = — —P L 1 -

(11)

1-B-a

ppjj is an important ve-hicle parameter. It is the amountof vehicle mass required per unitof payload and is a strong driverof vehicle capital costs. The low-est possible value that Wj/Wpj, cantake is l/l-B-a-6 for an MES/EVhybrid and 1/1-B-a for an EV. Thesevalues are reached in the impracti-cal limit of WS/WT •* 0. That is,

a vehicle with no R. Practical ve-hicles must have W S / W T > 0.

As Ws/WT is increased in re-sponse to increasing range andpower demand, then W T/W P L must in-crease as well. The upper limitfor WS/WT is l-B-a-6 for ME:-/EVhybrids and 1-B-a for EV's, be-cause as these finite levels areapproached W /Vipi, •* °°, which alsorepresents an impractical limit.Practical vehicles must have Ws/WT< 1-B-a.

Since a, B, 6 and WS/WT canbe assumed to be independent of W Tfor vehicles of the same generictype and design values, then W T /

is independent of W^ as well.

It is interesting to notethat if an MES device is added toan EV, while a + 3 are held con-stant, then Wr£./WpL would increaseunless there was a compensatingdecrease in W5/W1J.

Operating Energy Consumption. Thequantity of battery energy con-sumed per unit distance of vehicletravel, C, is given in terms ofthe present notation as.

IEV/W d]—Y. 1

nsnp

(12)

This equation shows thatC - WT-

Division of both sides ofthis equation by Wp L yields

W,PL nsnp "PL

(13)

x is the energy consump-tion per unit distance of payloadmass carried and is a useful pa-rameter for comparing the energy-performance of vehicles.

Since Ws/WpL is independentof Wi>, then C/WPL will be inde-pendent of WT if [Ev/WTd] andare independent of W T . / Tdecreases very slowly with in-creasing W T ' and this variationwill be ignored here. Also

18

should be independent of W T andthis will be assumed here as well.

The vehicle operating energycosts per unit of payload carriedvary directly with C / W P T . Substi-tution of Eqs. 10 and II into Eq.13 shows how C/WpL increases withincreasing Wg/WnT. Incorporationof an MES into an EV would producean increase in C/Wp^ unless therewere compensating changes in VIQ/VI^,and/or ngrip»

It is interesting to note thatif ns = 0-65 and np = 0.7 are sub-stituted into Eq. 12 along with[Ev/WTd] = 0.09 whr/kg-km, whichare appropriate for the lead/acidbattery EV following the SAEJ227a/C-cycle, Eq. 12 becomes C = 0.2 WT,whr/km, which is typical of thebetter EV values found experimental-iyB.

ANALYSIS

In this section, the impacton vehicle performance produced bythe addition of an MES device to anEV will be analyzed. The analyti-cal approach developed in the pre-vious section will be employed todo so. In particular, the minimumMES device characteristics requiredfor no change in vehicle W^/Wp^ andC/Wpj, — the "breakeven" situa-tions — will be established.

The nature of the impact de-pends on whether the EV is of power orrange-determined design, and thiswill be shown to depend on boththe mission specification (i.e.,range and drive-cycle) and the bat-tery electrical characteristics.

The Inf j.uence of Mission Require-ments and Battery Characteristicson EV Design. Prom Eq. 11, it canbe seen that for an EV to be phys-ically realizable, its Wg/W>p mustbe less than 1-B-a. Defining themaximum possible value of battery-device mass-fraction as (Ws/W T) m a x,then/

Thus, for every set of mis-sion requirements (i.e., specifi-cation of R and drive-cycle), thereare minimum values of batteryelectrical characteristics thatmust be exceeded if the EV contain-ing the battery is to be physicallyrealizable.

Defining the minimum value ofbattery gravimetric peak-power-density as (pVa.)min, then, fromEqs. 4 and 14,

min(15)

So, for physically realizable EV'sof power-determined designs, thedecrease of Ws/Wy below (Wg/W T) m a x

with increasing battery (pVct.) canbe expressed as,

ws/wT(p.

(ws/wT)max

(p.d.)(16)

Battery energy capacity isthe product of (es)p=g and ns«These appear as a product in thedefinition of Ruit (Eqs. 6 and 7 ) ,and therefore, Ruit increaseslinearly with increasing ns(es)p=0*Defining the minimum values ofultimate range and battery energycapacity for physically realizableEV's as (Ruit)min a n d [ns(es)p=olninrespectively, then from Eqs. 6, 7and 14,

<Rult>min

1-3-a(17)

and[E /W d] R

So, for EV's of range de-termined design

ws/wT

(ws/wT)(19)

maxRult

= 1-e-a (14)

at the impractical limit ofW T / W P L •+ co.

19

EXAMINATION OP THE ABILITY OF FU-TURE BATTERIES TO MEET FUTURE EVDRIVE-CYCLE POWER DEMAND

Urban/Freeway Commuter-Car DrivingPattern. The modest, for petrole-um-fueled autos, power demand ofthe Scott cycle will be taken hereas the model for urban/freeway com-muter car driving patterns. FromTable 1, to follow this cycle, anEV will have to have a road powercapability of 32 w/kg (or 0.02hp/lb). Note that this is well be-low the power of today's petroleum-fueled vehicles and only 2/3 ofthat forecasted for future petrole-um-fueled vehicles3. Further, itleaves no margin of "reserve power"for use should extraordinary driv-ing conditions occur during thenormal Scott cycle. Reserve powerwould be used if the EV, whiletraveling at high speed, had to ac-celerate to avoid having an acci-dent. For the Scott cycle, a =0.18.

Assuming it is necessary tohold EV £ to 0.45, and thus acceptlower structural values than thoseexhibited by petroleum-fueled ve-hicles,4 then 1-B-a = 1 - 0.45 -0.18 = 0.37. (Note that if thepower train energy-density increas-es in the future a will decreaseand 3 can increase while l-$-a re-mains at 0.37.) Assuming^that np =0.8, then from Eq. 15, (p.d.)min =32/(0.8)(0.37) = 108 w/kg.

When (pTct.) = (pTa.)min» %/WPL = "» an<^ this does not repre-sent a physically realizable ve-hicle. The actual Ws/WT must beless than 1-g-a. It will be as-sumed 5 that prudent E.V. designcalls for WT/WpL = 7 if the acqui-sition costs and operating energycosts (per unit payload mass) areto be reasonable. SubstitutingWT/WPL = 7 and 1-g-a =0.37 intoEq. 11 yields WS/WT = 0.23. [Notethat the lower value of WS/WT com-pared to present practice 2 stemsfrom the larger value of a re-quired by the Scott cyclercomparedto a = 0.07 required by present dayEV's following the low-power SAEJ227(a)C cycle.]

From Eq. 16, (p.d.) =

(Ws/WT)max/(Ws/WT), so (p?d\) =(108)(0.37)/(0.23) = 173 w/kg.This is the battery gravimetricpeak-power requirement that mustbe met if the EV containing it isto follow the Scott cycle and beof prudent (WT/WPL = 7) design.

Present day EV's, powered bylead/acid batteries, cannot evenfollow the SAEJ227"a"D cycle, sothere is no possibility for themto follow the Scott cycle. How-ever, batteries having better elec-trical characteristics hopefullywill become available in the future.Recently, a panel of battery ex-perts have projected, with opti-mism, the electrical characteris-tics of batteries that might beavailable for EV applications inthe 1990-2000 time frame3. The(p7a.)'.<3 are: lead/acid, 94.6w/kg;Ni/Fe, 103.8 w/kg, Ni/Zn, 139.9w/kg; Zn/Cl2* 88 w/kg; Li-Al/FeS,130 w/kg; Na/S (B-alumina), 130w/kg; Na/S (glass), 160 w/kg;Li/air, 129.4 w/kg.

It is clear that none of thesebatteries would be able to poweran EV through the Scoti cycle.That is, all of the (pTd.)'s arebelow (p.d.)mj.n = 173 w/kg. There-fore, if there can be no improve-ment, in fact, over these optimis-tic forecasts, then EV's of pru-dent design will not be able tofollow the Scott cycle — or anydriving-pattern with comparablepeak power demands — in thiscentury1

Consequently, only by employ-ing an MES device to provide thepeak-power demand of the Scott (orcomparable) cycle will electrical-ly-energized vehicles be able tocompete with petroleum-fueledautos in the urban/freeway commu-ter car market. This is a strongimperative for developing appro-priate MES devices.

Urban Commuter-Car Driving Pattern.The EPA urban cycle will be takenhere as the model for urban com-muter-car driving patterns (i.e.,with no freeway component). The(Pmax/WT> is 23 w/kg, which meansthat a is 0.13 so 1-3-a = 1 - 0.45-0.13 = 0.42. Assuming that ^=0.8,

20

then from Eq. 15, (p.d.)min = 2 V(0.8) (0.42) = 68 w/kg.

Assuming that prudent EV de-sign calls for WT/WPL = 7, thenfrom Eq. 11, for 1-3-a = 0.42, WsWrr= 0.28. From Eq. 16, (p"T5.) =(p7a.)min (Ws/WT)max/(Ws/WT) = 68(.42)/0.28 = 102 w/kg.

Comparing this value with theoptimistic values projected3 forbatteries in the 1990-2000 timeframe, it appears that EV's employ-ing lead/acid and Zn/Cl2 batterieswill be unable to follow normal/urban commuter-car driving pat-terms in this century, and those •employing Ni/Fe batteries will bemarginally acceptable. Should itbe possible to develop the otherbatteries cited to a point wherethese forecasted electrical pro-perties are actually realized, thenthey would be adequate for urbancommuter car EV's of prudent (Wf/WPL = 7) design.

In the 1985-90 time frame, theonly batterie^ that have adequateforecasted (pTd.)'s are possibly:the Ni/Fe, 101.0 w/kg; Ni/Zn,135.3 w/kg; Li-Al/FeSx, 115.1 w/kg;and Na/S (g-alumina), 115.1 w/kg.But if the Ni/Zn battery, for ex-ample, cannot be developed to thepoint where many hundred deep dis-charges are possible, then it prob-ably will lead to EV annual expen-ses that are too high to be eco-nomically competitive with petrole-um-fueled vehicles; and thus itwould not be used in EV's in the1985-90 time frame. In addition,should the Ni/Fe peak-power pro-jection be too high by several per-cent, and the Li-Al/FeSx and Na/S($-alumina) projections prove toohigh by 15 percent or so, the EV'scould not follow normal urban com-muter-car driving patterns.

Therefore, considering themany technical uncertainties in-volved in the development of thesebatteries, it would be sagaciousto hedge against the possible eco-nomic/technical shortcomings of thebatteries that are actually devel-oped by using MES devices to supplyEV peak power requirements. Other-wise, it might not be possible to

have, on the road within the nextdecade, electrically energized ve-hicles that can follow normal ur-ban commuter car driving patterns.

THE OTPIMALLY-DESIGNED EV BATTERY:INFLUENCE OF MISSION SPECIFICATION

The optimally designed EV bat-tery has electrical characteris-tics that result in a value ofvehicle Wg/Wip which simultaneouslysatisfied both the mission drive-cycle and range requirements.

Eqs. 16 and 19, in order tobe satisfied simultaneously re-quire

•'min

Jo p t

<Rult>(20)

min

where [(p.d.)/RuitJopt is the ratioof electrical values required ofthe optimum battery. SubstitutingEqs. 17 and 18, yields,

!

(pTd.) max

I opt[Ev/WTd]

P 1R

(21)

where[(p.d.)/ns(es)p_0]Ql:)t is thevalue of the peak-power/energy-capacity ratio of the optimum bat-tery. Note the strong influenceof mission on the characteristicsrequired for a battery to be op-timum.

Conversely, given batterycharacteristics and a drive-cycle,it is possible to calculate avalue of R = Ro.b. that would makethese battery characteristics op-timum. Thus, from Eq. 21,

TEv/WTd] %

"S* s P=0

(22)

For example, if an EV were tofollow the EPA^urban cycle and hadnp = 0.75 and ^p = 0.8^ then(Pmax/Wij.)/[Ev/W>jid] np/TVp = (23)(0.75)/(0.09)(0.8) = 240 w/whr-km.For this situation, Eq. 22 becomes

21

240 (22a)of R.

For the batteries having fore-casted^ values of (p?5.) > 102 w/kg(and therefore, able topower prudently designed EV'sthrough the EPA urban cycle beforethe end of the century,) values ofRo.b. a r e listed in Table 2 (fol-lowing References). The forecast3 n r discharge capacity^ is used forris(es)P=0 in Table 2.

An EV that can follow normalcommuter car driving patterns, butthat has a mission R < Ro.b. willbe of power-limited design. Shouldbatteries of greater (]?Vcl.) thanthe values listed in Table 2 bedeveloped in the time frames indi-cated, then the WT/Wp]j and C/WPLpenalties associated with R < Ro.b.operation will be reduced. Butshguld future battery actual(p.d.)'s be equal or less than theforecasted values, then the onlyway to satisfy R < Ro.b. missionwith electrically-energized vehi-cles, while avoiding these penal-ties, is to employ MES devices.

Graphical Representation of theVariation of Ws/tftp with R for anUrban Commuter Car Using a Ni/Zn(1985-90) Battery. For these con-ditions, Ruit =

r|s(es)p=onp/[Ev/WTd] = (150)(0.75)/0.09 = 1250 km.And if 1-B-a = 0.37 = Ws/WT, thenWT/WPL -* «.

Assuming that for prudent de-sign, Wtp/WpL = 7, then from Eq. 11,Ws/W-r = 0.28, so the maximum R fora prudently designed vehicle is(0.28)(1250) = 350 km. Thus, thevalue 350 km limits R in Fig. 1.

The (pTd.) for prudent designwas found to be 102 w/kg. There-fore, to meet the power demand,WS/WT = (0.28)(102)/135.3 = 0.21.This is the level of WS/WT thatyields R = Ro b = 266 km (fromTable 1).

For R > Ro.b.' W T / W P L is givenby Eq. 11. This is the region ofrange-determined design. For R <Ro.b.' W T / W P L B 0.21, independently

260 3'50

R, km

Fig. 1 W /W vs RS T

The potential benefit thatwould accrue in Wg/Wy by substi-tuting an MES-device (Ni/Zn battery)vehicle for a Ni/Zn battery EV inthe region of R < Ro.b. can beestimated by comparing the gap be-tween Wg/WT = 0.21 and the projec-tion of Eq. 11 into this region ofpower determined design.

.. N1MUM THRESHOLD MES DEVICE CHAR-ACTERISTICS FOR BENEFICIAL APPLI-CATION TO RANGE-DETERMINED EV's

Various improvements in ve-hicle characteristics (i.e., de-crease in Wy/WpL and C/WPL) can beproduced by MES device addition.However, the benefits will accrueonly if the MES device has char-acteristics above a certain thresh-old level. The threshold MES de-vice characteristics for "break-even" vehicle performance areanalyzed here for a baseline EV ofrange determined design (i.e., R >

Assuming that a remains un-changed, which is a realistic as-sumption for the present level ofanalysis, then for both the base-line EV and the vehicle with theMES device to have the samefrom Eqs. 10 and 11,

1-e-a-R 1-e-a-lult <Rult>

m

(23)

where: (<5)R is the MES device ve-hicle-mass-fraction allowable ifthe Wf/WpL of both vehicles ofrange-determined design are to bethe same; and (Rult)m *-s *-he ulti-mate range of the vehicle with theMJ3S device.

Solving Eq. 23 for (6") R yields,

< 6 > R = Iult ""'

(24)

Since R/Ruit = Ws/Wj (from

Eq. 7) and Rult/(Rult>m = n / n m

from Eq. 6), for (es)p_g the samein both vehicles, where n m is de-fined as the energy efficiency ofthe vehicle with the MES device,then Eq. 24 can be written as,

1-3-cx

6 —WT

0 0.5 1.0

T\/T\

Fig. 2. Breakeven 6 vs

n/nm for range limited EV

Dividing Eq. 27 by Eq. 24yields,

«>» - %(25) (28)

Equation 25 is displayed in Fig. 2.

For both vehicles to have thesame value of C/WpL, from Eqs. 10,11 and 13,

[Ev/wTd]/nm

1-a-B-RRult

l-g-P-(6)R- R

m

(26)

where (<S)R is the MES device ve-hicle-mass-fraction required forequal C/WpL.

Solving for (<5)R yields,

(6) R = (1-e-a) [1 - 3. ]m

(27)

The result of (<5)R > (6) R de-rives from the fact that improve-ment in energy efficiency benefitsC/WpL more strongly than W T/W p L.Thus, for the same 1 m a largevehicle mass-fraction can be de-voted to the MES device and stillresult in break-even Wc/WpL.

Assuming that n > i), then:

(1) for 6 < (6")R- (W T/W p L) m < (WT/

(C/WpL)W p L) and (C/W p L) m

(2) for (I) R < 6 < (6)R; (WT/WpL)m

> (WT/WpL) and (C/W p L) m

(c/wpL)

Equation 27 is displayed in Fig. 2.(3) for 6 > (6)R; (W T/W p L) m > (WT/

W p L) and (C/W p L) m > (C/WpL)

23

Illustrative Example of MES DeviceVehicle-Mass-Fraction Required forBreak-Even Performance: BaselineEV of Range-Determined Design. As-sume that an EV with Ni/Zn (1985-1990) battery has a mission of 300km following an urban commuter cardrive-cycle. Therefore, 1-6-a =0.37 and Ruit = 1250 km. Wg/WT =300/1250 = 0.24. (See Fig. 1)

Should the MES device have anin/out efficiency such that, as aresult of regenerative braking andbattery peak-power-load-levelling,ri/nm = 1/1.3 (which is reason-able3'6 ) then from Eq. 23, ( 6 ) R =(0.24) [1 - 0.771 = 0.055, and fromEq. 27, (6) R = (0.37) [1 - 0.77] =0.085.

This calculation shows that,for the same mission, a vehicleincorporating an MES device willhave a lower Wy than that of thebaseline EV if the device can pro-duce r\/r\t=lA-.'i while having a massless than 5.5% of W T. And the MES-device vehicle will consume lessenergy than the baseline EV if itsmass is less than 8.5% of W T.

Illustrative Example of Best Possi-ble Performance of Vehicle WithMES Device: Baseline EV of Range-Determined Design. For the situa-tion considered in above section,assuming that the gravimetric en-ergy and power content of the MESdevice is so great that 6 •* 0, thenin the limit of 6=0 Eqs. 10 and11 can be used to calculate thecorresponding (Wf/WpiJm,6=0 a n d

(C/WPL)m,6=o-

Since Rult = 1250 km; (Rult)m=

(1.3) (1250) = 1625 km, soR^(Rult)m=300/1625 = 0.185. From Eq. 10,(WT/WpL)m,6=0 = 1/(0.37 - 0.185) =5.4 so a (7 - 5.4)/7 x 100 = 22.8%reduction in W T is ideally possi-ble for the conditions assumedhere.

From Eqs. 10, 11 and 13

R

1 [0.37-0.24]1.3 [0.37-0.185] 0.54

(C),[1-g-a- ]

V [l-n-ct-rf-

where C m is defined as the energyconsumption of the vehicle con-taining the MES device. This re-sult can be represented as[1 - (C)m/C]100 = 46%. That is,46% reduction of energy consump-tion is possible for the condi-tions assumed here.

For actual MES devices having0 < 6 < 0.055, W T savings will de-crease and approach 0 as S ap-proaches 0.055. If 0.055 < 6 <0.085, the vehicle containing theMES device will have a greater massthan the baseline EV, but its en-ergy consumption will be belowthat of the baseline EV. Finally,MES devices requiring <5 > 0.085 inorder to produce nm/n = 1.3, can-not be beneficially employed to re-duce vehicle energy consumptionfor the situation considered inthis example.

RESULTS AND FUTURE WORK

Analytical energy and powerbalances were used to produce equa-tions describing vehicle mass andenergy consumption as a functionof payload and mission specifica-tion. They form the basis forevaluating the impact on perform-ance produced by adding MES de-vices to vehicles.

For EV's, it was found thatno battery forecast for develop-ment in this century had a peak-power capability sufficient for anurban/freeway commuter car of pru-dent design. Only by incorpora-ting an MES device could EV's per-form this mission with these bat-teries.

For missions that can be per-formed with EV's containing theforecast future batteries, (e.g.,an urban commuter car mission,) theimpact of MES device addition wasfound to depend on the baselineEV design. A criterion for theoptimum EV battery as a functionof mission specification was evol-ved. Should an EV be used in amission with a range less than

24

that derived from the criterion,it is of power determined design;if the range is greater, the de-sign is range-determined.

Time permitted only the range-determined case to be analyzed.The threshold MES device character-istics for breakeven vehicle per-formance was established as afunction vehicle energy efficiencyincrement. The impact produced inthe limiting case of MES devicegravimetric power and energy con-tent approaching °° was examined.

Future work includes a break-even analysis for EV's of powerdetermined design and use of the re-sults to examine the impact ofactual MES devices on both powerdetermined and range determineddesigns, and application of themethodology to establish the im-pact of MES-device addition topetroleum-fueled vehicles.

REFERENCES

1) Department of Energy, "Per-formance Standards for Demonstra-tion: Development of Energy'sElectric and Hybrid Veyicle Re-search, Development and Demonstra-tion Project,: Federal Register,Vol. 43, No." 104, May 30, 1978.

2) McAlevy, R.F. Ill and Bedrosyan,L., "A Critical Review and Evalu-ation of Published Electric-Ve-hicle Performance Data," Proceed-ings of the 13th Intersociety En-ergy Conversion Engineering Con-ference, Aug. 20-25, 1978, SanDiego, CA, pp. 655-661, SAE P-75,SAE Inc., Warrendale, PA, August1978.

3) Behrin, E. et al., Energy Stor-age Systems for Automotive Propul-sion, Vols. I and II, UCRL-52303,Lawrence Livermore Laboratory,University of California, Decem-ber 15, 1977.

4} McAlevy, R.F. Ill, "A Funda-mental Basis for Evaluating thePerformance of Electric (and OtherEnergy-Storage) Automotive Vehi-cles and its Use in Energy PolicyAnalysis," Proceedings of theNATO/CCMS Fourth InternationalSymposium on Automotive Propulsion

Systems, Washington, D.C., April17-22, 1977.

5) McAlevy, R.F. Ill, "Optimum De-sign of Automotive Vehicles Employ-ing Alternative Energy Sources ofLow Energy Density: Impact onSelection of an Energy Carrier forFuture Urban Vehicle Transporta-tion Systems," Proceedings of theInternational Conference on Alter-nate Energy Sources, Miami Beach,Florida, Dec. 5-7, 1977.

6) Sandberg, J.J. and Leschly, L.,"User Experience with On-RoadElectric Vehicles in the U.S.A.and Canada," Proceedings of the13th Intersociety Energy Conver-sion Engineering Conference: Aug.20-25, 1978, San Diego, CA, pp.644-654, SAE P-75, SAE Inc., War-rendale, PA, August, 1978.

7) Liles, A.W. and Fetterman,G.P., Jr., "Selection of DrivingCycles for Electric Vehicles inthe 1990's," Eleventh IntersocietyEnergy Conversion Engineering Con-ference Proceedings, Vol. II,Stateline, Nevada, Sept. 12-17,1976.

8) Private Communication, E.Behrin, May 23, 1978.

9) O'Day, J., et al., "A Projec-tion of the Effects of ElectricVehicles on Highway Accident Sta-tistics, " SAE Paper #780158, pre-sented at SAE Congress, Detroit,Michigan, Feb. 27-March 3, 1978.

25

Table 1. Typical drive-cycle characteristics

DRIVE-CYCLE

SAEJ227a/CSAEJ227a/DEPA Urban (noScott

freeway)

0.090.10.090.12

13202332

Table 2. Forecasted (Ref. 3) Battery Electrical Characteristics andEV Range Required of Urban Commuter Car for Batteries to be Optimum

Battery (p.d.),£j-

Ni/Fe (1985-1990)Ni/Fe (1990-2000)Ni/Zn (1985-1990)Ni/Zn (1990-2000)Li-Al/FeSx

(1985-1990)Li-Al/FeSx

(1990-2000)Na/S (g-alum):

(1985-1990)Na/S (g-alum):

(1990-2000)Na/S (glass):

(1990-2000)Li/air: (1S90-200O)

101101.8135.3139.9

115.1

130

115.1

130

160119

nS(es>

5998150153

100

110

82

123

150357

whrP=0' kg

.4a

.5

.5

.2

.1

(pTa.) wns ( es )p=o' w h r

1.71.0.9

0.91

1.2

1.2

1.4

1.1

1.10.33

R . ,kmo.b.

141240266263

200

200

171

218

218727

aCap"acity at 3 hr discharge

26

PROJECT SUMMARY

Project Title: Advanced Flywheel Energy Storage Unit for a High PowerEnergy Source for Vehicular Use

Principal Investigator:

Organization:

Arthur E. Raynard

Project Goals:

Project Status:

Contract Number:

Contract Period:

Funding Level:

Funding Source:

Garrett-Ai Research2525 w. 190th StreetTorrance, CA 90509(213) 323-9500

The project goal is to determine the benefits of a light-weight,hermetically-sealed energy storage unit for vehicular applications.

The tradeoff study and detail design are completed. The testequipment and the test specimen are being fabricated for developmentand performance testing. The testing will verify functional compat-ibility and measure both parasitic losses and input/output losses.Testing will start in December 1978.

University of California Purchase Order 9676603

May 1, 1978 - April 30, 1979

$470,182

Lawrence Livermore Laboratory

27

ADVANCED FLYWHEEL ENERGY STORAGE UNIT FOR AHIGH POWER ENERGY SOURCE FOR VEHICULAR USE

Arthur E. RaynardAiResearch Manufacturing Company of California

A Division of The Garrett Corporation2525 West 190th Street

Torranee, California 90509

ABSTRACT

The maximum benefits that may be gained by the incorporation of mechanical energystorage (MES) into vehicular propulsion systems are obtained by combining the conceptsof load-leveling the prime energy source(s), and recovering vehicle kinetic energy byregeneration. These benefits are intuitively achievable and large, and have beenindirectly demonstrated in past programs. There is a pressing naed for the verifica-tion of these benefits in a structured program designed for that purpose. The verifica-tion, or lack thereof, is needed to give direction and emphasis to future MES componentand system development. This paper describes an experimental program that characterizesthe magnitude of the benefits in a flywheel propulsion system. The first step in theprogram will be to experimentally determine the performance of a hermetically-sealedflywheel system that has been optimized for an electric or hybrid vehicle. The flywheelsystem has been designed and the design tradeoffs will be described.

INTRODUCTION

The need to reduce dependence onpetroleum sources for energy generationhas created a substantial interest in theinvestigation and development of energystorage devices, ihe flywheel energystorage unit can provide substantial bene-fits to transportation propulsion systems.The flywheel can supply high power demandsand thereby provide a method for loadleveling the primary energy supply. Thisenergy supply can be an electromechanicalbattery or a liquid-fueled heat enginepower converter. In addition, the fly-wheel can accept vehicle kinetic energyduring braking (regeneration) at a ratelimited only by the transmission powercapability. This method of vehicle energystorage can also be applied to third rail,electrically-powered vehicles, or station-charged electrical Iy-powered vehicles.

The one-year program that isdescribed in this paper will involvedesign, fabrication, and experimentaldetermination of the performance of anadvanced, hermetically-sealed energystorage unit that has been sized for atypical 3000-lb curb weight vehicle.

PROJECT GOALS

The principal goal is to providedecision-making information regarding thebenefits of a mechanical energy storagedevice as it applies to vehicle fuel con-sumption or vehicle range. Those benefitsmust be of sufficient magnitude to reducethe life-cycle cost of the flywheel pro-pulsion system below a conventional pro-pulsion system. The expected result ofthe development program is to stimulateindustry to market such a s/stem andthereby provide a method to reduce petro-leum dependence within the U.S.

BACKGROUND INFORMATION

A substantial number of programs haveinvestigated flywheel systems for the pur-pose of load-1 eveI ing the energy, or powersource, and using vehicular kinetic energyrecovery to improve the system propulsiveefficiencies. Table 1 lists the programsthat Garrett-AjResearch has managed orcontributed to that feature flywheeldevices. A demonstrated savings of 33percent was accomplished as a result ofbraking regeneration only in a six-monthtest program in The New York City subway

28

Table 1. Summary of load levelingdemon strat i on s.

PROGRAM

1. ENERGY STORAGE CAR

2. ADVANCED CONCEPT TRAIN

3. UNIVERSITY OF WISCONSIN

4. HEAT ENGINE/FLYWHEELELECTRIC TRANSMISSION STUDY

5. FLYWHEEL BUS STUOV

7. BATTERY/FLYWHEEL CAR

8. HERMETICALLY SEALEDENERGY STORAGE UNIT

SUBWAY

SUBWAY

PASSENGER CAR

PASSENGER CAR

URBAN BUS

POSTAL VAN

NEAR TERMELECTRIC VEHICLE

VEHICLEPROPUL5ION

RESULTS33% ELECTRIC ENERGYSAVINGS. NY. TEST

DEVELOPMENT TARGET33% SAVINGS

33% FUEL SAVINGS.F.U.O.C., OYNO TEST

CALCULATED «5% FUELSAVINGS, F.U.O.C.

CALCULATED, 36S.FUEL SAVINGSDEVELOPMENT TARGET.140H CYCLIC RANGE

DEVELOPMENT TARGET.140% CVCUC RANGE

DEVELOPMENT TARGET.66% FUEL SAVINGS

system. Two subway cars were includedin the test. These cars were in revenueservice in 10-car trains. Studies haveshown that it is possible to save as muchas 65 percent of the fuel normally usedin a passenger vehicle operating in a citydriving mode. Other methods of improvingfuel economy are being developed, butthese other methods are complementary tothe flywheel system and do not replacethe benefits associated with the use of aflywheel.

The flywheel benefits, however, canbecome insignificant if the weight, theparasitic losses, and input/output effi-ciencies are non-optimum. This programconcentrates on t'nese factors.

PROGRAM SCHEDULE SUMMARY

The program events are summarized inFig. 1. The design tradeoffs (Task 1) arecomplete and the results reported herein.The hardware design (Task 2) is also com-plete and the hardware has been ordered.The results of Task 2 are summarizedherein.

MAY 1. 1»7» '

TASK 1

DESIGN TRADEOFFS

STUDVREPORT.

TASK 2

HAROWARE DESIGN - - - .

TASK 3

HARDWARE FABRICATION .

TASK 4

TESTING AND EVALUATION..

TEST REPORT.

TASKS

PROGRAM MANAGEMENT

MONTHLY PROGRESS REPORT.

QUARTERLY ORAL

PRESENTATION. _

MONTHS AFTER START OF PROGRAM

197t

'M

BBl

j

m

—

J A

"Iurn

u

S

B

l |

o

BBl

N

BBl

I]

D

BBl

t»79J F M

\]

A

u

Fig. 1. Phase I program summary.

TRADEOFF STUDY SUMMARY

The purpose of the tradeoff study(Task 1) was to develop the parametric

design information to allow a state-of-the-art energy storage unit (ESU) to beevolved. The program goal is to providean ESU with an overalI energy density of3.0 w-hr/lb. This goal is to be achievedfor the indicated energy and powerrequirements. There were five main areasthat required investigation in order todefine the ESU:

(a) Flywheel tradeoffs

(b) Electrical input/output machinetradeoffs

(c) Bearing tradeoffs

(d) Vacuum tradeoffs

(e) Lubrication tradeoffs

The parameters investigated are indi-cated in Fig. 2.

• WEIGHT• RPM• MATERIAL• CONTAINMENT

L ENERGY STORAGE UNIT(ESUI

- • WEIGHT•RPM• EFFICIENCY CHARACTERISTIC

BEARING TRADEOFFS [ /

•RPM•LOS5E5• SIZE• MATERIALS

LUBRICATION TRAOEOFFS

VACUUM TRADEOFFS* MOLECULAR PUMP

PRESSURE VERSUSFORE CHAMBERPRESSURE

Fig. 2. Tradeoff study summary.

The flywheel tradeoffs consideredweight, speed, material, and containment.Two rotor rim materials were included,Kevlar and S-glass. The tip speed limitswere 2500 fps and 2200 fps, respectively.

The weight shown in Fig. 3 includesrotor, containment, molecular pump, and*lywheel housing. This weight is plottedagainst flywheel diameter and reaches aminimum at 42,000 rpm for the S-glass12-inch diameter configuration. However,Kevlar was chosen as the rim materialbecause the Kevlar rotor design has beendeveloped and can be geometrically sealedfor this application. The weight differ-ence between the two materials is insig-nificant and the program risk is minimizedby utilizing an existing design concept.

ELECTRICAL INPUT/OUTPUT MACHINE TRADEOFFS

Five types of machines were paramet-rically designed in terms of weight, speed,range, and part-power efficiencies. Theyare identified in Table 2. Each machine

29

FLYWHEELASSEMBLYWEIGHT,

LBS

Table 3. Bearing tradeoffs.

0 10 12 14 16 18 20 22 24

FLYWHEEL RIM DIAMETER, INCHES

Fig. 3. Flywheel tradeoffs.

Table 2. Electrical input/outputmachine tradeoffs.

MACHINETYPE

SALIENT POLESYNCHRONOUS

ROUND ROTORSYNCHRONOUS

INDUCTION

H0M0P0LARINDUCTOR

IRONLESSPERMANENTMAGNET

BESTWEIGHT

LB

28.5

23

16

25

NO SOLUTION

1C0% SPEEDRPM

25.000

36.000

42.000

60.000

14,000

CHARACTERISTICEFFICIENCY

%

89

83

92

89

-

best-weight, which corresponds to itsweight at the indicated speed, is Iistedalong with its characteristic efficiency.The induction machine, in combinationwith the flywheel, provides the optimumESU from a weight, size, and efficiencyStandpoint, and therefore was selected asthe ESU input/output machine.

BEARING TRADEOFFS

Bearings were investigated in termsof losses, sizes, and materials. Thedesign loads criteria were imposed alongwith the speed that was determined fromthe electrical machine and flywheel opti-mization. The study results are summar-ized in Table 3.

A 202-size-bear ing made from 52100steel was selected for the lower bearing.A 204-size-bearing made from M50 CEVMsteel was selected for the upper bearing^.The M50 material yields a 17,975 hr BtOlife. The total predicted maximum bearingtoss is 188 w.

bEARlMGLOCATION

LOWER

UPPER

BEARING

ZOO201

m204

302303304

BEARINGBOREMM

to2b7

•0bi

20

POWERLOSS

WATTS

303BS376

13590

120163

BIOLIFE

HOURS

6.6809.770

17.S3O"5.000

I7.9JU*

*.no_

22.450

•MAT I M50C£VM

VACUUM SYSTEM DESIGN

The ESU has a duaI chamber vacuumsystem during operation. The flywheelchamber is run at a pressure between 1.4and 50 microns while the electricalmachine chamber is at approximately 1000microns. The low pressure is maintainedin the flywheel cavity by a molecular pumpwith a rotor diameter of 6 in. The designcharacteristics are summarized in Fig. 4.This design provides a pressure cat'\o of700/1 at 100 percent speed and 20/1 at 50percent speed. The energy storage unitcrossection is shown in Fig. 5.

Fig. 4. Vacuum system design.

Fig. 5. Energy storage unitcross-section.

30

LUBRICATION TRADEOFFS

Five candidate oils are identifiedin Table 4. As of this date, the Fyrquel150 appears to be the best candidate oil.Material compatibility tests are currentlytaking place. Results will be, discussedin the next reporting period.

Table 4. Lubrication tradeoffs.

CANDIDA TES

MILL 23699IMOBIL.STD. OILPLUS OTHERS!

COhAV. GRADE 32EXXON

TERESSOJ79EXXON

BRAYCOB15ZBRAr O I L c o

FYRQUEL 150'STAUFFERCHEMICAL CO.

TVPE

SYNTHETICBASE

MINERALBASENAPTHENIC

PERFLUORIDATEDPOIV-ETHER

TR1-ARYLPHOSPHATE

VISCOSITYCENT'STOKES

5 ,

46

40

45

VAPORPRESSURE

TORR021D°F

2

2X1»3

5 X ID 3

1 XIO8

3.5X102• 380OF

LUBRICITY

EXCELLENT

GOOD

EXCELLENT

GOOD

CORROSIONSTABILITY

EXCELLENT

POOROXIDATIONRESISTANCE

EXCELLENT

GOOD

FLYWHEEL ROTOR CONSTRUCTION

The flywheel rim is a filament-wound composite material consisting ofconcentric outer rings wound of Kevlarroving with an epoxy binder, and one innerring wound using S-glass roving with anepoxy binder. The Kevlar rings have ahigher strenqth-to-density ratio and aresuperior for enerqy storage and fatiguelife. The inner ring of S-glass providesfor a dead loading of the outer rings andfor higher compressive strength in thearea of contact with the hub spokes.

The rim is mounted on a 4-spokedhub, machined from 7075-T7351 aluminumplate stock. There is a short protrusionon one side of the hub to receive thebolts that mount the flywheel to theshaft.

CONTAINMENT/VACUUM CHAMBER

The flywheel is mounted in an air-tight enclosure to provide for a vacuumenvironment surrounding the flywheel. ThewalIs of this enclosure are of steel andare of sufficient thickness to providefor containment of the flywheel in theunlikely event of a burst failure.

MOTOR/GENERATOR CONSTRUCTION

The motor/generator is of the squir-rel cage induction type. The rotor con-sists of a stack of silicon steel lamina-tions pressed onto a magnetic steel shaft.The rotor laminations are of specialdesign to minimize harmonic losses and

rotor heating caused by the waveform ofthe inverter. The current conductors andend rings are cast copper alloy with aconductivity related to pure copper of 60percent. A flow of oil is pumped throughthe inner diameter of the hollow rotorshaft to provide cooling. The rotor hasa maximum peripheral speed of 420 fps,which is consistent with other high speedrotor designs.

The stator consists of a stack ofsilicon steel laminations punched toreceive the stator windings. The statorinsulation system wilI be class 220°C andwill have HML-coated magnet wire andNomax-Kapton-Nomax laminated slot-linerinsulation. The stator will be vacuumimpregnated with ML varnish. The leadwires will have double Teflon insulation.

The stator 0D wiI I be ground for ashrink fit into the motor housing. Themotor housing is machined from aluminumalloy and contains oil passages aroundthe motor stator to provide for statorcool ing.

VACUUM SYSTEM

The housing for tie entire motor,flywheel, and pump is evacuated toapproximately one rorr (one millimeter ofmercury). In the unit shown, which is thelaboratory demonstrator, this process mayhave to be repeated occasionally becauseof the 0-ring seals in the housing. Asecond evacuation may be necessary tofacilitate disassembly for developmentchanges. The flywheel cavity is broughtto an even lower pressure of approximately0.001 torr by means of a molecular pump.This device, with its rotor adjacent tothe flywheel hub, propels molecules up aspiral path in the stator plate, therebycausing the molecules to move from theflywheel housing to the motor housing.No rubbing seals are required.

LUBRICAT 10N/C00LING SYSTEM

Oil is circulated through the motorcooling passages, the bearings, and theoil cooler by means of a centrifugalpump. The pump is gear driven from tnelower end of the flywheel shaft. Oil ispicked up from the oil sump fn the lowerend of the unit and discharged under apressure of approximately 10 psi intopassages in the aluminum housing. Fromhere it passes out through the oil coolerand back into the motor housing. One

31

passage leads through an external tube tothe jet in the oil sump cover. This jetshoots oil into the center hole in therotating flywheel shaft. The high speedof rotation causes the oil to flow on theID of the central hole and move to areasof greater radius. This oil is thendrained bacx into the oil sump throughthe outer annul us of the rotor shaft.

After the oi I reenters the motorhousing from the oil cooler, it entersthe motor stator cooling passages. Aftercooling the motor stator, the oiI isreturned to the sump by gravity. Thesplit of the oil flow into the two possi-ble paths is controlled by an orifice inthe housing that can be adjusted duringthe development program.

BEARING ARRANGEMENT

The fIywheeI rotor i s cant iI ever-mounted on the end of the motor shaft.A 204-size ball bearing is located at thelower end of the shaft. The bearing innerraces and other elements of the rotatingassembly are clamped tightly axially forincreased shaft cross section to providethe required stiffness. The upper bearingis resilient Iy mounted to control therange of critical speeds, reduce noise,and to increase bearing life. The lowerbearing has an axial preload spring withforce that is also transmitted to theupper bearing. Both bearings are shaft-cooled and mist-lubricated.

ESU INPUT/OUTPUT MACHINE PARAMETRICPERFORMANCE

For trade-off consideration, some ofthe machine parameters are plotted in Fig.6. The left-hand curve shows a plot ofmachine efficiency versus output power.Slip values are shown for the correspond-ing power outputs. Both the high-speedcase (in solid line) and the low-speedcase (in dotted line) are presented.

15 30 45POWER OUTPUT (KW)

15 30 45POWER OUTPUT |KW)

FUTURE WORK

The completion of the current contractis based on the following objectives:

• Prove composite rotor inoperating environment

• Design integration of allelements

• Determine parasitic losses

• Determine I/O losses

A test program will measure theenergy storage unit performance anddetermine the compatibility of the com-ponents with the cooling and lubricationsystem. The effects of the vacuum levelwill be observed and measured. Afterthese tests are completed, a series oflimited environmental tests will evaluatethe energy storage unit in terms of avehicle environment and operational cycle.

Follow through work to be consideredwill investigate the total propulsionsystem. This work will determine andoptimize the benefits associated with theflywheel concept. The work must be sys-tematic since the results will be stronglyinfluenced by the intended application.

A second element of work to be con-sidered will determine the added weightand cost of the total flywheel propulsionsystem.

HIGH SPEEDLOWSPEEO

Fig. 6. ESU I/O machine parametricperformance.

32

PROJECT SUMMARY

Project Title: Regnerative Flywheel Energy Storage System

Principal Investigator: E. L. Lustenader

General Electric CompanyCorporate Research and DevelopmentP. 0. Box 43Schenectady, NY(518) 385-3084

Laboratory test an improved flywheel energy recovery system sizedfor a 3000 pound class battery/flywheel electric vehicle. Teststo simulate an electric vehicle operating under the SAE J227aSchedule D driving cycle will establish the range improvementsattributed to the flywheel.

The design and detail drawings are complete, and componentmanufacturing is underway. The flywheel drive motor/alternatorwill be a 20,000 rpm six-pole synchronous inductor type machinepowered by an 8 SCR loan commutated inverter. The motor/alter-nator is coupled directly to a small steel disc flywheel designedto recover vehicle braking energy. The laboratory set up includesa 109 volt lead-acid battery bank, a new separately excited DCelectric vehicle propulsion motor, the flywheel energy storagesystem (motor/flywheel, PCU and control) and a load flywheel tosimulate vehicle inertia. Facility modification is underway.

Contract Number: University of California Purchase Order 8990503

Contract Period: Mar. 1978 - Mar. 1979

Organization:

Project Goals:

Project Status:

Funding Level:

Funding Source:

$450,905

Lawrence Livermore Laboratory

33

REGENERATIVE FLYWHEEL ENERGY STORAGE SYSTEM

E.L. LustenaderGeneral Electric Company

Corporate Research and DevelopmentP.O. Box 43

Schenectady, New York 12301

ABSTRACT

This paper describes the progress to date on the laboratory development and evalu-ation of a regenerative flywheel energy storage system. The system has been designedspecifically for a battery/flywheel electric vehicle in the 3000 pound class. Plannedlaboratory tests will simulate this electric vehicle operating over the SAE J227a ScheduleD driving cycle. The range improvement attributed to the use of the flywheel will beestablished. The flywheel energy storage system will consist of a solid rotor, syn-chronous inductor-type flywheel drive machine electrically coupled to a DC battery elec-tric propulsion system through a load commutated inverter. The motor/alternator unit iscoupled mechanically to a small steel flywheel which regenerates the vehicle's brakingenergy. The laboratory simulation of the battery/flvwheel propulsion system willinclude a 108 volt lead-acid battery bank, a separately excited DC propulsion motorcoupled to a flywheel which simulates the vehicle's inertia, and the flywheel energystorage system comprised of the motor/flywheel unit, the load commutated inverter and itscontrol.

INTRODUCTION AND BACKGROUND

In 1974, a flywheel energy storagesystem was conceived by the General ElectricCompany and proposed for demonstration tothe Department of Energy, Office of EnergyTechnology, Division of Energy Storage Sys-tems. The overall objective of the programwas to demonstrate new technology associatedwith a flywheel energy storage system con-sisting of a composite flywheel directlycoupled to an AC synchronous motor/alter-nator. The motor/alternator/flywheel unitwas hermetically sealed with the rotatingassembly operating in a low pressure heliumatmosphere. The motor/alternator receivedits power from a solid state inverter/recti-fier unit designed to provide the necessaryfrequency control from a constant DC batterypower source.

The objective of the DOE program wasto demonstrate the following new technology:

• Solid Rotor, inductor type synchronousmotor/alternator;

• Direct coupled composite flywheel ofnew "cross-ply" construction;

• Sealed rotor assembly operating in lowpressure helium;

• Load commutated inverter power supplyto couple the flywheel energy packageto a DC input/output power source;

• Novel force commutated circuit forstarting the synchronous machine fromstandstill;

• Novel control system not requiring theuse of rotor shaft position sensors.

This new technology was successfullydemonstrated on DOE Contract EY-76-C-02-4010.

A follow-on to the program was initi-ated on March 17, 1978. The new programhas the specific objective of laboratorytesting an improved flywheel energy recov-ery system sized for a battery/flywheelhybrid electric vehicle in the 3000 poundclass. Laboratory tests will be conductedto simulate the electric vehicle operatingunder the SAE J227a Schedule D drivingcycle. The objective is to simulate thevehicle operating over this duty cycle andto determine the range improvement that canbe attributed to the use of a small fly-wheel in combination with the battery bank.

Analysis has indicated that the fly-wheel/battery system can isolate the batteryfrom the acceleration power demands of thevehicle and can also recover a substantialportion of the braking energy. Thus, therange of the flywheel/battery electricvehicle is projected to be greater than thatof an all-battery electric car when used ina repetitive start-stop driving cycle.

SCOPE OF CURRENT PROGRAM OVERALL VEHICLE SYSTEM STUDIES

In achieving the above objective, animproved flywheel energy recovery systemhas been designed and is being fabricated.There are nine major technical tasks to beaccomplished. The scope of the overallprogram can be summarized as follows:

• Establish specifications for a regen-erative flywheel energy storage systemand prepare a test plan. These speci-fications would be based on tradeoffstudies between system weight, compo-nent efficiency, and performance.

• Design a new, improved, all steel,inductor motor/flywheel energy storagepackage for a 3000 pound class electricvehicle. Test results from the firstphase of the DOE program would be uti-lized in arriving at this design.

• Test the regenerative flywheel energystorage system in the laboratory. Inorder to simulate an actual drivingcycle, a load flywheel would be de-signed to simulate the inertia of the3000 pound vehicle. This would becoupled directly to a new separatelyexcited DC propulsion motor.

The fourteen SCR inverter/rectifiercircuit which was demonstrated on the firstcontract was completely redesigned andreplaced by a new 8 SCR unit with power con-tactors. Field reversal in the separatelyexcited propulsion motor obviates the needfor the six additional SCR's originallyproposed for braking.

The simulated propulsion system will consistof a 108 Volt lead-acid battery bank, a DCpropulsion motor with a load flywheel, andthe regenerative flywheel energy storagesystem. Tests will be conducted on theequipment to simulate the vehicle in oper-ation. The equipment will be operated toestablish the performance of the regenera-tive flywheel energy storage system as ifit were performing in a 3000 pound elec-tric vehicle. It will be operated oversimulated driving cycles, and measurementswill be made to determine the energy re-quired per cycle. Data will be reduced andthe results analyzed and compared to thepredicted performance. Results will pro-vide an estimated range and power consump-tion for a 3000 pound class electric vehiclewith this type of flywhee1/battery propul-sion package.

In analyzing the performance of aflywheel/battery powered vehicle operatingon a duty cycle such as the SAE J227a Sched-ule D, a multitude of operating nodes canbe assumed. On this contract, five modesof operation were selected for consideration.These operating modes are illustrated inFig. 1.

In Mode 1 all of the braking energystored in the flywheel is used to supplythe drive motor armature power during theinitial stage of acceleration. This resultsin power not being required from the bat-tery by the motor armature until some pointbeyond the "cornering point." "CorneringPoint" is defined as the vehicle speed atwhich the back EMF of the propulsion motorequals the battery voltage. This elimin-ates the need for an armature chopper inthe battery circuit, thus resulting in acost and weight saving as well as elimina-ting the losses ssociated with the chopper.In this mode of operation, as well as in

MOOES

Note: O denotes "cornering point"

Fig. 1. Power-Time Profiles for VariousDriving Cycles

35

other modes studied/ it is assumed that thebatteries will, at all times, supply thedrive motor field power and all other aux-iliary power requirements.

Mode 2 is similar to Mode 1 exceptthat once the cornering point is reached,the remaining available flywheel power isused uniformly over the remainder of theaccelerating period with the additionalrequired armature power being supplied bythe battery. In this way '.he need for anarmature chopper is still elminated but thepeak battery current is reduced with a re-sulting improvement in battery output.This is accomplished, however, at the ex-pense of added control complexity.

In Mode 3 all of the braking energystored in the flywheel is used to reducethe peak battery current. In this case,an armature chopper is required in the bat-tery circuit since battery power is usedprior to reaching the cornering point.However, since peak battery current islower, total battery output will increase.

In the last two modes of operation.Mode 4 and Mode 5, load levelling is used.This is accomplished by using the batteryto charge the flywheel during the idle and/or cruise segments of the cycle. Thisresults in minimizing the peak battery cur-rent. The tradeoff in this case, however,is between the improved battery output dueto lower peak current and the losses asso-ciated with charging the flywheel from thebattery. With the exception of the loadlevelling feature, Mode 4 is similar toMode 2, and Mode 5 is similar to Mode 3.

PERFORMANCE SIMULATION

In order to select the optimum mode ofoperation and to carry out tradeoff studiesbetween vehicle weight, battery weight,flywheel weight, flywheel in/out efficiency,etc., computer simulation runs were made.To accomplish this, four digital simulationprograms were set up; each one was derivedby making extensions to an existing computerprogram currently being used for designoptimization performance predictions of theelectric vehicle being developed by theGeneral Electric Company under DOE Contract.The modified programs are flexible enoughto accomodate a variety of vehicle para-meters including:

• Vehicle gross weight.

• Wheel rolling radius,

• Final drive ratio,

• Aerodynamic drag coefficient;

• Vehicle frontal area,

• Number of transmission speed ranges(gears),

• Speed ratio for each gear,

• Rolling resistance drag coefficients,

• Wheel inertia,

• Transmission/final drive efficiency,

• Motor design parameters (as definedby the motor model),

• Battery design parameters (as definedby the battery model),

• Electrical losses in drive train (asdefined by field and/or armaturechopper models),

• Auxiliary power losses (includingaccessories and ventilating blowers).

COMPUTER PROGRAM MODIFICATIONSFOR THE FLYWHEEL/BATTERY VEHICLE

Figure 2 shows schematically the over-all system modelled for computer evaluation.The flywheel package was considered as asingle subsystem, containing the motor/alternator/flywheel and power conditioningequipment. As such, only one parameter,the combined in/out efficiency, was requiredto define this subsystem. This value becamean input to the program. It was assumedthat the size and speed range of the fly-wheel would be designed to be sufficient tostore the braking energy.

RESULTS OF THE VEHICLE SIMULATION

The battery/flywheel vehicle assumedfor these simulation runs was taken as amodification of the all-battery electricvehicle which General Electric is develop-ing under a DOE/JPL contract. The vehiclehas an empty weight of 2000 lbs, a batteryweight of 1,100 lbs, and a passenger load of600 lbs for a gross total weight of 3700 lbs.

Load Leveling C M M Only - Mode* 4 and 5

Fig . Overall System Model

inertl*

AtroDrao

36

In assessing the results of the vehiclesimulation runs, two criteria were used:First, the actual range for a repetitiveJ227a Schedule D cycle, and second, howthis range canpares with that computed forthe pure battery electric vehicle. In thelatter case, no regeneration into the bat-tery was assumed for a referenced vehiclesince the amount of regeneration possibleis not precisely known at this time.

Originally, it was proposed that thetotal weight of the vehicle be kept con-stant even though a flywheel energy storagepackage may weigh more than the choppercontrol system used in a conventional bat-tery electric vehicle. The constant vehicleweight would be accomplished by removingone or more of the propulsion batteries.Computer runs for constant vehicle weightwere made, but runs were also made in whichthe vehicle gross weight was allowed toincrease by the amount by which the fly-wheel package weight exceeded the weight ofthe conventional propulsion system. Inthese latter runs, the comparison with thepure battery electric vehicle was made onthe basis of equal gross weight with theincreased weight of the battery vehicle_being made up of additional propulsionbatteries.

A typical set of results is shown inFigs. 3 and 4. These computations arebased on the Mode 3 type of operation indi-cated in Fig. 1. Since the flywheel energystorage system had not yet been defined,its weight was unknown. Therefore, com-puter runs were made with the assumptionthat the flywheel propulsion system wouldweigh betweeen 0 and 200 pounds more thanthe pure battery (chopper controlled) pro-pulsion system.

Calculations showed that for any fixedweight increase due to the flywheel packagedifferential the vehicle performance wasslightly better if the battery weight werekept constant as opposed to the gross weightof the vehicle being kept constant. If thevehicle gross weight were held constant, aslightly higher flywheel in/out efficiencywas required in order to break even with apure battery vehicle. The same conclusionwas found for other modes of operation.

, The shortcoming of Mode 3 and Mode 5operation is that an armature chopper orsome other means of controlling the DC volt-age to the traction motor is required.Realistically, this is also a shortcomingof Mode 1 and Mode 2 operation since littleenergy may be left in the flywheel after the

100

90

SO

40

J227a Schedule D Driving CycleVehicle Gross Weight Constant

at 3700 Pounds

MODE 3 OperationJ227a Schedule O Driving Cycle

Battery Weight Constantat 1100 Pounds

3700 Ib Vehicle (0 wt flywheel)3800 Ib Vehicle (100 Ib flywheel)

3900 Ib Vehicle (200 Ib flywheel)

0 20 40 60 80 100

In-Out Efficiency of Flywheel Energy Storage Package (%)

Fig. 3. Effect on Vehicle Range forConstant Vehicle Weight

0 20 40 60 80 100In-Out Efficiency of Flywheel Energy Storage Package (%)

Fig. 4. Effect on Vehicle Range forConstant Battery Weight

37

cornering point is reached even under idealconditions. For lower flywheel packageefficiencies or for interrupted drivingcycles, there may be insufficient flywheelenergy to reach the cornering point. Mode4, however, avoids this shortcoming by pro-viding for the transfer of energy from thebattery to the flywheel at a low currentlevel during the idle period and as afurther option during the cruise period.In this way, the flywheel has sufficientenergy at the start of the cycle to reachthe cornering point and make a substantialreduction in peak battery current under allconditions.

SELECTED OPERATING MODE

Operating Mode 4 with load l e v e l l i n gduring the i d l e period was, t h e r e f o r e ,se lec ted as the operat ing mode t o be usedfor the remainder of t h i s s tudy. Based ont h i s opera t ing mode, a computer s imulat ionstudy of the J227a Schedule D dr iv ing cyclewas made t o p r e d i c t the vehic le range fora single battery charge.

As a further output of the computerruns, detailed calculations of voltages,currents, e tc . , at discrete time intervalswere made, thus allowing subsequent designstudies of the various system components.The in/out efficiency of the flywheel pack-age was assumed to lie somewhere between 60and 100%. The recharging of the flywheelduring the vehicle idle period was at aconstant current for a 25-second interval.The idle period charging current was variedparametrically to determine the optimumflywheel recharge. Range was determined fortwo propulsion motor/gear ratios, 5.48 and7.307; the former corresponding to a vehiclespeed of 60 mph at full rated motor speed,and the latter to 45 mph at the same motorspeed and thus is a "lower" gear. The 7.307gear ratio (lower gear) proved to be betteradapted to the Schedule D driving cycle asthe cornering point of the traction motoris reached at a lower vehicle speed.Init ial computer studies show that the7.307 gear ratio achieved approximately 5miles better range than the 5.48 ratio.Therefore, the remaining computations weredone for a 7.307 gear ratio.

Plotting the calculated range vs. theidle time battery current showed a maximumrange occurred with between 50 to 60 ampsfed to the flywheel package. This maximumrange obtained at optimum idle rechargecurrent i s plotted in Fig. 5. On the basisof these results, the detailed design study

110

100

90

80

70 -

60 -

50

Zei > Weight Flywheel PropulsionPackage Differential

3700 Ib Total (vehicle andpassengers)

v200 Ib Flywheel PropulsionPackage Differential

3900 Ib Total (vehicleand passengers)

Mode 4 - 7.307 Gear RatioOptimum Battery Current During Idle

- Basic 3700 Ib Vehicle w/o Flywheel Packageor Battery Regeneration

-O-60 80

In/Out Flywheel Efficiency (%)

100

Fig. 5. Flywheel Augmented Vehicle Range

assumed 50 amps idle recharge current, aflywheel stored energy of 105 watt hours,and a gear ratio of 7.307.

FLYWHEEL ENERGY STORAGE PACKAGE

ELECTRICAL DESIGN

As far as the flywheel drive machineis concerned, the most difficult operatingpoint of the J227a driving cycle occursduring vehicle braking and flywheel motor-ing. From an electrical point of view, themost severe machine load occurs at 145% ofthe base (or minimum operating) speed,where the commutation requirements are 230amperes at 108 volts in the DC link. Thisparticular point is, therefore taken as thecritical design point for electrical trade-off studies.

NUMBER OF MAGNETIC POLES

The inductor machine which was builton the first DOE Contract (EY76-C-02-4010)had 8 poles. That design was chosen as acompromise between mechanical and elec-trical requirements. Maximum power con-ditioning frequency was limited to 1000 Hzand because of flywheel weight, the maxi-mum rotor speed was set at 15,000 rpm. Forthis application, a small steel flywheel is

38

Fig. 6. Machine with Maximum DesignSpeed of 20,000 RPM

used since less stored energy is required.The lower bearing loads in this case allowhigher rotor speeds. Thus, still withinthe constraint of maximum frequency of 1000Hz, it is possible to design either a 6-polemachine operating between 10,000 and 20,000rpm or a 4-pole machine operating between15,000 and 30,000 rpm.

Design tradeoff studies were performedfor 4-, 6- and 8-pole machines. The loadpoint at 145% of base -speed for the flywheelmachine during motoring was taken as adesign point. Results showed that the 6-pole machines provided the lightest weightfor this application. In order to assess

the tradeoff between weight and motor/alternator efficiency, electrical lossreduction was assessed in terms of overallvehicle weight. Vehicle system studiesshowed that an energy storage systemefficiency improvement of 1 percentage pointduring braking regeneration was equivalentto approximately 25 pounds of vehicle weightThis factor was considered in the overalltradeoff.

The electrical machine finally chosenfrom the tradeoff is a 6-pole machineoperating with a maximum speed of 20,000rpm, with a rotor radius of 3.6 inches anda rotor length of 3.44 inches.

MECHANICAL DESIGN

The basic mechanical design of theflywheel energy storage package now underconstruction is shown in Fig. 6.

The inductor machine portion of therotating assembly, including the six polesand the central shaft, is machined from amagnetic steel billet (AISI 4340). Steelshafts of non-magnetic, austenitic steelare welded to both ends of the pole bearingsection in order to minimise flux leakage,which tends to saturate the magnetic ironand magnetize the ball bearings, makingthem traps'"for magnetic wear particles.The stub shafts are hollow to reduce weight.

The rotor shaft is vertically orientedin order to minimize gyroscopic effectswhich would occur if the unit were operatedin a vehicle. Angular contact ball bear-ings are used. The upper bearing supportsthe weight of the rotor while the lowerbearing acts as a guide with a spring con-trolled preload. Bearing lubrication andcooling is achieved by an oil jet directedon the inner race. Most of this oil isthrown free of the bearing and returns tothe sump by a parallel path so as to mini-mize churning losses.

The oil is circulated by an internalgear pump driven through a gear reducerfrom the bottom of the main shaft. Becauseof the low atmospheric pressure in therotor enclosure, the pump will be operatedat low speed to prevent cavitation at thepump inlet. At full motor speed, the pumpwill operate at approximately 300 rpm. Thepump and gear reducer housing also servesas the oil sump and reservoir.

39

FLYWHEEL WINDAGE AND BEARING LOSSES POWER CONDITIONER & CONTROL

The outer sides of the machine polesare shrouded to reduce windage losses.Windage losses due to the flywheel can becomputed with reasonable accuracy. However,the windage loss due to the pumping actionof the lobed rotor can only be estimated onthe bases of test results from the previousmachine. For the 6-pole machine, the pres-sure within the housing will be set at 0.01atm helium. At this pressure, the directflywheel windage loss is predicted to be 38watts at the maximum rotor speed of 20,000rpm. This reduces to an average loss of 18watts over the flywheel duty cycle.

The average windage loss of the lobedrotor will be approximately 35 watts. Whenthe average rotor loss (35 watts) is addedto the average flywheel loss (18 watts) thetotal average estimated windage loss is 53watts.

DESIGN STRESSES

The energy stored in the inductor rotoris approximately 15 watt-hours over thespeed range of 10,000 to 20,000 rpm. Anadditional 90 watt hours will be stored inthe flywheel. This flywheel will be amultidisk shrunk on design made of vacuummelt AISI 4340 steel. It will operate ata relatively modest stress level in orderto produce a factor of safety of 2.0 rela-tive to the 10? cycle curve for alternatingstress. .Although a higher stress designwould be possible, it was not consideredworthwhile in this case since doubling thedesign stress would only remove approxi-mately 10% of the total package weight fromthe flywheel. The flywheel weight reduc-tion would be counteracted by the neod fora heavier containment ring. The hig;ierstressed wheel would also be larger dia-meter and occupy more vehicle space.

POWER CONDITIONER

The power conditioner, which is a loadconunutated inverter/rectifier, is based onthe system provided by General Electricunder the previous Contract (EY-76-C-02-4010). However, the new unit will belighter and smaller than the original loadcoounutated inverter (LCI) and will requireonly six power thyristors instead of twelve.Reversal of power flow is provided by ahybrid reverser (2 diodes and 2 contactors)rather than a second set of six thyristors.A block diagram of the entire system isshown in Fig. 7. The four major componentsfor the electrical portion of the drivesystem (battery, propulsion motor, hybridreverser, and load commutated inverter/rectifier) are shown together with theelectromagnetic contactors that connect thevarious components in the several modes inwhich the system operates. A contactclosure sequence for each mode is shown inTable 1. The designation Vj in Table 1 isthe motor rpm corresponding to the con-dition when the propulsion motor voltageequals the battery voltage. That speed isa function of the state of charge of thebattery and does not correspond to a fixedspeed for all driving cycles.

Table 1. Contact Closure Sequence

Mode

Motoring (Stop-V^)

Motoring (V^-Top)

Braking (Top-Stop)

Recharge at Stop

Initial Starting ofFlywheel

X = contactor closedContactor numbers are

Contactor1

X

X

X

shown

2

X

X

X

in

3 4

X

X

X

Fig. 7

Figure 8 shows the power and auxiliarythyristors of the load commutated inverter/

CommutatingCapKllor

Fig. 7. Schematic Diagram of SimulatedPropulsion System

Fig. 8.

Six flower Two Auxiliary

Thyristora Thyristors

Load Commutated Inverter/RectifierPower and Coiranutating Circuit

40

rectifier. In this system, power flow isaccomplished by reversing the direction ofthe DC voltage while maintaining the currentflew from the thyristors in the same direc-tion. The two auxiliary thyristors and thesingle commutating capacitor connected tothe synchronous motor neutral terminal forstarting are similar to that provided in theprevious contract.

This configuration of the power circuitwas chosen in order to provide the capa-bility to recharge the flywheel from thebattery during conditions of either zerospeed operation or full speed operation.

The electrical requirements of theload commutated inverter (relating to itsDC side) when operating over a simulatedSAE J227a Schedule D driving cycle are givenin Figs. 9, 10, and 11. Positive.currentindicates power flow from the flywheel tothe propulsion motor. Negative currentindicates power flow from the propulsionmotor, acting as a generator during braking,to the flywheel or from the battery to theflywheel. During the condition in whichthe battery is supplying power to the pro-pulsion motor or to the flywheel motor, thevoltage can be in the range of 75 to 108volts depending on the state of batterycharge. Figs. 9 and 11 provide the informa-tion necessary to select the power semi-conductors for the inverter/rectifier andtheir cooling requirements when repetitivelyoperating over the simulated driving cycle.

300

200

100

8S -100

250 A

10 sec

125 A

28 sec

10 20 30 40 50 60 70

Time —•-

80 90 100 110

-200

-300

Fig. 9. Inverter DC Link Current vs Time

CONTROL

The system control is required toregulate the operation of both the energystorage flywheel system and the vehicletraction motor. A number of variables inthe power circuit will be sensed to accom-plish this. In addition, it is desirableto establish the operating efficiency of

J227» Schedule D

t10

10 20 30 40 SO 70 80 90 100 110 120

Time (seconds)

Fig. 10. Flywheel Per Unit Speed vs Time

J227a Schedule D

10 20 30 40 50 60 70 80 90 100 110 120

Time (seconds) —»»

Fig. 11. Inverter DC Voltage vs Time

each major component and measure the energystorage efficiency of the flywheel energystorage package so as to determine theenergy improvement to the electric vehicle.

The control system will consist of aninverter current regulator with an internalfeedback loop using motor operating angleto maintain synchronism of the inverter tothe motor.

In addition to the primary controlsystem, two field current controls arerequired. One is for the flywheel motor,which will simply vary the flywheel motorfield current as a fixed function of in-verter DC link current, flywheel speed, andDC link voltage.

The second is for the traction motorfield which will remain fixed at low vehiclespeeds but must be varied at high speeds toregulate the armature current during theperiod of time that the battery suppliesall the power to the propulsion motor.

The major part of the system controlwill be implemented using the laboratoryhybrid controller. It allows for easilymade changes in control system configura-

41

tions and parameters as the test programdevelops.

SYSTEM TEST

An existing General Electric laboratorytest facility is currently being modifiedfor use on this contract. This laboratorysetup will include a new separately excitedDC propulsion motor, a flywheel simulatingvehicle inertia, an electrical load machineto provide road loss, and a torque trans-ducer to measure the propulsion motortorque. This equipment together with a 108volt battery bank, the inverter/rectifier,the inductor motor/alternator/flywheel pack-age, and the hybrid controller will beassembled in the laboratory for componentand system testing.

The system will be operated in 1979 todetermine the performance of the regenera-tive battery/flywheel energy storage pro-pulsion system as if it were operating ina 3000 pound battery electric vehicle.Performance will be established for the SAEJ227a Schedule D driving cycle.

ACKNOWLEDGEMENTS

The author wishes to acknowledge thetechnical contributions made to this pro-gram by the following General ElectricCorporate Research and Developmentpersonnel:

Mr. I.H. Edelfelt, System Analysis andComputer Simulation

Mr. D.W. Jones, Flywheel Energy Storage,Mechanical Design

Dr. A. Plunkett, Control System Design

Dr. E. Richter, Electrical Design ofInductor Type Synchronous Machine

Mr. F.G. Turnbull, Load Commutated inverterand Drive System Design.

REFERENCES

1. R.H. Guess, & E.L. Iiustenader,"Development of a High Performance andLightweight Hybrid Flywheel/Battery PoweredElectric Vehicle Drive," Fourth Inter-national Electric Vehicle Symposium,Dusseldorf, Germany, August 1976.

2. E.L. Lustenader, "Flywheel EnergyStorage System Development," FlywheelTechnology Symposium, San Francisco, CA.,October 1977. :

3. E.L. Lustenader, G. Chang, E.Richter, F.G. Turnbull, J.S. Hickey,"Flywheel Module for Electric VehicleRegenerative Braking," 12th IntersocietyEnergy Conversion Engineering Conference,Washington, DC, August 1977.

4. A.B. Plunkett and F.G. Turnbull,"Load Commutated Inverter/SynchronousMotor Drive Without a Shaft PositionSensor," 1977 IEEE/IAS Annual MeetingConference Record, October 1977, LosAngeles, CA., IEEE Publication No. 77CHI246-6-IA, pp. 748-757.

5. R.L. Steigerwald and T.A. Lipo,"Analysis of a Novel Forced CommutationStarting Scheme for a Load CommutatedSynchronous Motor Drive," 1977 IEEE/IASAnnual Meeting Conference Record, October1977, Los Angeles, CA., IEEE PublicationNo. 77 CHI246-8-IA, pp. 739-747.

6. E.L. Lustenader and E.S. Zorzi,"A Status of the 'Alpha-Ply' CompositeFlywheel Concept Development," Society forthe Advancement of Material and ProcessEngineering, 1978 National SAMPE Symposium.

7. A.B. Plunkett and F.C, Turnbull,"System Design Method for a \oad CommutatedInverter-Synchronous Motor Drive," IEEEIndustry Applications Society AnnualMeeting, October 1-5, 1978, Toronto, Canada.

8. R.L. Steigerwald, "Characteristicsof a Current-Fed Inverter With CommutationApplied Through Load Neutral Point,"GE Report 78CRD162, August 1978.

October 1978

42

PROJECT SUMMARY

Project Title: Low Cost Flywheel System Demonstration

Principal Investigator: D. W. Rabenhorst

Organization:

Project Goals:

Project Status:

Contract dumber:

Contract Period:

Funding Level:

Funding Source:

Applied Physics LaboratoryThe Johns Hopkins UniversityJohn Hopkins RoadLaurel, MD 20810(301) 953-7800

Develop and evaluate a flywheel capable of storing 20 watt hoursper dollar. Demonstrate this flywheel in a complete 115 vacsystem (1 KUH, 2 KM). Develop and evaluate low drag, long lifebearing systems.

All principal objectives are expected to be met. Approximately60 percent of the scheduled work has been completed in accordancewith the revised schedule of May 1978.

EC-77-C-01-5085

Oct. 1, 1977 - Mar. 31, 1979

5355,190

Department of Energy, Division of Energy Storage Systems

LOW COST FLYWHEEL DEMONSTRATION

D. W. RabenhorstThe Johns Hopkins UniversityApplied Physics Laboratory

Johns Hopkins RoadLaurel, Maryland 20810

ABSTRACT

The current Applied Physics Laboratory/Department of Energy Programinvolves the demonstration of a very low cost flywheel ($50/kwh) in acomplete energy storage system. The program also includes the develop-ment and evaluation of low loss, long life bearing systems. The energystorage system, operating at 115 VAC, has a capability of storing onekilowatt-hour of energy and of accepting or delivering this energy at anaverage rate of two (2) kilowatts power level. A final output of theprogram is the extrapolation of the flywheel design to a full size unitwith a storage capability of 10 to 100 kwh. In addition to meeting theprogram objectives, this program has resulted in the following notableachievements in related flywheel technology:

- A flywheel configuration has been demonstrated which allowsthe exploitation of a wide variety of applicable low costmaterials in a very inexpensive fabrication process.

- A flywheel suspension system has been developed which reducesbearing loads by more than 90%, thus permitting the use ofsmaller bearings at lighter loads and longer projected life.Resulting bearing performance approaches that of a threeplane magnetic suspension—without the disadvantages of themagnetic system.

- Several novel bearing systems were evaluated which haveexhibited further reduced losses while having an order ofmagnitude longer predicted life.

- A number of novel low cost flywheel materials were evaluatedwhich promise future flywheels having even lower cost.

INTRODUCTION

The Applied Physics Laboratory.The Johns hopkins University (APL)is currently engaged in a 15-monthprogram with the Department ofEnergy (DOE) which has as itsprimary objective the feasibilitydemonstration of a flywheel energystorage system utilizing a verylow cost flywheel. The principalobjectives of this program are asfollows:

1. Demonstrate a flywheelcapable of storing one kilowatthour at a cost of less than 20watt hours per dollar ($50 per

kilowatt hour).

2. Evaluate the character-istics of this low cost flywheel.

3. Demonstrate the flywheelin a complete home type energystorage system having one kilo-watt hour storage at a rate ofapproximately two kilowatts power.

4. Develop and evaluate longlife low drag bearing systems foruse in flywheel systems.

5. Provide design projec-tions into energy storage systemsof interest to DOE (e.g. 10 kilo-

watt hour, etc.).

6. Investigate and evaluatepotential low cost flywheelmaterials.

In addition to the finalreport on this program coveringthe foregoing subjects, theprincipal product of the programwill be a complete energy storagesystem operating at 115 voltsinput capable of receiving onekilowatt hour of energy, storingthis energy for an unspecifiedperiod of time, and deliveringthis energy to a load at 115 voltsAC. This energy storage systemwill demonstrate the generalfeasibility of nighttime energystorage in an individual home.

FLYWHEEL DEVELOPMENT

The Applied Physics Laboratoryhas been engaged for a number ofyears in the development of fly-wheel configurations which notonly permit optimal use of thefilamentary materials but willpermit their use in configurationshaving an absolute minimum fabri-cation cost. A detailed descrip-tion of the APL bare filament fly-wheel configuration is containedin the referenced report,1 whereit was concluded that the barefilament configuration offers thehighest possible energy per unitweight, energy per unit volume,and energy per unit cost of anyknown flywheel configuration.The principal reason for this isthat the performance of any woundflywheel is a function of thestrength to weight ratio of thematerial used in its construction.The typical wound multi-ring fly-wheel construction involves afilamentary high strength materialin a polymer matrix, with theratio of filament to polymerseldom exceeding 70%. In the APLconfiguration, on the other hand,the ratio of high strength fila-ments to polymer matrix is gener-ally of the order of-98 to 99percent. Thus, the performanceof the APL flywheel can be afunction of the strength to weightof the high strength filamentitself, without the detraction of

a considerable amount of matrixweight.

The hundreds of flywheel andmaterial spin tests which followedthe original bare filament fly-wheel concept have demonstratedthat there are a number of addi-tional advantages to the APL barefilament flywheel configuration.The problems of expansion and con-traction of the wound filamentsas the flywheel rotational speedis increased or decreased arereduced to an absolute minimum.In the bare filament configurationthe bulk of the wound filamentsare free to expand and contractwithout interfering with oneanother or transferring loads toone another. This expansion andcontraction is apparently donevery evenly, in view of the ex-tremely large number of filamentsinvolved. A further advantagearising from the relatively inde-pendent action of the individualfilaments is the fact that it istheoretically impossible to havea simultaneous failure of all ofthe filaments in the flywheel.A catastrophic failure of the barefilament wound flywheel has neverbeen experienced in any of the APLspin tests to date. In contrast,the typical failure pattern isthat the outer fibers will fail,leaving the inner fibers compris-ing the major portion of the fly-wheel assembly, intact. It hasalso been demonstrated that thefailure of the outer wound barefilament fibers in the APL fly-wheel does not necessarily meanthat the flywheel will go out ofbalance and be destroyed. Onseveral occasions a relativelylarge proportion of the fiberswas destroyed, but the rest ofthe flywheel remained intact, andwas brought satisfactorily torest without further damage. Onnumerous occasions the flywheelhub assembly was reused two,three, or even four times follow-ing a flywheel spin test todestruction.

It would appear, in view ofthe foregoing, that the APL barefilament configuration, in addi-tion to its performance potential

45

offers exceptional safety charac-teristics compared with othertypes.

MATERIALS EVALUATION

Of all the materials investi-gated for use in the subject pro-gram, three were selected as havingthe appropriate characteristics interms of performance, cost, fabri-cation simplicity, availabilityand safe failure mode. Althoughthe three materials selectedappear to have satisfactory char-acteristics in all of these areas,they represent respectively verydifferent material configurations.The vinyl impregnated fiberglassstrand material, hereafter refer-red to as vinyl-glass, actually isa bundle of as many as 1000 fiberseach having a diameter of .0003inch. The presence of the vinylin this configuration serves toprevent abrasion among the fila-ments, and also, to some degree,excludes water vapor from theglass filaments. In its optimumconfiguration the vinyl-glasswould be processed immediatelyafter drawing the molten glassfrom the platinum crucible, evenbefore the glass strand hasreceived any aqueous lubricant,which normally protects it as itis routed over numerous pulleysin the manufacturing process.

In the flywheel applicationthe vinyl impregnated glassappears to have a considerablyhigher usable strength than ordi-nary uncoated fiberglass, and thisleads to higher flywheel perfor-mance and/or lower flywheel cost.Usable strengths of the order of50% of the ultimate tensilestrength are believed to be pos-sible with this material configu-ration.

The second material whichqualified for the low cost fly-wheel bare filament configurationis steel wire, such as that usedin pneumatic and hydraulic hosereinforcement. This wire, calledhose wire, has a diameter of .012to .015 inch, and is usuallyplated with a very thin costing ofbrass to protect the wire and to

serve as a lubricant in the fabri-cation and processing operations.Although the hose wire is marginalin energy to cost potential, itwas nevertheless selected as theprimary material in the APL pro-gram, in view of the fact that itsstrength and cost characteristicsare easily determined, whereas thecorresponding characteristics ofthe other materials consideredinvolve a considerable amount oftheoretical projections.

The third material selectedfor the bare filament flywheelconfiguration is Metglas®. Gener-ically. Metglas has been termedas a metallic alloy with glass-like properties. The propertiesof most interest to the flywheelapplication, however, are the factthat the Metglas can be made tohave a high strength to weightratio at a relatively low projectedcost. It also offers the prospectof providing these capabilitiesin a configuration which alsoqualifies on the issue of safetyin the failure mode. The Metglasused in the program to date hasbeen a one-half inch wide ribbonhaving a thickness of .002 inch.This ribbon form gives the Metglasa distinct advantage over theother two materials mentioned, inthat the winding time to fabricatea flywheel of a given size can beas much as a hundred times lessthan the corresponding windingtimes of the other materials.Eventually this could have a con-siderable effect on the fabricatedcost of the flywheel in question.

VINYL-GLASS FLYWHEEL

The vinyl-glass used in thesubject program was provided atno cost to the program by PPGIndustries, Inc., Glass ResearchCenter, Pittsburgh, Pennsylvania.The material used up until thepresent time is 50% vinyl and 50%glass, primarily because thismaterial in that form was avail-able from other current applica-tions. PPG has indicated, however,that it will be a relativelysimple matter to produce thismaterial in a configuration con-sisting of 10% vinyl and 90% glass

46

especially for the flywheel appli-cation, and this material isexpected to be made availablelater in the current program.

A typical vinyl-glass testflywheel is illustrated in Fig. 1.

Fig. 1 Vinyl-glass flywheel.

Here it can be seen that thevinyl-glass is wound in a verydense pattern, and gives theappearance of being bondedthroughout. Actually the onlyareas of polymer bonding are atthe four narrow band radial wrappositions, and everywhere else onthe flywheel the vinyl-glassstrands are in the so-called barefilament condition. The vinyl-glass appears to be an extremelytough material. On at least oneoccasion the flywheel has beenre-spin-tested with only minorrepairs, after having been inad-vertently dropped to the bottomof the spin chamber at 20,000 rpm,and was bounced around inside thechamber for several minutes.

One characteristic of thevinyl-glass which represents adistinct advantage during thewinding and handling processes,may actually turn out to be adisadvantage in the final analysis.This property is the "tackiness"of the vinyl-glass strand as oneturn is laid on top of another inthe winding process. While thiscondition greatly facilitates the

process, it may result in effectivepartial bonding of these strandsin the flywheel, which is an unde-sirable feature of this flywheelconfiguration. This effect willbe evaluated in more detail in thebalance of the current program.

METGLAS FLYWHEEL

The typical Metglas test fly-wheel is illustrated in Fig. 2

Fig. 2 Metglas flywheel.

The Metglas provides an extremelydense, clean and attractive woundconfiguration, where the Metglasoccupies essentially 100% of thewound structure. This is in con-trast to the fiber and wire con-figurations, where the materialoccupies only about 80% of thewound structure.

The Metglas is a proprietarymaterial in experimental productionat the Allied Chemical Corporationof Morristown, N. J. All of theMetglas used in the current pro-gram was provided at no cost tothe program by Allied Chemical.The material provided to thepresent time is a ferrous alloywhose general strength character-istics correspond to an equivalentsteel. The mcnufacturing processinvolves extremely rapid chillingof the molten material at a rateof approximately two milliondegrees centigrade per second.The result is a smooth metallicribbon which has surprisingly

47

uniform dimensions and physicalcharacteristics. The configura-tion of the Metglas currentlybeing processed into flywheels iscalled the "As-Cast" condition.In this condition the Metglasapparently has microscopic cracksalong the edges, which preventachievement of the ultimate physi-cal characteristics of thismaterial. A process has beendeveloped at Allied Chemicalwherein these microscopic cracksare removed without appreciablyadding to the production cost ofthe material. Following thisprocess the Metglas ribbon isexpected to have usable strengthsof the order of 60% higher thanthe usable strengths in the."As-Cast" material received to date.Nevertheless, the experiencegained in processing and spintesting flywheels of the "As-Cast"Metglas material has beenextremely valuable, and hasallowed the development of fabri-cation techniques which will applyto the ultimate Metglas material.

STEEL WIRE FLYWHEEL

The typical steel wire fly-wheel is illustrated in Fig. 3

Fig. 3 Steel wire flywheel.

The wire is helically wound on tothe flywheel hub in a manner sim-ilar to that used for the vinylglass flywheel described in theforegoing. The steel wire usedin the current flywheel testing

at the Applied Physics Laboratorywas received at no cost to theprogram from two separate sources.The first source is the CentralSpecial Studies Group of IndustriePirelli S.p.A. in Milan, Italy.The second source is the NationalStandards Company in Niles,Michigan, U.S.A.

The fabrication of the steelwire bare filament flywheel hasinvolved the development of specialtechniques to accommodate thepeculiar characteristics of thissteel wire. Early attempts towind the bare filament steel fly-wheels were unsuccessful becauseof these characteristics; however,the problems have be*~>n resolved,and successful configurations arenow a standard achievement. Threeprincipal problems were as follows.First, a satisfactory wire termina-tion scheme had to be developedwhich would allow the terminationof the final end of the wire onthe inside of the flywheel, inorder to permit the maximum utili-zation of the tensile strength ofthe wire. The method developedwas to pre-machine a helical slotoccupying one quarter of a revolu-tion around the periphery of theflywheel hub. Thereafter, uponcompletion of the winding of thesteel wire, a final revolution washand wound to the center of thewound structure, whereupon the endof the wire was fed through thepre-machined slot and bondedsecurely. This method has provento be quite satisfactory, and, infact, was later adapted for use inthe vinyl-glass flywheel describedabove. The othex* two problemsconcerned the "slickness" and wire-cast conditions. It was found thatthese two problems working togetherresulted in a wound flywheelstructure which was staticallyunstable, when removed from thewinding mandrel. After a seriesof experiments, and careful con-sultation with the NationalStandards Company, appropriatewound geometries and processeswere established which eliminatedthis problem.

48

FULL SIZE FLYWHEEL

The characteristics of thefull size flywheel to be used inthe final energy storage demonstra-tion system are illustrated inFig. 4.

Tapered hub plates

Fiberglass caps

Radial wrapover these

Flat section is 1.5%\Subcircular = 9.6 in. radius

Hub nominalradius = 9.81 in.

Steel wire(2.25 x 2.25 in.)

Wet wind.-*JuTr

.75 in.-*J'1? in _ I J

Fig. 4 Full size one kWh rotor.

The physical dimensions of thisflywheel are those required tosatisfy the rotational speedrequirements of the motor genera-tor and control system. In thiscase a maximum rotational speedof 14,400 with a four to onespeed ratio resulted in a minimumrotational speed of 3600 rpm.Then, knowing the maximum designperipheral velocity of thedesired flywheel material, itwas a straightforward calculationto determine outside diameter ofthe flywheel. In this case,using the steel wire as the fly-wheel wound material, the outsidediameter was 24 inches, and thecross section dimensions of theflywheel are simply determinedfrom the amount of weight in thewound flywheel necessary to permitthe storage of one kilowatt hourof energy at the maximum rotationalspeed. Actually, the total amountof energy in the flywheel at thiscondition is approximately 1100watt hours, in order to permit thestorage of one kilowatt hour at afour to one speed ratio.

The material selected for thehub was the most effective studiedin terms of structural capabili-ties, damping qualities, minimumfabrication cost, and minimummaterial cost. Ic. this case thematerial is multi-ply BalticBirch plywood. Although thismaterial typically has plugs andoccasional voids internally, ithas proven to be a very consistentmaterial, probably because of thefact that in the total hub struc-ture, there are 27 plys of material.In fabrication a rigid compressionring is formed of the windingmaterial itself. The steel wireis wet wound for the first onequarter inch radial dimension,which forms a stiff ring bondedto the plywood hub materialthroughout its periphery. There-after, the winding is continuedwithout resin except for the verythin radial bands in the fourradial wrap positions.

Although the steel hose wirewas selected as the primarymaterial for this full scale fly-wheel, it will be of interest tocompare the properties of thisflywheel with those of similarones made of vinyl-glass andpolished Metglas. This comparisonis made in Table 1.

Table 1

Low cost flywheel materials comparison.

Tensile strengthIntrinsic energydensity - Wh/lbUsable energydensity - Wh/lb

Assumed materialcost — c/lbMaterial performanceWh/$Winding weight —Ibs-for 1 kWhRelative volumeRelative winding time

10/90vinyl-glass

264.00040.8

17.8

50

35.6

60

2.12.1

Polishedmetglass

0.02" x 0.5"

450,00025.2

15.8

50

31.6

67

0.67.02

Steel hosewire

368,00020.4

12.7

60

21.2

84

11

Considering first the steel wireflywheel, the relatively lowusable energy density and result-ing high winding weight are

49

relatively inconsequential for thestated application. Of particularimportance, however, is the factthat the material performance issomewhat marginal at 21.2 watthours per dollar, indicating thatthe steel wire flywheel cost maybe somewhat higher than desired.

The relative volume and rela-tive winding time are arbitrarily•set at unity for the steel wireflywheel, in order to best comparethis with the other types. Therelative winding time is a secondorder indication of total cost ofthe flywheel.

The vinyl glass bare filamentflywheel, on the other hand, hasabout 40% more energy density andnearly a 70% advantage in materialperformance cost. However, bothof these items are based upon theassumption that the vinyl-glassconfiguration will permit utiliza-tion of 50% of the ultimate tensilestrength, compared with 70% in thecase of the steel wire. Whilethis may be a perfectly validassumption, based on the prelimi-nary testing to date, a consider-able amount of additional testingwill be required in order to con-firm this assumption. It can beseen that the vinyl-glass flywheeloccupies more than twice thewound volume of the steel wireflywheel, and it is largelybecause of this that the relativewinding time is also greater thanin the case of the steel wire.

The polished Metglas flywheelappears to combine the bestadvantages of the other two, withsome additional advantages of itsown. Usable energy density issomewhat comparable to that ofthe vinyl-glass flywheel, as isthe material performance cost.But this latter factor is depen-dent upon a very importantassumption that the projectedcost of the Metglas material willbe $.50 per pound, which is twoorders of magnitude lower costthan it is at the present time.

Because of its very highusable strength, the Metglas fly-wheel occupies 50% less volume

than the steel wire flywheel.However, because this material isused in a relatively wide tapeform rather than wire form, itsrelative winding time is approxi-mately one fiftieth of the wind-ing time of the steel wire andl/100th of the winding time ofthe vinyl-glass. This could bean important factor in favor ofthe Metglas flywheel over theother types, inasmuch as thewinding time is a significantcost factor in the flywheel fabri-cation. It should also be pointedout that the Metglas can appar-ently be made available in oneinch widths, as opposed to theone-half inch widths currentlybeing used, which should resultin further reductions in fabrica-tion cost.

FLYWHEEL BEARING ANDSUSPENSION DEVELOPMENT

The design of the flywheelbearing system and suspensionsystem are inextricably tiedtogether. In the case of the APLsystem an attempt has been madeto minimize the axial and radialloads on the bearings, and therebypermit the use of much smallerbearings at a much lower thannormal load rating. Also thebearings themselves, have receivedconsiderable attention, anddesigns have been developed whichappear to offer somewhat lowerdrag but an order of magnitudelonger projected lifetime. Thesebearings together with the APLsuspension system are describedin the following.

BEARING TEST EQUIPMENT

The test rig illustrated inFig. 5 was used to evaluate thevarious bearing configurationsbeing considered.

50

SERIES BEARING CONFIGURATION

Figure 7 illustrates theAPL series bearing concept whichis, in effect, one bearing rota-ting inside of another bearing,with the resulting rotationalspeed of each being reduced by 50%.

Stationary housing

Identical highspeed bearings

Fig. 5 Bearing test equipment. Fig. 7 Series bearing concept.

The losses were determined byrespective measurements of bearingtorques using the sensitive torquewatch device illustrated in Fig. 6.

The test assembly of this conceptis illustrated in Fig. 8, where itcan be seen that the bearings arecompactly arranged so as to beaxially adjacent rather thanradially adjacent.

Fig. 6 Torque watch for measuring bearing.

This device has an apparentaccuracy of 1/100th of an ounceinch. The bearing test equipmentis so arranged that tests can beconducted either at atmosphericpressure or under vacuum condi-tions.

Fig. 8 Series bearing test unit.

In addition to providing an orderof magnitude increase in projectedlifetime by virtue of the greatlyreduced rotational speeds, thisbearing arrangement also providesimproved reliability (since it isunlikely that both bearings wouldfail simultaneously), and reduced

51

bearing losses.

The use of the series bearingconfiguration in the APL flywheelenergy storage system results insuch a low rotational speed on thebearings, that more exotic schemessuch as the Draper RetainerlessBearing Concept, are not believedto be required.

MAGNETIC SUPPORT SYSTEM

Since the flywheel spin axisin the APL energy storage systemis in the vertical plane, it ispractical to consider relievingthe gravity loads from the fly-wheel and motor generator with apassive magnet system. Forexample, if the rotating machineryweight were 85 pounds, and amagnetic support system wereutilized having a capability of8C pounds then the actual loadfrom gravity on the bearings wouldbe only 5 pounds. In other words,the load on the bearing would bethe same order of magnitude asthe desired preload on the bearings.The overall size of the rotatingmass is such that a permanent magnetsystem could be considered, andsuch a system is illustrated inFig. 9. Its magnetic forceproperties are illustrated inFig. 10.

140

l i20llOOI 80-I 60-

40

Magnet — Barium ferrite no. 5O.D. = 3.950"I.D. = 1.292"Thickness = 0.425"

Steel shell = 4.683" O.D.-Nominal design range

= 80 ± 5 Ib

2 4 6 8 10 12 14 16GAP—thousandths of an inch

18

Fig. 9 Magnetic support component

Fig. 10 Performance of magnetic support system.

The magnetic attractive mode wasselected for the system ratherthan the repulsive mode for anumber of reasons. First, in theattractive mode the magnet isstable in two planes and unstablein only one plane (the axialplane). On the other hand in therepulsive mode the magnet systemis stable in one plane (axial)and unstable in the other two.But perhaps even more importantis the fact that in the repulsivemode the optimum system wouldrequire two magnets; whereas onlyone magnet is required in theattractive mode. The effect onmagnetic system cost is obvious.

The Jobmaster Corporation inRandallstown, Md., who designedand built the APL magnetic supportsystem, conducted an in depthstudy of the advantages and dis-advantages of the rare earthmagnets versus the simple ceramicferrite magnets. These studiesindicated conclusively that theferrite magnet by far offers out-standing performance per unit cost.It has been projected that themagnetic support system illustra-ted in Fig. 9 and having the per-formance approximately the sameas Fig. 10 would, in mass produc-tion, cost approximately 75£.Thus, the addition of the magneticsupport system to the low coststationary energy storage systemflywheel represented an insigni-ficant increase compared with theadvantages gained.

52

SUSPENSION SYSTEM DEVELOPMENT

The basic APL concept forthe stationary flywheel energystorage system suspension isillustrated in the Fig. 11 sketch.

Drivemotor

"^-Energy absorbingelastomer

Relativelystiff shaft

Flywheel

Energy absorbingelastomer

Fig. 11 Flywheel suspension system concept.

The principal advantage of thisflexible shaft suspension systemis that the radial loads (orimbalance loads) on the flywheelbearings are reduced to anabsolute minimum, and actuallyare virtually eliminated. Asecondary advantage of thisarrangement is that the systemae.ommodates a relatively largeamount of imbalance without trans-mitting imbalance loads to thebearings.

SYSTEM DISCUSSION

Two types of damping areemployed in the APL flywheelsuspension system. There is hubdamping of the motion between theflywheel plane and the spin axis,and there is driver damping whichdamps motion between the overallsprung system and the supportstructure. Hub damping has beenapplied in the previous APL sub-systems by means of a rubbercoupling between the flywheel andthe flywheel shaft. In the pastsuch a coupling has been in theform of a typical Lord Corporationshock mount located at the fly-wheel hub, in order to permitgimballing, as well as isolationof the flywheel from the rotatingsystem. A second type, and the

one which is being used in thecurrent APL system, is the bondedhub arrangement, where the fly-wheel hub structure is bonded tothe metal hub assembly with a thinrubber slab located in between, inorder to permit accommodation <.-*differential expansion, and also toprovide damping.

A number of schemes have beentested to provide driver damping.One successful driver dampingarrangement is illustrated inFig. 12, and this is the arrange-ment currently used in the APLspin test facility.

Fig. 12 Demonstration system suspension system.

Here the turbine drive assemblyis isolated from the supportstructure through a solid rubbercoupling, which allows consider-able motion between the turbineand the support structure, whileat the same time providing thenecessary damping between thesestructures. A functionallysimilar arrangement is that em-ployed in the full-size energystorage system as a part of thecurrent program. Here the entireflywheel container assembly ismounted on suitable shock mountsto provide the same effect as thedriver damping.

These flywheel suspensionarrangements coupled with anappropriately designed flexibleshaft having critical speeds far

53

in excess of those expected to beencountered, permitted relativelysmooth flywheel operation through-out the design rotational speedrange.

ENERGY STORAGE SYSTEM DESIGN

Two complete energy storagesystems will be fabricated in thecurrent program. The first ofthese systems will be the so-called"battleship model" made withruggedized components for thepurpose of evaluating the electri-cal and thermal problems inadvance of the final system com-ponent availability. The princi-pal components of this energystorage system are illustrated inFig. 13.

Fig. 13 Energy storage demonstration system.

The flywheel in this system is alaminated aluminum disk which,while storing only about half ofthe final flywheel assembly,nevertheless operates at essen-tially the same rotational speedrange as the final flywheel system.The motor generator and controlsystem are essentially the sameas those which will be used inthe final system, and this equip-ment will be used to evaluate theoverall characteristics of thismotor generator and control system.The rotating machine is a squirrelcage induction motor designed tooperate between 3600 and about

15,000 rpm. Its design is basedupon an off-the-shelf unit havingmoderately high performance.

The control system receives115 volt, one phase, 60 hertzinput power, rectifies it to acontrolled level DC voltage, andinverts it to controlled frequencythree phase AC voltage. The con-trol system is arranged to provideconstant current per phase, sothat the motor produces a constanttorque up to the point where theline voltage reaches 230 VAC at120 Hertz. At this point themotor is producing in excess offour HP. Over the entire rota-tional speed range, the averagepower input and output is of theorder of two kilowatts.

The control system has beendesigned to accept the energy froma 115 VAC source, transfer thisenergy into the flywheel, allowstorage of the energy in the fly-wheel for an unspecified period,and finally to transfer the energyfrom the flywheel into a 115 VACelectrical load.

REFERENCE

1D. W. Rabenhorst and T. R.Small, "Composite Flywheel Develop-ment Program: Final Report"APL/JHU Report No. SDO-4616A datedApril 1977.

54

PROJECT SUMMARY

Project Title: Materials Program for Fiber Composite Flywheels

Principal Investigator: J. A. Rinde

Organization: Lawrence Livermore LaboratoryUniversity of CaliforniaP. 0. Box 808, L-338Livermore, CA 94550(415-422-7077, FTS 532-7077)

Project Goals: The goals of this project are threefold: (1) to accelerate thewidespread use of the fiber composite flywheel by developing thenecessary engineering design data on fiber composite materials,(2) to demonstrate the high energy kinetics attainable with fi' f*composite materials, and (3) to transfer the technology thusqained to the private sector.

Project Status: Our fiber composite materials program for flywheels is dividedinto the following areas: matrix resins, static engineeringproperties, stress rupture (lifetime tests at constant load),and dynamic fatigue. During the past year, we characterized arubberized epoxy resin that offers improved fracture toughnessand suitable performance at moderately elevated temperatures(up to 70 C). We also evaluated six epoxy resins for service at150 C. In addition, the rubberized epoxy resin was used as amatrix in Kevlar 49 composites and engineering design data weregenerated. We initiated stress rupture tests on E-glass com-posites at load levels of 60 to 85% of short-term strength;these tests will continue for several years. We also begandynamic fatigue tests on Kevlar 49 composites in the tension-tension mode. In addition, fatigue tests on a Kevlar 49composite rinq specimen at 50 to 75% of ultimate strength arein progress. We anticipate that these data will provide adirect estimate of flywheel performance.

Contract Number: W-7405-ENG-48

Contract Period: Continuing

Funding Level: $350,000 (includino rotor development)

Funding Source: U. S. Department of Energy, Mechanical Energy Storage Division

55

MATERIALS PROGRAM FOR FIBER COMPOSITE FLYWHEELS*

J. A. RindeLawrence Livermore Laboratory, University of California

Livermore, California 94550

ABSTRACT

Our fiber composite materials program for flywheels is divided into the followingareas: matrix resins, static engineering properties, stress rupture (lifetime testsat constant load), and dynamic fatigue. During the past year, we characterized arubberized eposcy resin thtL offers improved fracture toughness and suitable perfor-mance at moderately elevated temperatures (up to 70°C). We also evaluated six epoxyresins for service at 150°C. In addition, the rubberized epoxy resin was used as amatrix in Revlar 49 composites and engineering design data were generated. We initi-ated stress rupture tests on E-glass composites at load levels of 60 to 85% of short-term strength; these tests will continue for several years. We also began dynamicfatigue tests on Kevlar 49 composites in the tension-tension mode. In addition, fa-tigue tests on a Kevlar 49 composite ring specimen at 50 to 75% ultimate strength arein progress. We anticipate that these data will provide a direct estimate of flywheelperformance.

DJTRODOCTION

The fiber composites materials pro-gram at LLL was begun in 1975 with theprimary goal of providing meaningful andreliable engineering design data on com-posite materials specifically intendedfor flywheel rotors. The two main crit-eria for material selection for this fly-wheel program are high performance andlow to reasonable cost. Because thehigh-performance composites used by theaerospace industry are very costly, wecould not consider these materials, eventhough thorough design data exist.

Therefore, since 1975, we have:• Generated engineering design data forKevlar 49, S2-glass, and E-glass com-posi tes.• Spin tested thin-rim flywheel rotors ofKevlar and glass fiber composites andcorrelated the results with the fibers'static strength properties.• Developed and engineered new fiber com-posite flywheel design—a quasi-isotro-pic, laminated, solid disk rotor with atapered profile.• Characterized flexible and rubberizedmatrix resins• Investigated the transverse tensileproperties as a function of matrix resinmodulus for a series of S2-glass compos-ites.• Conducted stress rupture tc~ts on

Revlar 49 and E-glass composites.During the past year, we concentrated ourefforts specifically on new matrix res-ins, E-glass composite stress rupture,engineering properties of a compositemade with a rubberized resin, and com-posite fatigue lifetime. Each of theseareas is discussed below.

MATRIX RESINS*

We studied two types of epoxy re-sins this year, a rubberized resin withimproved fracture toughness, and hightemperature resins for service at 150°C.Complete details of our work are given inRefs. 1 and 2.

RUBBERIZED EPOXY 3ESIN

A rubberized epoxy resin is a nor-mal epoxy resin that has been modified bythe incorporation of a soluble carboxy-terminated butadiene acrylonitrile (CTBN)rubber. During the curing process, theCTBN rubber is forced out of solution toform a second phase of l-to-10-um dia-meter rubber particles. These particlesmodify the fracture process of the resinand provide improved fracture toughness.In our work we formulated a resin systemSix table for wet filament winding, con-ducted fracture tests to demonstrate the

•Principal Investigator, J. A. Rinde.

*This work was performed under bhe auspices of the 0. S. Department of Energy byLawrence Livermore Laboratory under contract No. W-7405-Eng-48.

56

resin's improved fracture toughness andthen characterized the mechanical andphysical properties of the resin.

Table 1 summarized the properties ofthis resin system. The epoxy resinXD 7575.03 (Dow Chemical) contains 10%CTBN rubber. This rubberized resin sys-tem offers the advantages of low viscos-ity and long gel time for easy processingas well as a moderately high tensilestrength of 76 MPa. However, since wecompleted this work, both the XD 7575.03and XD 7714 (Dow Chemical) have beenwithdrawn from the commercial market.An equivalent product for the XD 7575.03can be obtained by using 23% KelpoxyG-293 (Spencer Kellog Division ofTextron) plus 77% DER 332; XD 7114 canbe replaced by Wilmington ChemicalsHelox 68.

RESINS FOR HIGH-TEMPERATURE SERVICE

We have formulated and evaluated sixepoxy resin systems suitable for wet fil-ament winding of high-performance fly-wheels. We determined their processingcharacteristics as well as their mechani-cal properties. Specifically, we judged

the resin systems on five criteria: (1)low viscosity (1.0 Pas at 25°C), (2)long gel time (220 h for a 30-g mass at25°C), (3) high glass transition tem-perature (Tg 2 180°C), (4) high ten-sile strength with nigh modulus, and (5)good retention of mechanical propertiesupon accelerated aging (7 days at 175°C).Two of the resin formulations (1 and 2)have tensile strengths above 85 MPa andglass transition temperatures above 205°Cin both and as-cured and aged conditions.The resin systems formulated and some oftheir key characteristics are presentedin Table 2.

All resin systems were cured for 3 hat 70°C plus 2 h at 120°C plus 2 h at180°C to achieve a high Tg and to curethe resin at a temperature above the in-tended-use temperature of 150°C. Withthe exception of the resin systems curedwith APCO 2347 (modified imidazol), thesecure conditions produced a T g greaterthan 185°C. Both the APCO 2347 systems(resins 5 and 6) exhibited additionalcuring upon heat aging.

The six resin formulations were sub-jected to two mechanical-properties tests

Table 1. Properties of a rubberized epoxy resin system.

ComponentsResin: Dow XD 7575.03Diluent: Dow XD 7114Curing agent: Tonox 60-40

Cure cycle,h/°C

Viscosity at 25°C,Pa*s

Gel time for 30-g massat 25°C, h

Cured density at 25°C,Mg/m3

Water absorption (ASTMD-570-63), % gain/h

1.5/90+ 2/1300.95

29.3

1.19

2.02/8,3.13/24

Parts by weignt1006533.9

Tensile properties:

Stress at maximum, MPa

Stress at failure, MPa

Strain at maximum, %

Strain at failure, %Modulus, GPa

76.1

72.5

5.7

8.42.43

Glass transition temperature,°C 104 ,

Thermal coefficient of linearexpansion from 298 to377 K, in./in./°C 71 x 10"6

IZOD impact (ASTM D-256-73),J/m:Method A 47.7Method E 645.4

Flexural Strength at 5% strain,MPa 106.4

Tangent modulus, GPa 2.85

Compressive properties:

Maximum strength, MPa

Strain at maximum strength, %Secant modulus, GPa

Torsional properties:Maximum shear stress, MPaStrain at maximum stress, %

Tangent modulus, GPa

88.7

5.72.9

54.916.4

1.59

57

Table 2. Performance of the six epoxy resin systems formulated for elevated-temperature service.

Resin Resin Gelsystem components Viscosity, time,

No. (parts by weight) Pa-s h

Tensile PropertiesModulus,GPa

Stress,MPa

Strain,% Tg,°C

Ciba 0510/RD-2/APCO 2330 1.05 21.1(100/20/43.1): Cureda

Agedb

Ciba 0510/Tonox 60-40 1.33 25.1(100/49.8): Cured

AgedAPCO 2447/ERL 4206/Tonox 60-40 (100/20/ 1.08 38.639.9): Cured

AgedCiba 0510/DEN 438/RD-2/Tonox 60-40 (75/25/25/ 1.48 19.649.9): Cured

AgedCiba 0510/DEN 438/RD-2/APCO 2347 (75/25/20/ 0.88 34.412.2): Cured

AgedDER 332/RD-2/APCO 2347 0.90 48.5(100/15/9.5): Cured

Aged

5.44.8

4.14.1

3.93.9

4.04.2

3.13.4

2.82.6

89.584.7

86.873.4

83.571.4

81.078.5

74.873.2

59.853.7

2.02.0

3.22.2

2.42.0

2.52.8

3.72.8

2.72.5

205220

210226

190215

185213

115170

115145

a3 h at 70°C plus 2 h at 120°C plus 2 h at 180°C.b7 d at 175°C.

and to a thermal gravimetric analysis inboth the as-cured and aged conditions.Resin samples were aged in a constant-temperature, forced-air oven at 175°Cfor 7 days. In general, aging causedthe samples to darken and become brit-tle. The results of these tests aresummarized in Table 2.

The tests reveal that resin system1, cured with APCO 2330, meets our processsing requirements of low viscosity andlong gel time, and has the highest ten-sile strength and modulus, the secondsmallest loss in tensile strength uponheat aging, and the second highest T g

in both the as-cured and aged conditions.Resin system 3 also meets the processingrequirements, but has a lower T g, alower tensile strength, and a largerreduction in strength upon heat agingthan resin 1.

We conducted dynamic shear modulusmeasurements as a function of temper-ature; the results for all six resinsystems in the as-cured condition areshown in Fig. 1. We used these measure-

5

Pig.

300

Temperature, °C

1. Dynamic shear modulus for thesix epoxy resin systems formulatedfor high-temperature service (testedat 0.1 Hz). All resin were curedfor 3 h at 70°C plus 2 h at 120°Cplus 2 h at 180°C. Resin 1 isCiba 0510/RD-2/APCO 2330, resin 2is Ciba 0510/Tonox 60-40, resin 3is APCO 2447/ERL 4206/Tonox 60-40,resin 4 is Ciba 0510/OEH 438/RD-2/Tonox 60-40, resin 5 is Ciba 0510/DEN 438/RD-2/APCO 2347, and resin 6is DER 332/RD-2/APCO 2347.

58

100

98

a? 96

.i M3 92

90

88

- T

-1 1

1 1

55=• i

— —175°C

,275°C

i i i

150°C_

-* 200°C ~

225°C -

250°C• 1 i

20 40 60

Time, min

80 100

Fig. 2. Isothermal weight loss atvarious temperatures in air by ther-mogravimetric analysis for resinsystem 1, Ciba 0510/RD-2 APCO 2330.

ments to determine the Tg of the curedresins as well as to show the variationof modulus with temperature and to pro-vide a measure of cross-link density.From these results, we conclude that re-sins 1, 2, 3, and 4 would perform well at150°C in short-term application.

The effect of heat aging these re-sins is exhibited by the change in Tgand tensile properties (see Table 2). Inall cases, the Tg increases with agingbut the tensile strength and failurestrain decrease with aging.

Figure 2 presents the isothermalweight loss curves for resin system 1 attemperatures from 150 to 275°C asdetermined by thermogravimetric analy-sis. At 150°C, resin 1 is thermallystable after an initial weight loss ofabout 2%. Above about 200°C, thesample exhibits an accelerating rate ofdecomposition as a function of the in-crease in temperature. Therefore, weexpect that flywheels using this resinsystem would have a lifetime of severalyears at 150°C and be able to with-stand shorter exposures (~30 min) totemperatures up to 200°C (Tg = 205°C).

STRESS RUPTURE*

We are also characterizing the time-dependent strength of several candidatefiber composites being considered forflywheels. These data are needed to pre-dict the probability of failure for a

flywheel operating under various stressesassociated with the input, storage, andoutput of energy. We use. the results ofstress rupture tests at constant loadlevels as the baseline benchmarks. Suchtests are required because even a nominalvariation (typically less than 5%) instatic strength can lead to a largescatter in stress rupture life, often inexcess of 100% (see Fig. 3). Therefore,to provide the necessary statisticalparameters for the reliable design ofcomposite components, large data samplesfrom long-term testing are being accumu-lated in testing facilities designed forsimultaneous testing of dozens ofsamples. Figure 4 illustrates the typeof data being generated. From suchcurves, we can determine the amount ofstress-level derating that is requiredto obtain the desired degree of reli-ability in the component's operatinglifetime.

We are conducting the first stressrupture tests on E-glass/epoxy; E-glassis a low-cost, large-bundle fiber of in-terest for flywheel applications. Forthese tests, the glass yarn was inpreg-nated with resin (Dow DER 332/JeffamineT-403) on a filament-winder and the re-sultant composite strands were given amild heat cure. Strands were cut tolength, fixed with end tabs, and loadedin the test apparatus at 60 to 85% oftheir average ultimate failure strength.In Fig. 4, the stress rupture results areplotted as precent load vs log time.Curves for Kevlar 49/epoxy and S-glass/epoxy composites also are included forcomparison.

•at 5 5 0

500

450

400

rrcv<

y cv1

1 . • . i

>iro%\

1 1 1 Xl.

0 100 500

Time to failure, h

1000

•Principal investigator, L. Penn.

Fig. 3. Nominal scatter in staticstrength data that can result inlarge scatter in lifetime predic-tions .

59

1h

I ' I I I I I I ' I *>l. I I I I I

5010 -3

i il l i 11 i i i 11 i i i 11 i i i 11 i \ i i i

10- 10r 1 10° 101

Lifetime, h102 103 1O4 105,

Pig. 4 Stress-rupture lifetime data for several composite materials. The 2-to-100%bands for S-glass and Kevlar 49 composites are displayed; data points are forE-glass composites.

The most noticeable feature inFig. 4 is the much broader distributionof break times in the E-glass compositeas compared to the S-glass and theKevlar 49 composites. This was expectedbecause E-glass is a low cost fiber andis known to have a higher variabilitythan the more expensive, aerospace gradeS-glass fiber. Some of the variabilityin E-glass results may have been causedby an inaccurate load applied to thesample. Such inaccurate application ofload is possible because of the leverarm arrangement of the apparatus. Wewill perform some dead-weight load teststo verify or disprove this hypothesis.Because so few data are available at thelower load levels at this time, we candraw no conclusions from these data con-cerning the useful stress levels forE-glass composite flywheels.

ENGINEERING PROPERTIES*

We have also conducted an extensiveevaluation of the physical and mechani-cal properties of a Kevlar 49 compositeusing the rubberized epoxy resin systemdescribed above. The primary advantageof this composite over previously testedKevlar composites is the higher Tg ofthe rubberized matrix resin.

Composites specimens 60-to-70-vol%Kevlar 49 fiber (1420 denier withoutsizing) were filament wound. The resinsystem was XD 7575.03/XD7114/Tonox 60-40(100/65/33.9). Composites were cured for4.5 h at 60°C plus 3 h at 120°C. Thefabrication and testing methods are re-ported elsehwere.3'4

Elastic constants and ultimatestrength properties of composites of 60-,65-, and 70-vol% fiber are presented inTable 3. The stress-strain curves of the65-vol% composite are shown in Fig. 5.The performance of this composite is com-pared to two other Kevlar 49 compositesusing different resin systems (DER 332/Jeffamine T-403, and XD 7818/JeffamineT-403) in Table 4. This comparison re-veals that the rubberized epoxy yieldslower shear properties than the othertwo matrices; other mechanical proper-ties, however, are comparable.

FATIGUE LIFETIME*

Knowledge of the time-dependent de-formation and strength properties isessential for the design of flywheels toensure minimum dimensional change, maxi-mum dynamic stability and long-termsafety of operation. Among the candidate

•Principal investigator, L. C. Clements. *Principal investigator, E. M. Wu.

60

Table 3. Mechanical properties data for a composite of Kevlar 49 in a rubberizedepoxy resin, XD 7575.03/XD 7114/Tonox 60-40.

Fiber content, vol%

Property 60

16852.6

8 .70.17

19.21.426.82 .7

(CV8)

(5.9)(5.2)

(8.8)(11.4)

(2.8)(3.7)

(2.4)(12.3)

Hb

55

55

7777

65 (CV)

80.52d (8.2)7.82 (15.5)1.7 (2.0)0.367 (3.96)0.037 (7.33)

1814 (4.1)2.6 (3.5)220 (5.8)0.35 (7.2)6.4 (10.5)

40.3 (9.1)1.4 (15.4)16.2 (3.8)1.2 (4.7)22.55 (1.2)2.0 (3.6)

N

2258

225

66665

558888

70 (CV)

1920 (3.0)3.3 (2.0)

4.8 (10.6)

16.2 (4.3)1.2 (7.9)21.36 (1.5)2.0 (8.9)

N

44

5

8888

Elastic Constants0

Longitudinal young's modulus, GPaTransverse Young's modulus, GPaShear modulus, GPaMajor Poisson's ratioMinor Poisson's ratio

UltimatesLongitudinal tension: stress, MPa

strain, %Longitudinal compression: stress, MPa

strain, %Transverse tension: stress, MPa

strain, %Transverse compression: stress, MPa

strain, %Shear at 0.2% offset: stress, MPa

strain, %Shear at failure: stress, MPa

strain, %

^ V coefficient of variation, in percent.TJumber of specimens tested.cElastic constants are assumed to be valid for both tension and compresion.Estimated from tests at 60 to 70 vol% fiber, normalized to 65 vol%.

2000

1600

1200

800

400-

Longitudinaijtension

'65% fiber volume

- j»Longitudinal "ompression^Transverse tension _.5tf i i- snear

250

200

150

100

50

0 1 2 3Strain, %

Fig. 5. Mechanical properties of65-vol% Kevlar 49/epoxy compositein tension, compression, and shear:rubberized epoxy was XD 7575.03/XD7114/Tonox 60-40.

flywheel composite materials, only com-posites of Kevlar 49 exhibit time-depen-dent deformation in both the fiber-con-trolled and the matrix-controlled proper-ties. Therefore, Kevlar 49 compositesare being studied extensively because oftheir apparent high engineering poten-tials.

We are characterizing the time-de-pendent properties of Kevlar 49 com-posites in the fiber direction by con-ventional operations using linear visco-elasticity. Tests being conducted areillustrated in Fig. 6. Creep compli-ances are obtained from static-weighttests. The onset cf nonlinearity isdependent on both stress level and load-ing rate and is being identified by com-paring creep tests and ranp loading testsat several loading rates. We are usinglow-frequency fatigue tests (1000 s/cycle) to identify the effects of cyclicstress history on the composite deforma-tions and lifetimes. By comparing thefatigue data to our stress rupture.data,we will obtain the design curves used toestimate composite lifetimes in accor-dance with current engineering practice(e.g., S-N or strain vs number of cyclesto failure curves). We are also explor-ing a theoretical correlation via the

61

Table 4. Mechanical properties of 60 vol% Kevlar 49 composites in three resinsystems, two rigid matrices and the rubberized matrix.a

Resin matrixProperty Der 332/T-403 XD 7818/T-403 Rubberized

Elastic Constants'3

Longitudinal Young's modulus, GPaTransverse Young's modulus, GPaShear modulus, GPa

UltimatesLongitudinal tension:

Transverse tension:

Shear at 0.2% offset:

stress, MPastrain, %

stress, MPastrain, %stress, MPastrain, %

81.85.101.82

18502.237.90.1624.41.55

75.14.561.89

14001.712.40.2833.671.98

80.57.81.7

16852.68.70.1719.21.4

aRigid matrices: DER 332/T-403 (Dow Chemical/Jefferson Chemical), XD 7818/T-403(Dow Chemical/Jefferson Chemical). Rubberized matrix: XD 7575.03/XD 7114/Tonox 60-40(Dow Chemical/Dow Chemical/Oniroyal).

^Elastic constants are assumed to be valid for both tension and compression.

damage function formulation in an attemptto increase the utility of these fatiguedata in generalized applications.

A second part of our fatigue programis to test small Kevlar 49 rings by loadcycling in the tension-tension mode atthe stress levels expected in operatingfiber composite flywheels (i.e., 50 to75% of ultimate stress). We believe that

CO

1Creep and creep rupture

«• • • • ^ M w » m^^ «HW M^^B av

•••

•* .• Ramp loading

/ / ..•** Fatigue

Time — » •

Fig. 6. Mechanical test performed todetermine the time-dependent be-havior of Kevlar 49 composites inthe tension-tension mode.

this cyclic test simulates the ring-typeflywheel and that the test results willbe useful as a direct estimate of fly-wheel structural performance.

In conventional engineering prac-tice, the time-dependent strength of acomposite is characterized in terms ofthe input loads {e.g., the S-N curvesand the stress rupture curves). In ourwork, however, we are relating materialsresponse properties to the input loads.In this manner, we are not only producingimmediately usable engineering data, weare also laying the groundwork for ageneralized failure theory for fiber com-posites. Much of our effort and re-sources have been directed toward thenecessary instrumentation and data aqui-sition systems to measure the strain re-sponses of the composite. Examples ofthe creep and fatigue responses are shownin Pig. 7. The experimental results ac-cumulated to date are shown in Fig. 8:the open data points are from creeprupture tests and the solid points arefrom fatigue tests. In Fig. 8a, we pre-sent the failure points in the conven-tional format of stress level (i.e., in-put load) vs the time to failure {i.e.,number of load cycles). In Fig. 8b, wepresent the failure points in terms ofthe materials response (i.e., the failurestrain) vs the time to failure.

62

250 U.Ul

0.008

Strain

, %

P

P

0.002

n

• ' • ' • ' ' ( b )

-

-

-

i i

2.4 4.8 7.2 9.6Strain response to fatigue, h

120 15 30 45 60

Strain response to creep, h

Fig. 7. Strain response to creep (a) and to fatigue (b) of Kevlar 49/ epoxy strands.Fatigue testing was conducted with square wave cycling.

10"1 10 102 103 104

90

a 85 -

80

840

£ 820

•S 800

780

1 1 1 11 . i i • i

W'20%1 1 ' 1

^ ^

i I 1

30'% Failed'

^ - A

1 1 1 i 1 i

• > • 1 i

^= = :A : : :^^5«! !

i , . 1 ,

(a)

i i i

I ' 1 r 1 I I i

(ma

n t

o faili

Strai

760

740

720

im-2

70a10

± ±

g } Creep

• Cyclic fatigue

• • . 1 •

10"1• •

1 10310 10£

Time to break, h

Fig. 8. Creep and fatigue of Kevlar 49/epoxy strands tested at 25°C: (a) stress(i.e., input load), and (b) strain (i.e., materials response).

104

63

Plots of stress vs time to failura(Fig. 8a) ace useful as design chartswith which to estimate the lifetime re-duction of the composite component due tocreep and fatigue loads. On the otherhand, plots of fatigue strain (Fig. 8b)suggest the generalized observation thatcomposite strain compliance at failureis weakly dependent on stress level andstrongly dependent on load history.

FUTURE WORK

We are continuing our stress ruptureand dynamic fatigue programs to obtainmore baseline data for predicting thelifetimes of fiber composite flywheels.Kevlar 49 and E-glass composites are nowbeing tested. New stress rupture andcyclic fatigue tes ts also will be con-ducted at elevated temperatures and undervacuum. Some of this work will be doneunder contract at the Oak Ridge NationalLaboratory. We plan to continue char-acterizing the matrix resin formulatedfor service at 150°C. We also will in-vestigate resins with improved bonding toKevlar 49 in an attempt to increase thetransverse tensile and shear propertiesof the composite. Engineering propertieswill be determined on S2-glass/high-temp-erature-service resin composites at roomtemperature and at elevated temperatures.Some work may also be done on low-costgraphite fiber/epoxy composites.

REFERENCES

1. J. A. Rinde, E. T. Mones, R. L.Moore, and H. A. Newey, An EpoxyResin-Elastomer System for FilamentWinding, Lawrence Livermore Laboratory,Rept. UCRL-81245 (1978)

2. J. A. Rinde, E. T. Mones, and H. A.Newey, Filament Winding Epoxy Resins forElevated Temperature Service, LawrenceLivermore Laboratory, Rept. UCRL-52577(1978).

3. L. L. Clements and T. T. Chiao,"Engineering Design Data for an OrganicFiber/Epoxy Composite," Composites 8,87-92 (1977).

4. L. L. Clements, R. L. Moore, and T.T. Chiao, "Elongated-Ring Speciman forTensile Properties of Filament-WoundComposites," in Materials Review '75,Proc. 7th Natl. SAMPE Tech. Conf.,Azusa, CA, 1975, p. 188.

REPORTS, PUBLICATIONS, ADO PRESENTATIONS

1. C. C. Chiao and T. T. Chiao, AramidFibers and Composites, LawrenceLivermore Laboratory, Rept.UCRL-80400 (December 1977); to be achapter in 27K Handbook on Fiber-glass and Plastic Composites.

2. T. T. Chiao, J. H. Rinde, and E. T.Mones, Epoxy Resins for Fiber Com-posite Flywheel Rotors, LawrenceLivermore Laboratory, Rept.OCRL-79573 (October 1977), presentedat the 1977 Flywheel TechnologySymposium, San Francisco, CA.

3. T. T. Chiao, Material Propertiesof Composite Flywheels, LawrenceLivermore Laboratory, Rept.UCRL-80515 (April 1978), presentedat The Middle Atlantic RegionalMeeting of the American chemicalSociety, Hunt Valley, MD.

4. T. T. Chiao, Some InterestingMechanical Behavior of CompositeMaterials, Lawrence LivermoreLaboratory, Rept. UCRL-80908 (April1978), presented at the VS-VSSRSeminar on Fracture of CompositeMaterials, Riga, USSR.

5. R. M. Christensen, J. A. Rinde, andE. T. Mones, Transverse TensileCharacteristics of Fiber CompositesUsing Flexible Resins, LawrenceLivermore Laboratory, Rept.UCRL-80241 (October 1977),presented at the 1977 FlywheelTechnology Symposium,San Francisco, CA and accepted forpublication in J. Polymer Eng. Sci.

6. L. L. Clements, Fiber CompositesFlywheel Program - Filament WoundComposite Data Sheets, LawrenceLivermore Laboratory, Rept.UCID-17874 (August 1978).

7. L. L. Clements, "Problems inTesting iVramid/Epoxy Composites,"Lawrence Livermore Laboratory, Rept.UCRL-79450 (November 1977), presen-ted at the AIMS Failure Modes inComposites IV, Chicago, IL.

8. L. L. Clements and R. L. Moore,Comparative Engineering Propertiesof Fiber Composites for Flywheels,Lawrence Livennore laboratory, Rept.UCRL-79575 (October 1977), presentedat the 1977 Flywheel TechnologySymposium, San Francisco, CA.

64

9. L. L. Clements and R. L. Moore,Composite Properties for S2-Glassin a Room-Temperature Curable BpoxyMatrix, Lawrence LivermoreLaboratory, Rept. UCRL-8I517(August 1978), to be published inSAMPB Quart.

10. S. Kulkarni, Interlaminar andStacking Sequence Considerationsfor Composite Flywheels, LawrenceLivermore Laboratory, Rept.UCRL-13888 (July 1978).

11. L. S. Penn and T. T. Chiao, BpoxyResins, Lawrence LivermoreLaboratory, Rept. UCRL-79815(October 1977), to be a chapter inThe Handbook on Fiberglass andPlastic Composites.

12. L. S. Penn, PhysiochemicalCharacterization of Composites andQuality Control of Raw Materials,Lawrence Livermore Laboratory, Rept.UCRL-81081 (Hay 1978), presented atthe 1978 ASTM 5th Conference onComposite Materials, New Orleans,LA.

13. J. A. Rinde, B. T. Hones, and H. A.Newey, Filament Winding BpoxyResins for Elevated TemperatureService, Lawrence LivermoreLaboratory, Rept. UCRL-52577(October 1978).

14. J. A. Rinde, E. T. Hones, R. L.Moore, and H. A. Newey, An BpoxyResin-Elastomer System for FilamentWinding, Lawrence Livermore Labora-tory, Rept. UCRL-81245 (September1978), prepared for presentation atthe 34th Annual Conference of SPIReinforced Plastics/Composites In-stitute, New Orleans, LA (January1979).

15. E. M. Wu, Failure Analysis ofComposites with Stress Gradients,Lawrence Livermore Laboratory, Rept.UCRL-80909 (August 1978), presentedat the US-USSR Seminar on Fractureof Composite Materials, Riga, USSR.

NOTICE"This report was prepared as an account of worksponsored by the United States Government.Neither the United States nor the United StatesDepartment of Energy, nor any of their employees,nor any of their contractors, subcontractors, ortheir employees, makes any warranty, express orimplied, or assumes any legal liability or respon-sibility for the accuracy, completeness orusefulness of any information, apparatus, productor process disclosed, or represents that its usewould not infringe privately-owned rights."

PLL

65

PROJECT SUMMARY

Project Tftle: The Laminated Dfsk Flywheel Program

Principal Investigator: R. G. Stone

Organization: Lawrence Livermore LaboratoryP. 0. Box 808, L-123Livermore, CA 94550Telephone: (415) 422-8284

Project Goals: (1) Develop the technology of laminated disk flywheels.(2) Demonstrate a prototype of a reliable, economical,

high-energy density flywheel based on the developedtechnology.

(3) Evaluate and disseminate this technology.

Project Status: This program was started very recently, July 1978. Initialanalyses have been performed and are continuing. Manufac-turing development has been started. One flywheel has beenbuilt and tested.

Contract Number:

Contract Period:

Funding Level:

Funding Source:

W-7405-Eng-48

July 1978 - July 1980

$775,000

Department of Energy

S

67

THE LAMINATED DISK FLYWHEEL PROGRAMA ROTOR DEVELOPMENT PROJECT BY LLL AND G.E. CO.*

Richard G. StoneLawrence Livermore Laboratory

P. 0. Box 808, L-123Livermore, CA ol»550

ABSTRACT

The Lawrence Livermore Laboratory and the General Electric Company** have initiateda joint program to develop the technology of fiber-composite, laminated disk flywheelsfor energy storage applications. The 2-year program was started in the simmer of 1978.LLL and G.E. will participate about equally in applying their complementary capabilitiesto the overall program. LLL is developing analytical methods applied to contoured diskwheels. LLL will also investigate bonded hub attachment methods and will select materialsand define processing requirements. G.E. is developing an alpha cross-ply laminate disk,a rim overwrap, and a mechanical hub attachment. G.E. will also develop manufacturingprocesses and will test the developmental and prototype flywheels. Concepts originatedin prior efforts and analyses are presented. Current efforts include the testing of acontoured, laminated disk flywheel.

INTRODUCTION

A number of studies have concludedthat energy storage flywheels have thecapability to conserve energy in a varietyof applications. Some of these studieshave concluded that incorporating a fly-wheel in a battery-powered vehicle willnot only conserve energy but will alsoprovide the acceleration and hill—climbingperformance demanded by users. Fiber-composite materials are most attractivefor energy storage flywheel construction,having the potential of high-energydensity, relative safety, and economy.However, the potentially high-energydensities have not been realized thus farin efforts to develop fiber-compositeflywheels. In particular, little atten-tion has been given to the developmentof laminated disk flywheels.

The Lawrence Livermore Laboratoryand the General Electric Company haveinitiated a joint program to develop thetechnology of fiber-composite, laminateddisk flywheels for energy storage applica-tion. Since this 2-year program was

recently started, this paper will sum-marize the concepts and analyses leadingto the program, describe the plannedprogram, and report on the program workaccomplished thus far.

BACKGROUND

For several years the LawrenceLivermore Laboratory and the GeneralElectric Company have been involved inprograms related to the development offiber-composite flywheels. LLL has beendeveloping fiber-composite materials dataand processing methods applicable to fly-wheels. We have also been engaged in theproject to develop and evaluate mechanicalenergy storage subsystem technology forapplication to electric and hybrid ve-hicles, technical support to the Electricand Hybrid Vehicle Demonstration project,and analysis of the role of energy storagepower systems in transportation. In thecourse of this work, Christensen and Wuperformed a design analysis of fiber-composite flywheels1 and Toland surveyedrecent developments in the application offiber-composite materials to flywheels.2

*This work was performed under the auspices of the U.S. Department of Energy by LawrenceLivermore Laboratory under contract No. W-7^05-Eng-l(8.**Reference to a company or product name does not imply approval or recommendation of theproduct by the University of California or the Department of Energy to the exclusion ofothers that may be suitable.

68

G.E. has been active in flywheelenergy storage projects including develop-ment of inductor/motor/alternator/flywheelenergy storage systems for automotiveapplications and development of variousfiber-composite flywheel concepts underG.E. company funding. An attractive con-cept reported by Lustenader and Zorzi1* isthe "alpha cross-ply" pseudo-isotopic,laminated disk flywheel.

As a consequence of these studies,the two organizations independentlyarrived at the conclusion that a mostpromising rotor design is a disk typecomposed of a number of laminas of fibercomposite materials oriented to maximizestrength. Disks of right-circular andof tapered (contoured) cross sectionsare of interest.

PROJECT DESCRIPTION

GOALS

The objective of this 2-year programis to develop the technology for high-energy density, fiber-composite flywheelsbased on the laminated disk concept andto demonstrate a prototype, practicalflywheel of this design. Our goal isa working capacity in prototype wheelsexceeding 50 Wh/kg of total storagecapacity in the 1- to 2- Kwh range, al-though we expect to demonstrate modelwheels failing at energy densities wellabove 90 Wh/kg. Our goal with respect tothe practicality of the prototype wheelsis a design directed to optimize theoverall flywheel energy system withrespect to low volume, low weight, manu-facturability and economy.

TASKS

The program is planned in five tasks:(l) establish analytical models; (2) manu-facture and test model flywheels; (3) de-velop hub attachment; {k) design, manufac-ture, and demonstrate performance ofprototype flywheels; and (5) assessment andrecommendations for the developed tech-nology. Tasks 1, 2, and 3 are closelyinterrelated and thus will be conductedconcurrently.

1. Analytical Models. We will use ana-lytical models capable of predictingstress and energy density of laminatedflywheels to make predictions of flywheelperformance. LLL is concentrating on theeffects of contouring on stress and energy

density. A two-dimensional finite-elementcode that uses uniform in-plane propertiesand uniform, but different, axial proper-ties has been applied to our first modelflywheel designs. Work is underway onmore detailed analyses of critical areastaking into account the anisotropic mate-rial properties and the combined inter-laminar and continuum stress fields.

G.E. has developed a two-dimensionalorthotropic finite-element simulation ofthe alpha-ply flywht-sls. They are currentlyanalyzing the effect of hoop-wound over-wrap on stress concentration at hub attach-ment holes. They are studying overwrapmaterials of higher modulus, such asKevlar 4 9 and graphite, with various de-grees of interference fit.

LLL and G.E. will compare and assesstheir analytical models and will applythem cooperatively.

2. Manufacture and Test Model Flywheels.In this task about 30 model flywheels willbe designed, built, tested, and evaluated.Sizes will range from 10 to 30 in. indiam. by 1/U- to 1-in. thick. These testswill verify the analytical models, thesuccess of manufacturing methods, and theeffects of hub attachment perturbations.

All of the flywheel tests will beconducted at the G.E. Corporate Researchand Development Center, Schenectady, NewYork.

LLL will produce about 10 model fly-wheels for testing. The first test wheelswill be directed toward verifying theanalytical model. Subsequent wheels willbe designed to evaluate anisotropy effects,optimized contouring, and manufacturingoptions.

G.E. will produce about 20 modelflywheels for testing. One group ofwheels will evaluate the effect of variousalpha-ply angles. Another group will beused to study the effectiveness of outerwrap rings of various radii and inter-ference fits. Hub attachment methodswill be evaluated in a third group.Finally, preferred alpha-ply, outer wrapcontouring and hub attachment combinationswill be tested and evaluated.

Manufacturing development is anecessary and important part of thistask. G.E. will investigate four areasof manufacturing: (l) the lay-up of

69

fiber prepregs into composite slabs,(2) outer ring winding and press fitting,(3) contoured wheel production, and(It) attachment fabrication. During manu-facturing development, G.E. will do thefollowing:

• Make flat slabs about 12-in. diam. and1-in. thick with the objective of assuringslab uniformity, low void content, andfiber parallelism.• Take micrographs to assure low voidcontent.• Evaluate resin characteristics andpost cure heat treatments.• Apply manufacturing processes to theouter ring winding.• Study winding directly onto wheelsand winding into rings for subsequentpress-fitting onto the wheels.• Develop contoured wheel manufacturingmethods.• Examine the effects of machining thecontoured wheel.

3. Develop Hub Attachment. The LordCorporation and David Rabenhorst of theApplied Physics Laboratory, Johns HopkinsUniversity, have developed an adhesive-bonded elastomeric hub attachment methodwhich has been used successfully intesting laminated disk flywheels.5 LLLwill use this method for their earlytest wheels. This design, as well asother potential bonded-hub designs, willbe investigated for application to workingflywheels.

The G.E. overwrap concept alleviates,to some extent, the stress concentrationat holes drilled in the wheel for hubattachment. Other methods of mechanicalhub attachment designed to minimize thisstress concentration will be investigated.Installing slender axial rods withoutcutting fibers during lay-up of the diskis one possible approach.

h. Prototype Flywheels. We will applythe information gained in the previoustasks to design, build, and test proto-type, practical flywheels. The designwill be directed to optimize the overallflywheel energy storage system withemphasis on the goals of low weight, lowvolume, reliability, manufaeturability,and economy. We expect that we willarrive at two designs of 1- to 2- kWhcapacity an*3. will build and thoroughlytest two wheels of each design.

5. Assessment and Recommendations. Thistask concludes the program with an evalua-tion of the technology developed andrecommendations for design and manufactureof these high-performance wheels, forapplications of this flywheel technologyto Department of Energy programs and forfurther flywheel development.

ORGANIZATION AND MANAGEMENT

As stated previously, this will bea joint LLL/G.E. program under the co-ordinating leadership of LLL. The analysisand design efforts will be about equally-divided between the two organizations.Although concentrating on their specificproblems, the approaches and results willbe shared by LLL and G.E. for maximumadvancement. LuL will concentrate itsefforts on contoured wheels, bonded-hubattachment, and materials selection andprocessing. The G.E. effort will bedirected toward laminate disk manufacturing,rim overwrap development, mechanical hubattachment, and component and system de-sign. G.E. will perform spin testing andwill be responsible for manufacture ofthe prototype flywheels.

The program will be coordinated byquarterly review and planning meetingsand ad hoc technical meetings. LLL willbe responsible for reports and programreviews.

ACCOMPLISHMENTS

This program was approved and fundedby the Department of Energy in June 1978.The contract with General Electric Com-pany was executed in August 1978. Thus,this 2-year program has just gotten under-way.

G.E. had developed and analyzed thealpha cross-ply concept previously andhad conducted a few verification tests.Since August they have continued analyt-ical work, started developing manufacturingmethods and building model alpha cross-plyflywheels for spin testing.

LLL had developed and analyzed theirquasi-isotropic, contoured concept; ob-tained material; and was building a testwheel under their materials and processingprogram. Under this program, this wheelwas completed and tested; and the resultsand material problems are under investigation.

70

GENERAL ELECTRIC CO. PROGRESS

The G.E. alpha-ply concept definesthe orientation of the fibers in adjacentlamina as

Angle between lamina = 90 .921N

where H is the alpha number. Thus, analpha-3 lay-up has the fibers of adjacentlamina displaced 60° from each other, analpha-9 is at 80°, etc. Their analysisshows improved energy potential foralpha values above J as shown in Fig. 1.Very preliminary tests appeared to verifythis trend.

2500

2000

1500

0

With Kevlar wrap

Without Kevlar wrap

a-3 a-5 cn-7

Alpha construction

a-9

Fig. 1. 2-D finite-element analysisof E-glass flywheels, with center hole.

Included in the G.E. continuinganalytical effort is a study of theeffect of a hoop-wound overwrap on thelaminated disk. The winding has twopurposes. One is to reduce "shredding"of peripheral glass fibers of thelaminate observed in some unwrapped fly-wheels. The other is to moderate theeffects of stress concentration at thehub mounting holes. Material propertiesof overwrap, thickness of overwrap, inter-ference fits, and materials costs arebeing evaluated.

Recent effort has concentrated onmanufacturing process development directedtoward high-quality production of alphacross-ply wheels for spin testing. Theprocedures developed thus far includestamping out circular lamina, accuratepositioning in the lay-up, evacuatingand hydroclave pressing and curing.Quality and uniformity are indicated bythe data in Table 1. The data are for

four samples taken from a 1-in. thick,10-in. diam. disk made from S-2 glassuniply-prepreg. Two alpha-9 wheels havebeen produced for spin testing. Bondedhubs, applied by Lord Corporation, willbe used in testing these wheels.

Table 1. Test data of a laminated wheelby hydroclaving.

Location

Property 12 3 6 £

% Glass 65.56 65.50 65.1*1 65.5>t

Measured 1.8001 1.8023 I.8O33 1.8017density

% Voids 2.062 1.9*123 1.781 1.974

LLL PROGRESS

The LLL concept is a quasi-isotropielaminate fabricated into a rotor havinga contoured shape similar to the idealizeduniform stress rotor (Stodola) for iso-tropic materials. The "proof-of-prlnciple"flywheel was, in fact, manufactured verynearly to this idealized contour. Excep-tions were that the central area was flatto accommodate the bonded hub and thatthere was no attempt to modify the contourto compensate for a finite radius versusthe idealized infinite radius. Figure 2is a sketch of the wheel "as built."

The laminated slab from which thewheel was machined was made of Celion6000 carbon fiber prepregged with Harmco5213. It consisted of 160 lamina orientedat 9o/!t5°/90o/135°, etc. symmetrical aboutthe center plane of the wheel. This slabwas fabricated by the Babcock and WilcoxCompany. The hub was manufactured andbonded to the slab by the Lord Corp., andmachining and balancing were performed atLLL. As machined, the wheel was verynearly in balance. The minor correctionrequired was accomplished by "moving"the hole in the hub. There was novibration problem during the spin test.

Figure 3 shows the finished wheel.The appearance of the numerous concentricrings is the effect of contour machiningexposing in turn each lamina from thecentral area out to the edge thickness.Nonuniformity of these rings, like con-tour lines on a map, indicate nonuniformityof the lay-up. Although noticeable, thiswas not believed to be serious.

71

-24 in. dia.-

Lord Corp. rubberbonded hub

Test wheelMaterial — Celion 6000 carbon fiber prepreg with Narmco 5213Lay-up — 0°/±45°/90°; center plane symmetry; 160 LaminaWheel weight - 11.5 Ib.; fiber volume 62%; density 0.056 Ib./cu. in.

Fig. 2. The Lawrence Livermore Laboratory "proof-of-prineiple" flywheel.

Fig. 3. The LLL flywheel for test No. 1.

The flywheel was tested at theApplied Physics Laboratory, Johns HopkinsUniversity. Failure occurred at 36,000rpm and was sudden and catastrophic.Energy density at this speed was 62.6 Wh/kg(28.h Wh/lb) for the entire wheel. Exami-nation of the debris indicated that thehub bonding had performed satisfactorily

and that fracture of the wheel probablyinitiated near the center.

A two-dimensional finite-elementanalysis of the wheel stresses andstrains had "been performed prior to thetest. This code uses uniform in-planeproperties and uniform, but different,

72

lower stresses and strains at failurethan the 0° tests. We conclude thelack of strength isotropy contributedsignificantly to the lower-than-expectedperformance of the rotor.

Our conclusions on the results ofthis first test flywheel are:

• The slab of material from which thewheel was machined had strength propertiesonly about 2/3 of what one should expectof this material.• The selected lay-up of O°/±lt5°/9O°wassomewhat less isotropic in strength (in-plane) than we expected.• The design and analysis of the wheelappears to be verified. LLL will publisha complete report on building and testingthis flywheel.

axial direction properties. As-builtdimensions and "published data" propertiesof the material were used. This analysisgave an expected maximum energy densityof 120 Wh/kg (5iu 5 Wh/lb). We anticipateda lower energy density due to in-planeanisotropy of the material and manufac-turing defects in the rotor althoughwe did not observe appreciable defectsduring machining of the rotor.

There was adequate excess materialfrom the lay-up slab to obtain "actual"material properties. However, thesehad not been tested prior to the wheelspin test. These tests have since beenconducted and we found that the "actual"material properties, stress and strainat failure, were only about 2/3 of the"published data" properties. Table 2gives a comparison of published data,actual material data, and spin test data.Based on these rather limited data, wefeel that the rotor performance wasseriously reduced due to a materialproblem. Discussions with the materialfabricator have not revealed the causeof the problem. It is speculated thatlengthy or improper storage of theprepreg material or an inadequate curecycle may have contributed to the problem.

To further evaluate the materialproblem, we obtained test panels of thesame material from Celanese Company, thematerial supplier. Samples of a uni-directional fiber lay-up and a 0°/±l)5°/90o

lay-up were tested and gave resultsessentially equal to the published data.In addition to testing the 0°/±2i5o/90o

material at 0°, we tested samples at22-1/2°. These 22-1/2° tests gave much

Table 2. Comparative properties of the O°/±k5°/9O° laminate and flywheel test data.

Published data

Sample from flywheelblank

Calculated from spintest

Spin test as percent

Failurestrain, %&

1.20

0.78

o.6o

78

Umaxialstress, kpsi

77-8

51.1

—

Biaxial stressestimate, a/(l-v)

112.7

78. h

61.7

78of flywheel material

aAt 0 deg.

73

REFERENCES

1. Christensen, R. M. and Wu, E. M., Optimal Design of Anisotropie (Fiber-Reinforced)Flywheels, Lawrence Livermore Laboratory, Rept7 UCRL-52169, November 1976.

2. Toland, R. H., Current Status of Composite Flywheel Development, Lawrence LivermoreLaboratory* Rept. UCRL-806CA, January 1978.

3. G.E. Company, Volume I - Executive Summary — Demonstration of an Inductor Motor/Alternator/Flywheel Energy Storage System, Phase I - Final Report, January 27, 1978.

h. Lustenader, E. L. and Zorzi, E. S., "A Status of the a-ply Flywheel ConceptDevelopment," Proceedings of the Society for Advancement of Materials and ProcessEngineering, May 1978.

5. McGuire, D. P. and Rabenhorst, D. W., "Composite Flywheel Rotor/Hub Attachmentthrough Elastomeric Interlayers," 1977 Flywheel Technology Symposium Proceedings,October 5-7, 1977-

NOTICE"This report was prepared as an account of worksponsored by the United States Government.Neither the United States nor the United StatesDepartment of Energy, nor any of their employees,nor any of their contractors, subcontractors, ortheir employees, makes any warranty, express orimplied, or assumes any legal liability or respon-sibility for the accuracy, completeness orusefulness of any information, apparatus, productor process disclosed, or represents that its usewould not infringe privately-owned rights."

Reference to a company or productname does not imply approval orrecommendation of the product bythe University of California or theU.S. Department of Energy to theexclusion of others that may besuitable.

74

PROJECT SUMMARY

Project Title: Flywheel Technology for Automotive Use

Principal Investioator: B. B. Smith/F. W. Jones

Organization:

Project Goals:

Project Status:

Union Carbide Corporation Nuclear Div.Y-12 PlantP. 0. Box YOak Ridge, TN 37830(615) 483-8611 Ext. 3-7385

Assist DOE in development of composite flywheel technologyfor automotive use; conduct testing operations to assessperformance, determine failure modes, measure transientloads transmitted to vehicular-type containments; generateengineering design data for vehicular systems.

A prototype composite flywheel was designed, fabricated,and evaluated in FY 1O76-76T. An improved flywheel andan instrumented cc-^nment assembly were designed andfabricated in FY 1977. Testing of these components willbe performed in m e UCC-ND flywheel test facility builtin FY 1978. This facility is approximately 90% completed.

contract Number: W-7405-ENG-26

Contract Period: Oct. 1, 1977 - Sept. 30, 1978

Funding Level: $170,000 (FY 1978)

Funding Source: DOE, Division of Energy Storage Systems, AdvancedPhysical Methods Branch

75

FLYWHEEL TEST FACILITY

B. B. SmithUnion Carbide Nuclear CorporationP. 0. Box Y, Bldg. 9998, MS 001

Oak Ridge, Tennessee 37830

ABSTRACT

A flywheel test facility was built at the Y-12 Oak Ridge Plant during Fiscal 1978.The facility is designed for 2 KWH energy level with minimum speeds to 60,000 RPM. Thetest machine is located below grade in a concrete pit. The objectives for constructingthe facility are to provide a safe facility for pe forming dynamic testing while develop-ing a composite flywheel and to provide test data for containment and mounting studies.The facility is located tn an uncleared area to allow relatively easy access to industri-al representatives and other government agencies.

INTRODUCTION

A facility for testing and evaluationof high performance composite flywheelshas been installed at the Union CarbideCorporation-Nuclear Division, Y-12 Plant.The flywheel test machine is designed towithstand a torque of 400,000 ft. lbs.,and is mounted on a 12-inch high sub-base .to allow for quartz windows throughthe floor of the machine for futuredynamic studies using high speed photog-raphy. The 4.5 inches of containmentis comprised of two rotatable linersinside a 1.5 inch thick vacuum vessel.The inside diameter of the 2 inch thickinner liner is 37 inches. The vacuumsystem is capable of maintaining 10 mm Hg.

The test flywheel will be driven maximumtest speeds of 60,000 RPM with a 4-inchBarbour-Stockwell air turbine. The testfacility is designed for in-place bal-ancing using distance detectors, oscillo-scope, x-y plotter, and a trackinganalyzer.

A precision balance machine is availablefor pre-balancing flywheels prior to test-ing.

DESIGN CRITERIA

The design specifications were toprovide a spin test stand capable oftesting a flywheel in accordance with the

following specifications. The specifica-tions were also selected such that a nomi-nal 0.5 KWH vehicular flywheel could bedeveloped and failure tested at speeds andenergy levels consistent with full utili-zation of the high-performance compositematerials (i.e., 40 to 60 watt-hrs/lbs)without being limited by test standintegrity.

1. Maximum kinetic energy - 2 KWHNo combination of speed, weight, andgeometry can exceed a stored energylevel of 2 KWH [i.e., KE = Funct.(speed, weight, geometry) <_ 2 KWH].

2. Maximum flywheel O.D. - 30 in.3. Maximum RPM - 60,000 rpm.4. Maximum weight - 50 lbs.5. Vacuum - <10 mm Hg.6. Adequate ports and space for mirrors

and associated equipment to permituse of high-speed photography infuture activities.

7. Freely rotatable crash ring design,8. Constrained arbor design.9. 4" Barbour Stockwell turbine.

10. Capability to structurally attachand effectively test the FY77 fly-wheel and containment assembly.

i

76 ~

DESCRIPTION

The flywheels to be tested will bedriven with an air turbine mounted on thetop flange of a test machine as shown InFigure 1.

Figure 1. Cross Section of Test Machine

The horizontal test machine provides forquick access to the test flywheel througha 2-1/2 inch thick steel flange. The topflange is secured to the vacuum casingwith 12 one inch SAE grade 5 bolts.whichwill withstand an axial force of 188,000Ibf with a factor of safety of 4.

The test machine has two rotatablecontainment liners and a 1-1/2 inchthick casing wall. The machine ismounted to a concrete mass with 16 oneinch SAE grade 5 bolts. A sub-base 1sprovided to accommodate optical viewingof the flywheel through the bottomfiance for dynamic analysis and to studyfailure data of a composite flywheel.

A vacuum system will provide anatmosphere of about 10-' mm Hg.

The mass of the concrete base isgreater than the machine mass. Theconcrete mass is 43,000 1bsm.

flfflkTFI TBT FACILITY

DttlCN EMMY LEVEL

D H I M FACTO* » SAFETY

D M I M MOUNT T O M M

EXPECTED Hourr TOMUC

DESIfN MOMENT TO HOUHT

EXPECTED HOHENT TO IfOUNT

DfS IM RADIAL i M D

CONTAINMENT THtCKMM - HWIAL

Tor F L M K

BUI IWt

2 MM

H

400,004 rt-utf

150/000 FT-Up

287,000 rr-Up

253,000 FT -u ,

100,000 FT-Up

4.5 INCHES

2.5 INCHES

2 INCHES

Figure 2. Qe|1gn for Flywheel Test Facility

It IIC

1977

KINETIC EHERSY

VEI«HT

BURST ENEMY

OELAMINATION

SPEED • 1.12 nm

EXPECTED TOMUE

FLYMHEEL

DESIGN

1.12 KM

25 u,,

4 4 - H / U ,

<35,Q00RPN

38,350 RFM

123,840 FT-Up

IAXIHUM TORQUE • BURST SPEED 10,209 x EQ6 F T - U .

Figure 3. 20 Inch Flywheel Design Criteria

W1«CH

KINETIC ENEMY

HEISHT

SELAHIKATION COLLAPSE LOAD

SPEED 8 2 KMH

RADIAL LOAD F - NE»*

CASINS HOVEHENT CAUSED > Y O R J I T U MFLYNHEEL

nvmn

2KHH

50 Up

88,000 Up

22.458 RPN

2,507,968 Up

0.078 INCHES

Figure 4. 30 Inch Flywheel Design Criteria

77

v~;--'--i^tt::-J-^-:..-ii.--'

Figure 5. Flywheel Test Facility Figure 6. Flywheel Test Machine Mountedin a Concrete Pit

78

SESSION I I ; FLYWHEELS

79

PROJFXT SUMMARY

Project Title: Mechanical Energy Storage Technology Development

Principal Investigator: R. 0. Woods

Organization: Sandia LaboratoriesAlbuquerque, NM 87185(505) 264-7553

Project Goals: The selection of those mechanical components of flywheel energystorage systems that can be carried to their next stage ofevaluation by efforts within the scope of our budget; thefunding of developmental work treating such components. As acorollary, the documentation of the state-of-the-art inspecific component technology where such information is scatteredor fragmentary.

Project Status: Efforts to date have treated the following components:

Rolling Contact BearingsMagnetic Bearings - Active and PassiveComposite Wheels - Hardware development, development of analytic

and testing techniques.

Composite Materials - Analysis, improved properties, vacuumproperties, fabrication and thermal stresscharacteristics

Ferrofluidic Seals - Power dissipation, lifetimes, permeationrates

Funding has been provided to other agencies for the testing ofoverall energy storage systems,and the groundwork has beencompleted for a program to deal with the problens of vacuumtechnology.

Contract Number: 5OL 78

Contract Period: FY79

Funding Level: $1,107 M (BO) *

Funding Source: DOE, Division of Energy Storage Systems

*Included in this project are the following:E. D. Reedy, Jr . , "Sandia Composite-Rim Flywheel Development," page 87;A. Keith Miller, "Structural Modeling of a Thick Rim Rotor," page 93;Melvin R. Baer, "Aerodynamic Heating of High-Speed Flywheels in

Low-Density Environments," Page 99.

81

OVERVIEW OF COMPONENT DEVELOPMENT

Robert 0. WoodsSandia Laboratories

Division 1+715Albuquerque, KM 87185

ABSTBACT

Sandia Laboratories is charged with advancing components technology for flywheelenergy storage systems. Portions of this work are being done in-house, the balance oncontract with outside agencies. At this writing, seventeen specific efforts can beidentified. These comprise seven tasks under the heading of composite wheel develop-ment, four in bearings, three in vacuum seals, and three in general vacuum technology.Each of the major efforts will be represented at this conference by a paper presentedby a representative of the organization actually performing the task. The intent ofthis presentation is to give an overview of the program's organization and objectives.

INTRODUCTION

The functional components of a fly-wheel energy storage system are represen-ted in Fig. l. We have shown, by puttingthe labels in boxes, the subsystems whichhave been identified as needing furtherdevelopment and that have not been worked-over to the point where prohibitivelylarge expenditures will be needed toachieve minor quantitative improvements.These items are the ones upon which wejudge money will be spent to best advan-tage .

Figure 2 indicates the time frameprojected for the program. All of the ef-forts spanning FY78/FY79 are in fact underway. Negotiations for some (but not all)of those starting in FY79 have been ini-tiated. Note that late Spring 1979 repre-sents a well-defined transition. At thatpoint the most fundamental componentsstudies will have been completed; althoughdevelopment of selected components willcontinue, it will at that time be possibleto incorporate study results into the de-sign of a next-generation flywheel moduleintended for evaluation and demonstration.

WHEEL DESIGN

All efforts in wheel development havethus far addressed the problems of wheelsmade of composite materials. This is notonly because it was Sandia's experiencewith composites which first involved usin the flywheel effort. Composite wheelsare undergoing development because they

meet the criterion of not having alreadybeen evolved beyond a point of diminish-ing return. The alternative to the com-posite wheel is the steel wheel; therethe technology is well understood andmajor advancements are unlikely at ourfunding level. In addition, compositewheels may have attractive features suchas greater energy density and safetywhen compared to steel wheels.

Sandia Laboratories, in addition toits own in-house efforts aimed at develop-ing analytic techniques and at betterunderstanding the behavior of compositesin flywheel applications, is currentlyfunding the development of four new wheelconcepts. Each of those will be describedin detail by the contracting organization.Together, these concepts exhaust the rangeof geometries proposed to date. The tim-ing of these programs is such that preli-minary information regarding each of thedesigns will be available when engineer-ing of the demonstration module begins.

BEARINGS

These programs will also be de-scribed by the contractors. In the caseof rolling contact bearings, Sandia hasfunded a combined laboratory and theore-tical study which has explored the capa-bilities of what may be the next-genera-tion design for ball bearings. At thesame time, this study provides inputs toa computer code that will, for trade-offstudies, predict bearing operating char-acteristics over a range of parameters

82

such as speed, size and loading.

Two studies are also being funded toexplore the characteristics of magneticbearings — both active (servoed) andpassive. It is generally thought thatsuch bearings will find use only in sta-tionary applications. At some time inthe future, probably mid-1979> we hope toalso fund work on hydrostatic bearings.The results of such work could find ap-plication in a later generation flywheeldemonstration unit.

SHAFT SEALS

We have concluded, and it appearsthat our judgement is supported by amajority of the workers in the field,that for high-performance (read "highspeed") flywheel systems, energy inputand output will probably be electrical,using motor/generators sealed within thevacuum housing. If, however, it werepossible to produce rotating shaft sealscapable of the low permeation rates, highsurface speeds, and long wear lifetimesneeded for flywheel applications, a num-ber of options would be opened thatmight make electric input/output lookless attractive. With the current state-of-the-art in seals technology the pos-sibility of using a shaft and directly-coupled mechanical transmission existsonly for a limited range of applications.In the interest of establishing (ordemolishing) the credibility of mechani-cal input/output concepts, we are fundingtwo seals studies and hope to fund athird. The first two will establish theoperational envelope for present-genera-tion shaft seals and perhaps producenext-generation hardware. The later study,that of the so-called "dynamic" seal isproperly part of a vacuum technologyprogram which is not yet under way.

VACUUM TECHNOLOGY

One of the companion papers to bedelivered at this conference will reporton a study of the aerodynamic heating ofcomposite wheels. The unavoidable conclu-sion arising from this work is that thevacuum requirements of a high-performancecomposite wheel are stringent to thepoint of entailing a major effort if suchwheels are ever to evolve beyond the stateof a laboratory curiosity. The main pro-blems in vacuum technology are the dis-mally poor energy efficiencies of present-day pumping equipment, the outgassing

properties of the composites themselves,the effects of shock and vibration en-countered in a vehicle environment, andthe close tolerances (hence high prices)found in most rotating vacuum equipment.The only hard datum is that a vacuumon the order of 10"* torr will be re-quired if the potential of a compositewheel is to be realized. This informa-tion in itself is not adequate to speci-fy the problem. What is also needed isinformation regarding pumping rates atthat pressure. These will be controlledby the outgassing properties of the com-posites and the permeation rates of theseals. No reliable information exists,although we have made a first step byfunding a series of outgassing testswhich investigated representative compo-sites .

The current status of the vacuumtechnology program is that we are active-ly trying to locate organizations work-ing in the field and to interest themin undertaking all or any part of theproblem.

CONCLUSION

This is an open-ended program whoseobjective is to upgrade a wide range ofcomponents for flywheel energy storagesystems. There is no single touchstoneby which the success of the program maybe judged; rather, the objective is toproduce a continuous progression oftechnological improvements. By mid-1979it should be possible to incorporate theoutputs of the program as of that dateinto a demonstration unit. That unitshould represent an improvement over whatcould now be produced. At the veryleast, it will represent a permutationof components which can credibly becalled "optimized".

83

SHAFT SEALS• FERROFLUIDIC• FACE

MOTORS/GENERATORS

FAILURE SENSORS

CO

HEAT TRANSFER• AERO HEATING• INTERNAL

TEMP. FIELD

VARIABLE SPEED TRANSMISSIONS

POWER CONDITIONINGCIRCUITRY

WHEEL DESIGN• CARBON SPOKE/RIM• KEVLAR SPOKE/RIM• RIMLESS• SPECIAL CONTOUR• TEST & ANALYSIS

GsaSKVACUUM

• PUMPING• OUTGASSING• SEALS

CONTAINMENT

BEARINGS• PASSIVE MAGNETIC• ACTIVE MAGNETIC• ROLLING

'BACKGROUND TECHNOLOGY"SANDIA FLYWHEEL PROGRAM

00i n

ro

FY78

TASKS O N D J F K A M J J A S

WHEEL DESIGN

• StA Design and

•SLA Tes t ing « •• *•

• Conrcercial "hoW-nr/lb" Wheel

Carbon Spoke -

Kevlar Spoke -

R i a l e s s - Over-

Rialesa -

ContourTest and

Analysis

BEARINGS

•hydrostatic

SHAFT SEAIS

•Dynamic (cf."Vacuum")

VACUUM TECHHOMGT

• Pmnping

PROCCCWEDEVEIOFMEBT

•Engineering

•Fabrication

•Teatiag *aiDnonstratloa

FY79O N D J F M A M J J A S

FY80

O N D J F M A M J J A S

FY81

0 H D J F H A M J J A 8

SANDIA COMPOSITE-RIM FLYWHEEL DEVELOPMENT*

E. D. Reedy, Jr.Sandia Laboratories^

Composite Materials DevelopmentDivision 5844

Albuquerque, New Mexico 87185

ABSTRACT

This paper will briefly discuss a series of spin tests completed in March, 1978(just prior to the last DOE Flywheel Contractors' Meeting) and report on the status ofSandia's current rotor development program. Two designs incorporating a 20 in. O.D.graphite/epoxy rim were tested. Each design utilizes a different method of attachingthe rim to an aluminum hub with Kevlar 49/epoxy bands. Data on rotor performance,dynamics, and failure modes were gathered. Two rotors with a pin-wrapped hub designwere spun to tip speeds in excess of 2500 ft/sec. This corresponds to an energy den-sity, based upon total rotor weight, of approximately 20 Wh/lb. These tests have helpedidentify areas requiring further research. In particular, there is a need for predic-tive dynamic analysis for this class of fJywheel.

INTRODUCTION

A hybrid propulsion system whichutilizes a flywheel for energy storage re-quires a rotor which can both store energyefficiently and fail in an easily con-tained manner. A flywheel design whichincorporates a circumferentially wound rimwas identified as being likely to satisfythese design requirements as well aslending itself to existing fabricationtechniques. An analytic study was per-formed to identify feasible design andmaterial choices for a filament wound rimwhich is of a size appropriate for ahybrid vehicle.1 Based upon this analy-sis, two flywheels were designed to have(1) a total weight less than 25 lbs, (2)a rim swept volume of approximately 0.5cubic feet, and (3) a storage capacity of0.56 kWh at 31,000 rpm in order that 0.5kwh could be delivered with a 3 to 1 speedreduction. Both designs incorporate a 20in. O.D., high strength graphite/epoxy rimwhich has an inner-to-outer radius ratioof 0.7625. The rim has a semi-ellipticalcross section with its 3 in. high flatedge facing the hub. The designs differin the method in which overwrapped bands

This work was supported by the U.S. De-partment of Energy (DOE) , under ContractAT(29-l)-789.

A U.S. DOE facility.

attach the rim to an aluminum hub. Kev-lar 49/epoxy bands were chosen to matchthe radial displacement and tip speedcapability of the rotating rim.2 TheAlleghany Ballistics Laboratory of Her-cules, Inc. fabricated two rotors incor-porating their suggested pin-wrapped hubconnection (Fig. 1). The hub region of a

Figure 1. Pin-Wrapped Flywheel

Pin-Wrapped flywheel is shown in Figure 2.Two rotors which utilize six radial over-wrapped bands to connect rim and hub(Figs. 3,4) were fabricated by the De-fense Division of the Brunswick Corpora-tion. Due to its appearance, this rotordesign was designated the Wagon Wheel...

87

Figure 2. Pin-Wrapped Flywheel Hub

The results of the spin tests of theserotors as well as the current programgoals will be discussed below.

SPIN TESTS RESULTS

Four flywheels, two of each design,were tested at Sandia Laboratories, Liver-more. These tests are described and dis-cussed in detail in Ref. 3. The resultsof the tests are summarized in Table 1.The highest rotational speeds were ob-tained by the Pin-Wrapped rotors (29,000and 30,100 RPM). The lower speeds reachedby the Wagon Wheel rotors may be due, inpart, to the presence of pre-existing rimflaws which apparently grew during spin-up.Also, the first Wagon Wheel rotor tested(Test #2) experienced a migration and lossof balancing weights during spin-up. Themaximum tip speed was reached by a pin-wrapped rotor, 2625 ft/sec (Test #4). Atfailure this rotor stored 0.532 kWh, whichis within 5% of the design goal of 0.56kWh. This corresponds to an energy den-sity of 21.3 Wh/lb based upon total rotorweight.

Rotor dynamics were monitored byproximity gauges which measured the hori-zontal displacement of the adapter con-necting the flywheel to the spinner shaft.In all tests the flywheels initially spunstably as their speed was increased. How-ever, each test (except Test #1 in whichthere was a facility failure) was termi-nated by excessive shaft rurout. Thecause of this increasing 2r-»i runout hasnot yet been determined. It appears thatit could have been caused by either (1) anincreasing rotor imbalance due to a struc-tural failure or a migrating balancevreight, (2) approaching a critical speed,

Figure 3. Wagon Wheel Flywheel

Figure 4. Wagon Wheel Flywheel Hub

or (3) a combination of the above. Thereis evidence of an increasing rotor im-balance in Test #2 due to migratingbalance weights, in Test #3 due to fray-ing bands, and in Test #2 and #5 due togrowing rim flaws (see Table 1). On theother hand, calculations by C. W. Bertet. al.1* indicate the existence of criti-cal whirl speeds of 23,000 EPM for theWagon Wheel design and 23,500 EPM for thePin-Wrapped design. These predictionsmay be altered to give better agreementwith experiment when more accurate esti-mates for rotor stiffnesses become avail-able.

One rotor of each design was spununtil its runout failed the breakawayshaft and dropped it into the spin pit.Although the failures generated may notrepresent a spontaneous rotor burst, theymay be typical of those induced by agrowing flaw which causes increasing

38

TEST

1

2

3

4

5

TABLE 1 .

FLYWHEEL

Pin Wrapped#1

Wagon Wheel#1

Pin Wrapped#1

Pin Wrapped#2

Wagon Wheel#2

C O M P O S I T E - R I M

MAX SPEED (RPH)

23,800

17,900

29,000

30,100

22,100

COMMENTS

Terminated by facility failureBands abraided by instrumentation debris

Pre-existing surface flaws (rim)Terminated after balance weights thrown offRim separations

Same flywheel used in Test #1Terminated by excessive shaft runoutBands frayed where abraided in Test #1

Best performanceDriven until shaft failed

Pre-existing internal flaws (rim)Driven until shaft failedRim separations

runout to break the shaft. In both tests,the bands were ripped from the rim andshredded. The rim remained intact (Fig.5). This failure mode is conducive tocontainment since there are no pieces withhigh radial momentum.

CURRENT PROGRAM

Since, as discussed above, the originof increasing shaft runout is not known,its identification is a principle goal inthe current program. An intensive effortis underway to better understand rotordynamics. Bert et_. al^.1*'5 are usinglumped mass techniques which use calcu-lated estimates for distributed band,shaft, and support flexibilities to modelthe rotor. His calculations incorporategyroscopic effects and centrifugal bandstiffening. A modal analysis of the WagonWheel design is being pursued by Miller.6

His finite element calculations are beingcalibrated against the results of anexperimental modal analysis. This experi-ment was performed on a non-rotating fly-wheel with traction-free boundaries. Thenext series of spin tests will attempt toidentify the source of the excess shaftrunout. One facet of these tests will beto try to observe experimentally theanalytically predicted changes in theWagon Wheel's dynamics when the bandstiffness is increased or the hub's dia-metrical moment of inertia is decreased.

In addition to dynamic analysis, amore detailed structural analysis of theflywheels is planned. In particular, a

3-D finite element model is being used tostudy the interaction between the rim andbands. Such analysis could be used tooptimize the choice of band material andrim geometry.

Methods for predicting and control-ling rim fabrication stresses are beingpursued as this may provide a .•neans ofextending a rim's operating speea. Anexperimental program to measure rim re-sidual stresses is underway. In thesetests the hoop strain at the rim I.D. ismonitored as layers of the outer surface

Figure 5. Failed Pin-Wrapped Flywheel(Test #4)

89

are machined away. These data can beused to infer the initial radial stressdistribution. Predictive analytic methodswhich correlate with the experimental dis-tributions are being sought.

Techniques for measuring and predic-ting a rotor's aerodynamic heating arebeing developed. As pointed out by Woodset. al.,7 aerodynamic heating may be par-ticularly severe for a resin matrix com-posite flywheel. These composites cansurvive only within a limited temperaturerange. To further compound the problem,hoop wound composites have a poor radialthermal conductivity which tends to allowthe temperature to build up on the peri-phery. Analytic methods similar to thoseused to analyze re-entry vehicles havebeen developed by Baer8'9 to calculate theeffect of aerodynamic heating on a com-posite flywheel. Initial experimentalefforts to measure rim temperature by on-board thermocouples during rotor spintests have proven difficult.3 At highrotational speeds lead wires and gages areripped from the rim. Recently, an infra-red radiometer was added tc ^andia Liver-more 's spin test facility in order thatrim temperatures can be monitored withouton-board instrumentation. The data col-lected will be useful in verifyinganalytic predictions.

A materials program is being con-ducted by Allied, et. al. 1 0' 1 1 to bothcharacterize and improve through materialsmodifications the transverse strength offilament wound composites. Often thetransverse strength of such materialslimit the energy storage capacity of acomposite-rim flywheel design. Theeffects of (1) matrix toughening, (2)resin/hardener chemistry, and (3) fila-ment surface treatment on transversestrength are being studied. One aspectof this work was the successful develop-ment of a self aligning specimen fixtureto test hoop wound tubes in transversetension.

SUMMARY

The favorable rotor performance andfailure mode encourage the continued de-velopment of composite-rim flywheels.Future work will.concentrate on gaininga better understanding of (1) rotordynamics, (2) rim-band interactions, (3)rim residual stresses, (4) flywheel aero-dynamic heating, and (5) the factors whichaffect composite transverse strength.

REFERENCES

1. E. D. Reedy, Jr. and F. P. Gerstle,Jr., "Design of Spoked-Rim CompositeFlywheels," Proceedings of the 1977Flywheel Technology Symposium, SanFrancisco, CA, Oct. 5-7, 1977, pp.99-110.

2. E. D. Reedy, Jr., "A Composite-RimFlywheel Design," SAMPE Quarterly,Vol. 9, No. 3, pp. 1-6.

3. E. D. Reedy, Jr., and H. K. Street,"Composite-Rim Flywheels: SpinTests," to be published.

4. C. W. Bert and T. L. C. Chen, "Lateraland Tilt Whirl Modes of FlexiblyMounted Flywheels: Analysis and Ex-periment," Presented at the 49th Shockand Vibration Symposium, Washington,DC, Oct. 17-19, 1978.

5. c. W. Bert, T. L. C. Chen, and C. A.Kocay, "Critical Speeds and NaturalFrequencies of Rim-rype Composite-Material Flywheels," OU-AMNE-73-3,September, 1978.

6. A. K. Miller, "Structural Modeling ofa Thick-Rim Rotor," Proceedings of theFirst Annual Mechanical and MagneticEnergy Storage Contractors' Informa-tion-Exchange Conference, Luray,Virginia, October 24-26, 1978.

7. R. 0. Woods and F. P. Gerstle, Jr.,"Sandia Basic Flywheel TechnologyStudies," Proceedings of the 1977 Fly-wheel Technology Symposium, San Fran-cisco, CA, Oct. 5-7, 1977, pp. 315-322.

8. M. R. Baer, "Aerodynamic Heating ofHigh-Speed Flywheels in Low-DensityEnvironments," Sandia Report, SAND 78-0957, October, 1978.

9. M. R. Baer, "Aerodynamic Heating ofHigh-Speed Flywheels in Low-DensityEnvironments," Proceedings of theFirst Annual Mechanical and MagneticEnergy Storage Contractors' Infor-mation-Exchange Conference, Luray,Virginia, October 24-26, 1978.

10. R. E. Allred, R. F. Foral, and W. E.Dick, "Improved Performance forHoop-Wound Composite Flywheel Rotors,"Proceedings of the 1977 Flywheel

90

Technology Symposium, San Francisco,CA, Oct. 5-7, 1977, pp. 377-392.

11. R. E. Allred and H. K. Street, "Im-provement of Transverse CompositeStrength: Test Specimen and MatrixDevelopment," Proceedings of the 24thNational SAHPE Symposium, San Fran-cisco, CA, May 8-10, 1979.

91

STRUCTURAL MODELING OF A THICK-RIM ROTOR*

A. Keith MillerApplied Mechanics Division 5521

Sandia LaboratoriesAlbuquerque, New Mexico 87185

ABSTRACT

A NASTRAN structural model has been constructed of the Sandiadesigned, thick-rim rotor having six discrete overwrapped bands formingtwelve pairs of spokes—the wagon-wheel rotor. The results of an experi-mental modal analysis of an actual rotor, using Past-Fourier transformtechniques, are being used to aid in the definition and refinement of thenumerical model. A description of the structural model is given in thiswork, and the resulting calculated normal mode shapes and frequencies arepresented. These mode shapes and frequencies are compared to thoseobtained from the experimental modal analysis.

INTRODUCTION

The objective of the program isto develop a flywheel rotor capableof being used in hybrid heat-enginevehicles. One promising proposeddesign of such a rotor involves athick-rim, composite rotor havingdiscrete composite spokes attachingthe rim to an internal hub. Theproposed rotor, shown in Fig. 1, has

Fig. 1. Thick-Rim Wagon-Wheel Rotor

This work was sponsored by theUnited States Department of Energy,Dr. Robert 0. Woods, programmonitor.

an outer diameter of approximatelytwenty inches and weighs approxi-mately twenty three pounds. Theflywheel is designed to store 0.5kWh of energy when rotating at32,000 rpm.

Because the success of anyrotor design is ultimately deter-mined by the dynamic performance ofa flywheel system, it becomes nec-essary to attempt to understand thedynamic characteristics of a rotorstructure before it is incorporatedinto the larger system. Once thestructural dynamic characteristicsof the rotor are properly identi-fied, then one should be able todetermine, with some confidence,the interaction between a flywheelsystem and a specific roto:.. Oneapproach to quantifying tne struc-tural characteristics of a flywheelrotor, in a manner such that theinformation can be used in a seriesof flywheel system designs, is toconstruct a numerical, finite-element model of the rotor usingexisting general purpose computercodes, such as NASTRAN.

Experimental modal analysis ofa wagon-wheel rotor has been usedto verify the various assumptionsinvolved in the assembly of thenumerical structural model (espe-cially items; involving connectingregions). When the numerical model

93

predicts mode shapes and frequenciesof the rotor structure which arewithin reasonable agreement with theexperimental results, it is believedthat the structural model is suffi-ciently accurate to be used infuture analyses of flywheel systems.

ROTOR COMPONENTS

A sectioned view of the wagon-wheel rotor is shown in Fig. 2.The wagon-wheel consists essentiallyof three components: the circularwound, carbon/epoxy rim; the uni-directional Kevlar/epoxy spokes;and the hub which includes thealuminum cylindrical hub, the alumi-num spin collar adaptor plate, thecatcher plate and plug, and thesteel spin collar. The mechanicalproperties of the graphite/epoxy,the Kevlar/epoxy, and aluminum usedin the structural model are givenin Table 1.

The rim, as shown in Fig. 2,has a semi-elliptical cross sectionhaving a semi-major axis of 2.4inches directed radially, and asemi-minor axis of 1.5 inches. Theinner radius of the rim is approxi-mately 7.6 inches.

The twelve overwrapped Kevlar/epoxy spokes have a nominal widthof 0.5 inches, and a nominal thick-ness of 0.1 inches. The spokeswere measured to be approximately50% thicker than the nominal valueon the rotor used in the experimen-tal model analysis. (The measuredthickness of the spokes was used inthe numerical model of the rotor.)The spokes are continuous, and passthrough the hub in a series of slots

- SPIN COLLAR

-WOUHTIHG I1LOCK

- CMrnrft r-tuG

Fig. 2. Cross-Sectional View ofthe Wagon-Wheel Rotor

which are machined into the ends ofthe hub.

The aluminum hub is a cylinderhaving an inner diameter of 3.5inches and a wall thickness of 0.5inches. The hub is 4.3 inches longbut has slots of depths rangingfrom 0.15 to 0.65 inches machinedfrom the ends to allow the over-wrapped spokes to pass through.The aluminum catcher plate is 2.5inches thick and 4.5 inches indiameter with a 0.6 inch diameterattached ball affixed to one side.The aluminum spin collar adaptorplate is 1.5 inches thick, with aninner diameter of 1.75 inches andan outer diameter of 4,5 inches,and is affixed to the opposite end

Table 1. Mechanical Properties of Materials Usedin the Thick-Rim Wagon-Wheel Rotor

Property

Longitudinal Modulus E.(106 psi)

Transverse Modulus E_(106 psi)

Shear Modulus G(106 psi)

Density (lb/in3)

Graphite/Epoxy

18.00

1.23

0.85

0.054

Kevlar/Epoxy

11.00

0.71

0.30

0.050

Aluminum

10.50

10.50

3.80

0.100

94

of the hub. To this is attachedthe steel spin collar which is essen-tially a hollow truncated conicalsection having a major outer diam-eter of 4.5 inches, a minor outerdiameter of 1.75 inches, an innerdiameter of 1 inch, and is 2 incheslong. All dimensions are approxi-mate.

More detailed dimensions anda description of the fabricationprocedure of this wagon-wheel rotorcan be found in Reference 1.

NASTRAN MODEL

The finite element structuralmodel which is currently being usedto describe the wagon-wheel rotorconsists of a series of beam ele-ments, plate elements, and solidelements. The graphite/epoxy rimis modeled as a series of twelvebeams each having a cross-sectionalarea and area second moment equiv-alent to the physical rim. Actualorthotropic material properties ofthe graphite/epoxy are used forthese beam elements.

Each of the twenty four Kevlarspokes were also modeled as a seriesof beam elements again having ortho-tropic mechanical properties. Thehub was modeled as a series of plateelements consisting of four ringswith twelve elements defining eachring. The catcher plate, the spincollar adaptor plate, and the spincollar were all described in themodel by three-dimensional solidelements.

The modulus and mass propertieswere included as material propertiesfor the beam elements of both therim and the spokes. The computercode then distributes the mass ofeach element equally at each nodepoint at the ends of the bars. Onlythe modulus of the material used inthe solid elements was input as amaterial parameter. Equivalentmasses were directly input to thenode points describing the hubassembly so that the moment of iner-tia of this assembly could beadjusted to better model the struc-ture. The masses are not distri-buted precisely enough for the hubassembly when done by the NASTRAN

algorithms. The moment of inertiaabout a diametral axis for the hubassembly is normally too large whenthe computer algorithms distributethe mass to the node points.

EXPERIMENTAL MODAL ANALYSIS

During the experimental modalanalysis, a thick-rim, wagon-wheelrotor was suspended from the catcheradaptor plate by surgical tubing tosimulate a free condition of therotor structure. The natural fre-quency of the rotor on the tubingwas roughly measured to be 1.2 Hz.A mounting block (See Fig. 2) wasattached to the catcher platethrough which a circumferentialimpulse was input to the rotor hubvia a hard plastic hammer having aforce transducer mounted to thehead.

Triaxial accelerometer blockswere mounted at ninety degree incre-ments around the inside of the rim,and at the top and the bottom of thehub assembly.

As an impulse was supplied tothe structure, the time responsesof the accelerometers were measuredat a sampling rate of 5000 Hz (onceevery 0.2 milliseconds). The res-ponses were averaged from ten im-pulses, and experimental transferfunctions for each accelerometerwere obtained from a Hewlett-Packard5451B Fourier Analyzer, using a_ HPsoftware package. Fig. 3 is a typi-cal plot of an experimental transfer

0.9

0.4

0.3

0.2

O.I

3 o

2-o.i

-0.2

-0.3

-0.4

-0.5

1

1 k

1.0

.8

.6

.4

.2

] 01 - . 2

-.4

-.6

600 1000 1500 2000(HZ)

REAL

i

i1r

i

B00 1000 1900 2000(HI)

IMAGINARY

Fig. 3. Experimental TransferFunction on the Spin-Collar-End ofthe Hub for the Wagon-Wheel Rotor

95

function; this particular one isfor an axial accelerometer mountedto the spin collar side of the hubassembly. The plot on the left ofFig. 3 represents the real portionof the transfer function, while theplot on the right represents theimaginary portion. Once the experi-mental transfer functions areobtained, a series of analytic trans-fer functions are "fit," with theaid of the Fourier Analyzer, tomatch the experimental ones. The"fitted" transfer functions are theones from which the model frequen-cies and damping factors are actuallyread.

The peaks in the transfer func-tions, such as are shown in Fig. 3,locate the frequencies of the reso-nant vibration modes in the struc-ture. The amplitude of a particularpeak gives an indication of therelative strength of that mode. Thewidth of a peak is related to theamount of damping in the structureassociated with that mode.

The modes for the wagon-wheelrotor were, in general, found to bevery distinct and nonoverlapping,which indicates that the structurewas behaving essentially linearlyfor the input excitation used.Although the normal modes up to2000 Hz can be identified from thetransfer function shown in Fig. 3,only those below 1000 Hz (60,000rpm) are examined. Modes at fre-quencies above 1000 Hz are consi-dered to be sufficiently above theupper limit of the proposed oper-ating range that they will probablynot be excited during the rotor'soperation. Therefore, only thefrequencies of the first four normal

modes, and descriptions of the modeshapes determined from the experi-mental analysis are given in Table2.

The fourth mode, which is awhirl mode of the hub with respectto the rim, may be one of some con-cern. Although the frequency ofthis mode (630 Hz) is beyond theproposed operating range for therotor, it is important to considerthe fact that this mode was identi-fied while the rotor was in a staticcondition where centrifugal stiffen-ing of the spokes is not present(which may raise the frequency ofthis mode), and while the rotor isin a free condition—not attachedto the air turbine used in the spintesting. When the rotor is attachedto the air turbine the frequency ofthis mode will be lowered becausethe additional flexibility of theturbine shaft is in series with therotor spokes for this mode shape.

It is presently unclear howstrongly centrifugal stiffening ofthe spokes will affect the whirlfrequency. However, a preliminaryanalysis using the present NASTRANmodel of the rotor and a currentestimate of the air turbine struc-ture indicates that the whirl fre-quency may be decreased as much as30 percent when the rotor is pendu-lous ly attached to the turbine. Ifcentrifugal stiffening of the spokesdoes not increase the whirl frequen-cy, a 30 percent reduction of thisfrequency places it near the 25,000rpm range. As was noted by Reedy3,previous spin tests of a wagon-wheelrotor indicated possible dynamicinstabilities near this angularspeed.

Table 2. Frequencies and Normal Modes of theThick-Rim Wagon-Wheel Rotor

Experimental Numerical

ModeNumber

1234

Frequency(Hz)

102376605630

ModeNumber

12

Frequency(Hz)

90373

634

Description

Axial motion of hub w.r.t. rimTorsional motion of hub w.r.t. rimHub oblonging (highly damped)Whirl of hub w.r.t. rim

96

The whirl mode is also the oneeasily excited by gyroscopic momentswhich develop from mass eccentrici-ties as a rotor is spun. Therefore,it is highly desirable to have thefrequency of this mode as far aspossible from the operating range.The relatively massive catcher plateand plug were removed from the rotorto investigate the possibility ofraising the frequency of the whirlmode. The experimental analysis wasrepeated and the whirl frequency wasfound to be raised 80 Hz (13%) whilethe rotor was in the static, freetest state.

NASTRAN MODEL NORMAL

The frequencies of the firstthree normal modes and descriptionsof the mode shapes calculated fromthe NASTRAN model are also given inTable 2. In general, the frequen-cies predicted by the model agree tobe within 10 percent of the experi-mental modal frequencies.

The structural model does notpredict a mode shape as observed inthe experimental analysis for modenumber three (605 Hz). Instead, theshape of mode number three predictedby the numerical model correspondsto the shape of mode number fourfrom the experimental analysis.When a particular mode is predictedby the numerical model which is notobserved in the experimental analy-sis, an immediate question arises ofwhether the mode was sufficientlyexcited by the chosen mechanicalinput to be detected by the trans-ducers. In this situation, an inputis selected which will sufficientlyexcite the mode, if it exists.Normally the mode does exist and isdetected when the alternate mechani-cal input is chosen. However, whena mode is detected by the experimen-tal analysis which is not obtainedfrom the numerical model (as in thecurrent case), then both the modeland the experimental data must bere-examined. Because the predictedmodal frequencies from the numericalmodel are in good agreement with thecorresponding frequencies of theexperimental analysis for the remain-ing mode shapes, a re-examination or

the transfer function shown in Fig.3 seems appropriate.

In Fig. 3, both of the peaks inthe real and imaginary portions ofthe transfer function associated withthis questioned mode (at 605 Hz) areof relatively small amplitude, indi-cating the response of the rotor atthis frequency is not great. Also,the peak in the imaginary trace iswide, indicating a large amount ofdamping is associated with this mode.The mode at 605 Hz is obviously notstrong, nor is it a clean mode. Thepeak may be either the result ofsome nonlinear behavior of the rotorstructure when it responds to themechanical impulse input or a smallbeating mode resulting from twohigher frequency modes which areexcited by the impulse. In anycase, this mode is not of seriousconcern. If the mode is the resultof some nonlinear response, thestructural damping associated withit is large enough that th° responseto possible excitations should remainsmall; or if the mode is the resultof two higher frequency modes, thosehigher frequency modes are well abovethe proposed operating speed of therotor and will not be excited; there-fore, no beating will exist.

It was discovered during theassembly of the structural modelthat the distance from the planeformed by the rim (plane A-A inFig. 2) to where the spokes attachto the hub strongly influenced thewhirl frequency predicted by themodel. The further from this plane(plane A-A) that the spokes werefastened to the hub, the lower wasthe whirl frequency. The pointswhere the spokes connect to the hubin the actual rotor vary in distancefrom the plane of the rim. Theinclination of the spokes vary inuniform increments from being essen-tially parallel to the rim plane,to vectors outwardly inclined appro-ximately four degrees when trans-versing from the rim to the hub.It is anticipated that the resultsof this discovery will be furtherconsidered if it becomes necessuryto increase the whirl frequency forthis design of rotor.

97

FUTURE EFFORTS

The thrust of the effort todate has been to assemble a struc-tural model of the thick-rim, wagon-wheel rotor which is accurate andcan be used reliably in futurestudies of flywheel systems. Theimmediate future effort will be toincorporate the current rotor modelinto a larger structural model con-taining the spin pit dynamic char-acteristics. The enlarged modelwill be used in an attempt to pre-dict possible dynamic instabilitiesas the rotor is spin tested. Itemssuch as the effects of centrifugalstiffening of the spokes and theeffects of removing the catcher plugon the dynamic characteristics ofthe flywheel system are expected tobe addressed. Frequency responsestudies of various force inputs tothe flywheel system are also anti-cipated.

The structural model can alsobe used to investigate the dynamiceffects of possible design changesof components of the rotor inadvance of fabrication of a newrotor. Dynamic control studiesbased on accepted structural modelsare also possible future efforts.

ACKNOWLEDGEMENTS

The author wishes to thankMr. A. R. Nord and Mr. C. M. Grasshamlor the excellent experimental modalanalysis. The author also wishes toacknowledge the many useful discus-sions with Dr. E. D. Reedy concern-ing the structural modeling of thewagon-wheel rotor.

REFERENCES

JE. D. Reedy. SAMPE Quarterly, 9,3,^(April 1978)2E. D. Reedy, Proceedings of theFirst Annual Mechanical and Magne-tic Energy Storage Contractors'Information-Exchange Conference,Luray, Virginia (1978).

98

AERODYNAMIC HEATING OF HIGH-SPEED FLYWHEELS INLOW-DENSITY ENVIRONMENTS*

Melvin R. BaerFluid Mechanics and Heat Transfer Division 5512

Sandia LaboratoriesAlbuquerque, New Mexico 87185

ABSTRACT

This study addresses the aerodynamic heating of high-speed flywheels in low-density environments. A computer code has been developed to predict temperaturefields in flywheels of variable geometry and consisting of multiple composite materialswith nonisotropic, temperature-varying thermal properties. Allowances have beenincorporated for variable environmental conditions, time-varying spin rates, and thechoice of slip-flow or free-molecular aeroheating. Major results from the code indi-cate that environmental pressures below 10~3 torr are necessary to avoid steady-statetemperatures exceeding 50°F above ambient. Protruding surfaces have the majorpotential to cause thermal problems.

NOMENCLATURE

a speed of soundCp specific heatG gap distance between flywheel and

housingk thermal conductivityp gas pressurePr gas Prandtl numberq heat fluxr radial coordinater1 recovery parameter5 reduced Mach numberSt Stanton numberT temperaturez axial coordinatea. accommodation coefficienty ratio of specific heats6 continuum boundary layer thicknesse emissivityH gas viscosity\ gas mean-free path(T Stefan-Boltzmann constant

Sandia Laboratories is a U. S.Department of Energy (DOE) facility.This work was supported by the USDOEunder Contract AT(29-l)-789.

flow angle of incidenceflywheel angular velocity

SUBSCRIPTS

crHWCO

f1

convectiveradiationhousing inner surfaceflywheel surfaceambientfiber directionnormal-to-fiber direction

INTRODUCTION

The introduction of high-strength,low-density materials such as Kevlar-49or other fiberglass/carbon-fiber com-posites has greatly improved the energystorage capabilities of high-speed fly-wheels. However, the requirements ofa long system lifetime and a high power-conversion efficiency constrain their useas a storage unit. At high rotationalspeeds, aerodynamic considerations areimportant because of the induced parasitic

99

frictional losses and the associated ir-reversible conversion of kinetic energyto heat.

It has been suggested that theseundesirable effects can be virtually elim-inated by operating the system within alow vacuum environment (below P < 10"^torr). ' However, this vacuum require-ment may be impractical or unrealisticfor some systems. For example, out-gassing characteristics of the organiccomposite flywheel may limit the achiev-able degree of vacuum.

Although drag losses can be re-duced, heat-transfer effects may besignificant since aerodynamic heatingoccurs over a long operating lifetime,and this integrated heating determinesthe temperature fields within the flywheel.The heating effect is of major concernbecause composite materials are heat-sensitive and experience severe degrada-tion of strength when the temperatureexceeds 150°F.2 Also, the possibilityof a pressure runaway condition existsbecause of an increased outgassing rateat elevated temperatures.

This study addresses the predictionof temperature fields within high-speedflywheels spinning in low-density environ-ments. A computer code was developedfor this purpose that incorporates theallowances of composite materials withnonisotropic temperature-varying prop-erties, variable environmental conditions,time-varying spin rates, and differentaxisymmetric geometries. Steady-stateor transient time calculations can bedetermined for slip-flow or free-moiecular aeroheating. Results pre-sented herein typify the predictions ofthe heat-transfer code.

THEORY

Figure 1 shows a typical flywheelgeometry examined in this study. Understandard conditions of pressure and tem-perature, a> continuum boundary layerflow exists adjacent to the flywheel sur-face as it is spun. Schlichting3 has

defined the laminar boundary layer thick-ness for the rotating flow as

Note: This characteristic dimension isindependent of flywheel radius.

COMPOSITE FLY WHEEL

Fig. 1. The geometry of the flywheelsexamined in this study.

As the pressure within the housingis dropped (maintaining the ambient tem-perature), the mean-free path betweenconsecutive molecular collisions becomeslarger. For an ideal gas, this dimensionis given as

X = 1.26N/yji/pa

Slip and/or transitional flow is en-countered when the mean-free path ap-proaches the boundary layer thickness.At this condition the boundary layer isdiffuse, the flow appears to be "slipping"along the rotating surface, and the gasexperiences a temperature discontinuityat the gas-surface interface.

Free-molecular flow is reachedwhen the mean-free path is large. Forthis rarefied flow, momentum and energyexchange occur strictly from molecular-surface interactions, and the incident

100

flow is undisturbed because of the pres-ence of the rotating surface.

Figure 2 depicts the subdivision ofthe flow regimes gauged according to theratio 6/X. Under conditions suggestedfor practical application (~10-2 to 10-5torr), the noncontinuum flow regimesare of most importance and the heattransfer associated with free-molecularand slip/transitional flow follows in thenext sections.

flux of energy is deduced from the ac-commodation property of the heatedsurface (determined experimentally andrelated to the rate if the molecules werereemitted with a Maxwellian velocitydistribution at the heated surface tem-perature). Typical values for the accom-modation coefficient are a = 0.85 to 1.0.The heat flux into the flywheel is thenrepresented in terms of a modifiedthermal recovery factor, r \ and aStanton number. Sj:

sI

1 rfr

•20,000 rpm

CONTINUUM FLOW

101 lO2 id 3 io<GAS PRESSURE (lorrl

105 W6

Fig. 2. Flow regimes at various ambientpressures and spin conditions.

FREE-MOLECULAR AEROHEATING

and

The expressions for r1 and St are givenin the Appendix.

SLIP/TRANSITIONAL HEAT TRANSFER

Within the pressure range p = 10" *to 10"3 torr, the effects of intermolecu-lar collisions begin to be important.Contrary to free-molecular flow, the gapbetween the housing and the flywheel isimportant since a conduction path is pro-vided to the surrounding ambient tem-perature.

To treat the heat transfer for slipflow, Couette flow between two parallelmoving surfaces of different tempera-tures4 is examined (refer to Fig. 3).The following heat-transfer relationshipincorporates the heat-conduction effectwith the viscous dissipation in the flow:

The convective heat transfer to arotating flywheel in the free-molecularrange is determined by the influx ofenergy transmitted by the bombardmentof molecules with the surface and is cor-rected for the energy that is reemittedinto the incident stream. This reemitted

C (1+2X/G)2G

Derivation of the above equation requiresthat the accommodation coefficient, a,and the momentum transfer coefficient,c , are uni ty .

101

Flywheel rotor

Fig. 3. Flywheel-housing geometry usedin the slip-transitional flow aeroheatingrelationship.

THERMAL RADIATION CONSIDERATION

The radiant interchange between theflywheel and housing is of major impor-tance in the heat-transfer modeling be-cause it equilibrates the system to tem-peratures much lower than the recoverytemperatures of the flow. Without theradiation cooling, the steady-state tem-peratures would approach the recoverytemperatures that are typically 1000°F.

The radiant exchange between graysurfaces5 is given as

W

where e w» £ H a r e respectively thewheel and housing emissivity and a isthe Stefan-Boltzmann constant.

Note: To maximize radiation cooling, ahigh emissivity of the housing is re-quired.

COMPUTER MODELING

The previous heating relationshipswere incorporated in a con^-wter codethat analyzes the steady-state or tran-sient temperature field within axisym-metric flywheel geometries. Optionsto the code include the allowances ofvaried geometry (including variablethickness), composite materials, non-steady spin rates, flow regime (pureslip flow or pure free-molecular flows)with variable gas properties, excessheating at the rotor tip, and nonisotropictemperature-dependent materials. Theresults that follow in the next sectiontypify the predictions of the case.

RESULTS

The first series of results are theequilibrium predictions of maximumtemperature for a flywheel composedof a material of low thermal conductivity(an adiabatic surface results from azero-valued thermal conductivity). Sur-face temperatures are determined bymatching the convective aeroheating tothe radiation cooling.

Displayed in Fig. 4 are thesemaximum temperatures at variousangular speeds at the tip of a flywheelhaving a radius of 1 ft and subjected tofree-molecular aeroheating. The heatingeffects are small below a gas pressureof 10~5 torr. For an air environment ata pressure 10~3 torr, the free-molecularcalculations predict a rise in temperatureof 50 °F above ambient at a speed between40, 000 to 50, 000 rpm. For other radii,the tip-speed velocity should remain thesame, with the angular speed adjustedaccordingly.

A modificftion of the above resultsis necessary for aeroheating surfaces notparallel with the incident flow. Someexisting flywheel designs incorporatesurfaces oriented 90° to the incident flow.For example, Fig. 5 depicts a portion ofa flywheel that has an outer wrap as asupporting spoke. This effect of surfaceorientation in the aeroheating is displayedin Fig. 6.

102

700.

600.

rr- HO.

|

§400.

§300.

I

'p-10'1 torr

Air envlromentFlywheel radius • 1 '

a "1.0; «"1.0ADIABATIC SURFACE

200. -

100. .

30.000 40,000 50.000 60,000 70.000 80,000ANGULAR SPEED (rpm)

Fig. 4. Equilibrium surface tempera-tures at various angular speeds andambient air pressures.

i - 01

\

spoke

\

Incident flow

Pig. 5. A flywheel design that incor-porates a spoke surface oriented 90°to the incident flow.

1000

500

50

9-90° (Air)

»"0° (He)

Flywheel radius - 1 ft• 40.000 rpm

io4 io3

AMBIENT PRESSURE (torr)

Fig. 6. Maximum temperature atvarious ambient pressure and flowincidence angles for environments ofair or helium.

Also included in the same figureis the effect of introducing an environ-ment of a low molecular weight; i. e . ,the inert gas helium. Hydrogen is per-haps the best environment from a molec-ular weight point of view; however, itshighly reactive nature may cause prob-lems.

The next section of results (Figs.8 to 12) includes the effects of thermalconductivity in a flywheel geometryrepresented in Fig. 7. Thermal prop-erties of a Kevlar-49/epoxy compositeare used in the calculation and are givenin the Appendix.

Fiber orientation plays an impor-tant role in this modeling since thethermal conductivity changes by an orderof magnitude with direction. For ex-ample, circumferentially orienting thefibers places low thermal conductivityin the r-z directions, and the tempera-ture fields are well-represented by theadiabatic predictions of the previoussection. The higher thermal conductiv-ity is in the fiber direction, and itseffect to.relax the temperature fields is

103

30,000 rpm

Fig. 7. The flywheel geometry used inthis thermal analysis.

insignificant since there are no thermalgradients in the azimuthal direction dueto symmetry. However, a radial orien-tation of the composite fibers imposes ahigher radial thermal conductivity andthe radial distributions of temperaturebecome flattened. Figure 8 comparesthe surface distributions for various k rand kz values. Linear profiles are pre-dicted, and the profiles cross at a com-mon location that divides the region atwhich the aeroheating dominates theradiation cooling, and vice versa.

k <BH)ltllhr> m

0.7 0.B OS 1.0

NORMALIZED RADIAL COORDINATE I i V l

Fig. 8. Radial surface temperatureprofiles for various thermal conduc-tivities.

As suggested earlier, the effect ofprotruding surfaces may have a pro-nounced impact on the temperature fields.To investigate this behavior in a two-dimensional model, the outer surfaceat the rotor tip was impressed with a

convective heating load correspondingto a 90° angle of incidence to the flow.This modeling approximates the tem-perature field near a supporting spoke.The surface temperatures with and with-out excess heating are shown in Fig. 9.

tz.

E

UTI

CO

in

-ua*

1

;

220

200

180

160

140

120

100

-

-

Excess heating at outer radius (k •

-

" Excess heating at outer radius (k -1.0: k-0.11

1> '—— /y

_ _ . - - * " NO excess heating (k

-

i

/ ^y

-

-r - l . 0; I^-0.1)

-

0.9

NORMALIZED RADIUS r/r

1.0

Fig. 9. Radial temperature profiles forvarious flywheels with and without excessheating at the rotor tip.

Representative times to reachsteady-state are shown in Fig. 10.Normalized temperatures; i. e . , tem-perature rise above ambient to maximumtemperature difference, are plotted forgas pressures of 10~2 and 10-3 torr.Steady-state conditions are reachedafter ~ 1 hr of continuous spin. Whenexcess heating is imposed at the outerperiphery, additional time is necessaryto relax the temperature fields.

Both a uniformly thick plate geo-metry and the temperature distributionwithin a Stodola disk geometry in whichthe thickness varied inversely to thesquare root of radius were examined todemonstrate the effect of variable fly-wheel thickness. The temperature pro-files are shown in Fig. 11. Near therotor tip of tne Stodola disk, less

104

cross-sectional area exists to conductthe heat radially; consequently, the tem-perature fields are higher.

P-O.OI / / / , P- 0.001 TOHftINOEXUSS HEATING -~V / • I™ (AUCMENIEO HEATING ATAT OUTER SURFACE! J / / OUTER SURFACE!

0.001 IORRINOFXCESS HEATINGAT OUTER SURFACE!

Finally, a slip-flow calculation isshown in Fig. 12. Maximum surfacetemperatures at various housing/flywheelgaps are depicted for a Kevlar wheelspinning at 40, 000 rpm within an air en-vironment at p = 10~3 torr. Gap dis-tances have to be lower than the mean-free path to provide an effective con-duction path to the housing, assumed tobe maintained at 70°F. Also, as thedistance becomes smaller, the heatingcontribution due to viscous dissipationis reduced. These calculations serveas a lower bound to the temperaturefields since a slip flow is not absolutelydefined at this pressure.

Fig. 10. Transient temperatures atvarious pressures with or without excessheating at the rotor tip.

100

>qccmvl iqrad<qconv1

0 2 0.4 0.6 0.8 1.0

NORMALIZED RADIUS r/r

Fig. 11. The effect of a change ingeometry on the radial surfacedistributions cf temperature.

P- in " ' (orrwCOOO rvn

stipaow

AdWwttct Hftce

/

/ _ . . - -

HlllimMMv

Fig. 12. Maximum steady-state tem-peratures for slip/transitional aero-heating demonstrating the effect of thehousing flywheel gap spacing.

CONCLUSIONS

A computer program has been gen-erated for predicting flywheel tempera-ture subjected to free-molecular or slip-flow aeroheating. Allowances were madeto modify flow regime and flywheel geom-etry for either a steady-state or atransient-time calculation. Additionalallowances were incorporated for com-posite materials with temperature-varying spin rates and variable gasproperties.

The following conclusions havebeen demonstrated by the code:

105

1. For free-molecular aeroheat-ing, gas pressures from 10~3to 10~5 are necessary to avoidsteady-state temperatures>50°F above ambient.

2. A gas of lower density producesreduced heating effects.

3. Protruding edges can causetemperature-related problems.

4. Increasing the thermal conduc-tivity in the radial directionrelaxes the thermal field andlowers maximum temperatures.

5. Time to reach steady state istypical of ~1 hr.

6. For slip-ilow calculations, thetemperature field is decreasedby reducing the housing/flywheelgap distance.

7. Since thermal radiation plays amajor role in the heat transfer,a nigh housing emissivity isdesirable.

APPENDIX

FREE-MOLECULAR HEATINGRELATIONSHIPS

The heat-transfer parameters r'and St have been derived by A. K.Oppenheim7 for a flat plate inclined atan angle 8 to the incidence flow (negativeangle yield back-surface heating). Theseare given as:

2S2 t -

1 + 4n (S sin 9)( 1 + erf (S sin 9) exp (3 sin 9) )

s t = ijm j e x p ( - ( s sin e)2)

+ sfff (S sin 6) [l + erf (S sin 9 »

where

S = rwvfy/2a

THERMAL PROPERTIES OF KEVLAR49/E POXY COMPOSITE

The calculations of this study usethe following thermal properties:

Fiber direction

k = 1. 39 + 1. 11 x 10" 3T + 0. 442 x 10~6T2

(Btu/ft/hr/°R)

Perpendicular to fibers

k = 0.074 + 0.22 x 10"3T + 0.023 x 10"6T2

(Btu/ft/hr/°F) (T is in °R)

p = 86. 4 lb/ft3; C = 0. 3 Btu/lb/°R;P

e = 0. 8; a = 1.0

REFERENCES

1. J. A. Kirk, "Flywheel EnergyStorage - I and II. " Int. of J. Mech.Science, 19, pp. 223-245, 1977.

2. J. A. Rinde, Fiber-Composite Fly-wheel Program-Quarterly ProgressReport, January-March 1977,Lawrence Livermore Laboratory,UCRL-50033-77-1.

3. H. Schlichting, Boundary LayerTheory, 6th Ed., (McGraw-HillBook Co., New York, 1968), pp. 213-218.

106

REFERENCES (cont)

4. S. A. Schaaf and P. L. Chambre,Flow in Rarefied Gases, No. 8,Princeton University Press,Princeton, NJ, 1961. p. 38.

5. R. Stegel and J. R. Howell,Thermal Radiation in Heat Transfer,(McGraw-Hill Book Co., New York,1972), p. 282.

6. M. R. Baer, Aerodynamic Heatingof High-Speed Flywheels in Low-Density Environments, SAND78-0957,Sandia Laboratories, Albuquerque,NM., October 1978.

7. A. K. Oppenheim, "GeneralizedTheory of Convective Heat Transferin a Free-Molecular Flow, " J. ofAero. Sci., 49, 49-58, January 1953.

107

SESSION I I I : FLYWHEELS

109

PROJECT SUMMARY

Project Title: The Application of Flutd Film Bearings and a PassiveMagnetic Suspension to Energy Storage Flywheels

Prfncfpal Investigator: Martin W. Eusipi, Larry Martin, andDr. Amit Ray

Organization: Mechanical technology Incorporated968 Albany-Shaker RoadLatham, NY 12110(518) 785-2211

Project Goals: The goal of this program was to establish realistic fluid filmbearing performance parameters over a range of energy storagelevels from 10 KW-hrs. to 100 KW-hrs. A passive magnetic liftwas to be incorporated in the suspension to minimize the size ofthe fluid film thrust bearing.

Project Status: A final report including detailed parts drawings of a bearingtest model was included in the program goals.

Contract Number: 076997

Contract Period: Jan. 1978 - Sept. 1978

Funding Level: $55,000

Funding Source: Sandia Laboratories, Albuquerque

111

THE APPLICATION OF FLUID FILM BEARINGS AND A PASSIVE

MAGNETIC SUSPENSION TO ENERGY STORAGE FLYWHEELS

by

Martin W. Eusepi, Larry Martin, and Dr. Amit Ray

of

Mechanical Technology Incorporated

968 Albany-Shaker RoadLatham, New York 12110

518/785-2211

ABSTRACT

This paper presents the results of an investigation to establish realistic fluidfilm bearing parameters and a passive magnet suspension system for the support of energystorage flywheels in the range of 10 kW-hr to 100 kW-hr. A specific design for a threelobe preloaded journal bearing and a shrouded step thrust bearing was established for a10 kW-hr flywheel which weighs 1600 lb and has a maximum rotation speed of 12,500 rpm.Equations are presented to assist in establishing the bearing design requirements forother flywheel sizes. The magnetic suspension system designed to support 90 percent ofthe flywheel weight operates in an attractive mode and incorporates a novel configurationwhich provides a positive stiffness slope for the attractive magnet assembly.

INTRODUCTION

Point of consumption energy storagecan produce worthwhile economies. The useof flywheels as the storage means is at-tractive but is limited by the continuousenergy drain of frictional losses. Thisreport will present the results of a studythat established realistic fluid film andmagnetic bearing performance parameters,over a range of sizes of energy storageflywheels from 10 kW-hr to 100 kW-hr. Atthe 10 kW-hr size, a flywheel weight of1600 lb (30 in. dia and 6 in. thick) anda maximum speed of 12,500 rpm, in a verti-cal orientation, are representative condi-tions. A vacuum environment of 10"3 Torrand an external ambient temperature rangeof 60°F to 100"F are assumed for designpurposes.

Long life and reliability, coupledwith minimum power loss, are the designgoals. Power consumption is crucial,since it represents a continuous drain inthe stored energy. In order to minimizepower consumption, a passive magnetic lift,supporting up to 90 percent of the flywheelwe ight and utilizing no direct electrical

is incorporated in the design.

The principal performance factors governingthe flywheel suspension design were: powerloss; magnet support capability; magneticsupport geometry; fluid film bearing geo-metry; lubricant selection and distributionsystem.

SUSPENSION CONCEPT SELECTION

FLYWHEEL SIZING

Prior to establishing the final con-figurations for both the magnetic suspen-sion and fluid film bearing designs, it isnecessary to determine the range of supportneeded, as the storage capacity of a fly-wheel is varied "over the range 10 kW-hr <E < 100 kW-hr. For any flywheel, thestored energy level is governed by the pro-portionality relationship: E % R^u^t;where R = radius, u> = rotation frequency(sec"1), t = thickness. In addition, thedisk stress can be expressed as: a •*• R2&)2and the weight by: W i. R^t.

If the allowable stress level is main-tained constant for all flywheels, then theproduct of rotational speed and diameter(DN) is a constant. From this assumptionand keeping the R/t ratio constant, it can

112-

be shown that the following relationshipsexist:

• Flywheel weight; H - W [|_]

° o• Flywheel Diameter; D - Do [f-]

1 / 3

o

• Flywheel Thickness; t • to l|-) 1 / 3

• Maximum Flywheel Rotational Speed; N • t)Q l-jrl1'3

At C/R - 1.0 x 10~3 and W/DL - 500or 250 psi, the load equation reduces to

W x - 1.62 @ 250 psi W 2 - 3.24 @ 500 psi.

A total radial load of 60 lb, 14 lbfrom a 1° tilt and 46 lb from a 12 x 10"6

in. unbalance has been chosen as a reason-able design value. The calculated bear-ing data is presented in Table 1.

Calculations of flywheel size andweight are based on a 10 kW-hr flywheelwith the following characteristics:

Wo =

o

JOURNAL

1600 lb

6 in.

BEARING

Do

No

SIZING

= 30

- 12,

in.

500 rpm

As a means for generating bearingloss data prior to the selection of afinal flywheel bearing configuration, aclearance ratio (radial clearance of thejournal bearing (C) to the bearing radius(R)) of C/R = 1.0 x 10"3 and a room temp-erature lubricant viscosity of Z = 5 centi-poise was selected. This viscosity ischaracteristic of commercially availablegyro-lubricants having a vapor pressureof approximately 0.001 mm Hg at 181°F.The bearing calculations are based on aloading pressure of 250 psi and 500 psifor the 10 kW-hr flywheel size.

For an elliptical bearing , the di-mensionless load parameter is given by

W = wMNLD

(C/R)'

LD

Viscosity,lb-sec/in.2

Length, in.Diameter, in.

The power loss is determined fromthe equation

3 2HP = 0.0042 [M —

. 3 T, and its flowc

rate from Q = 7 NDLCQ where f = fric-tion coefficient.

The extension of the performance pre-dictions listed in Table 1 to other fly-wheel sizes is performed as follows.

From the bearing's dimensionless load,power and flow rate equations and the fly-wheel extension equations, the bearing per-formance parameters can be reduced (forW/DL = constant) to:

Load Parameters W = W o [^Ho

Power Loss H = H Q [-|-] [=-];o o

Flow Rate Q = Q [--] [-^].o E -

o Qo

The calculated performance of theelliptical bearing extended over the fly-wheel energy storage range 10 kW-hr < E <100 kW-hr is based on the above equations.The quantities W, T, and Q may be differentfor different bearing designs and cannotbe related directly to a change in flywheelsize; however, for detailed design, a de-signer must be aware that specific bearinggeometries will produce somewhat differentnormalized results requiring less generalcalculations.

THRUST BEARING SIZING

The frictional power loss generatedby a thrust bearing is expressed as:

2-2r

' w h e r e V - viscosity; A

Table 1. Journal Bearing Dimensional Performance Data; 10 kW-hr Flywheel.

BearingDesign Pressure

psi

Length andDiameter

in.

MaximumSafe Load

lbPower Loss

Watts

RequiredLubricant Flow

in.3/sec

OperatingFilm Thickness

10"3 in.

250

500

.49

.34

210

105

29

15

6.3 x 10

3.6 x 10'

r30.25

0.17

113

active area; h = nominal film thickness,

u = rotational frequency; r = mean radius.

Thrust bearing area is established bysetting the nominal design average pressureto: p a v e = 800 psi and by establishingthe criteria that the thrust bearing in-ner area diameter will be one-half itsouter diameter. In addition, since themagnetic suspension is to provide a sig-nificant portion of the gravity thrustload, the fluid film thrust bearing loadcan be expressed as W^ = aWf. Therefore,the bearing load is reduced by a and thebearing diameter by [a]-*-/2.

The thrust bearing power loss, as afunction of energy level to be stored inthe flywheels cannot be calculated untilthe operating film thickness is establish-ed. The calculation of film thickness isbased on Fuller2 where an approximate filmthickness for a hydrodynamic thrust bear-ing is given by the relationship:

1/2

, where:

aveI = mean bearing length;n = pad size factor; taken as 0.44;k =hydrodynamic factor; taken as 0.04;u = mean velocity.

The smallest thrust bearing (10 kW-hr flywheel @ a = 0.1) with a mean radius,r = 0.69 in. at one-third its maximumspeed of 12,500 rpm has a mean velocityof u = 301 in./sec, a mean bearingof I = 0.434 in., and a calculated minim;film thickness of h = 1.2 x 10"4 in.

The bearing load, film thickness andpower loss equations can be used to cal-culate bearing power loss as a function offlywheel size. To calculate the effect ofspeed variation on power loss, the effectof speed on film thickness must be takeninto account. It can be showi. that for afixed flywheel the power loss can be ex-pressed as:

3/2

The results of the parametric thrustbearing power loss study show that unlessfluid-film bearing loads are substantiallyreduced below the actual flywheel weight,excessive bearing losses are experienced.As an example, at full speed a 1600 lbthrust bearing load would produce a 3500watt loss while a 160 lb load producesonly a 6 watt loss.

MAGNETIC SUSPENSION SIZING

Magnetic suspension design became asignificantly important study when it be-came apparent that a magnetic lift wasessential to minimize the size of the fluidfilm bearings. Both passive or active mag-netic suspension can be used for flywheelload support.

The electrical techniques for servoingmagnetic bearings have reached the levelof sophistication that it is possible toobtain magnetic suspension actively. Thegreater the ability to maintain passiveforces, the lesser the dependence on servo-ing and the associated electrical equipmentpower requirements and propensity for in-stability or non-reliability. Therefore,the present study considers only the pas-sive magnetic lift where no servo techniquewill be utilized.

As a part of this study, several typesof magnetic support bearings for the groupof flywheels in the energy storage rangeof 10 kW-hr < E < 100 kW-hr were consider-ed; these were: conventional attractivedesign; repulsion design; new attractiveaxially stable design.

Conventional attractive magnetic bear-ing designs all show axial instability ora negative spring rate. This bearing type,therefore, was considered and subsequentlydiscarded for this application. However,the repulsion type magnetic suspensionich results from the installation of like

magnetic poles across a bearing gap wasanalyzed and solved for high coercivityrare earth permanent magnets.

The following design equation, result-ing from this analysis is (in the MRS sys-tem) used to calculate <he quasistaticbearing capability: F = 3.375 x 104 w2l,w = width of magnet, I = dimensionlessforce. For calculation of nonr;aal or tan-gential forces, the appropriate value ofI is obtained from Figure 1.

The 10 kW-hr size flywheel with aweight of 1600 lb requires a magnet bear-ing support of 1440 lb. If a magnet widthof 1 in. is selected, then the total forceequation reduces to: EFZ = 70 n and thetotal number of magnets required is n = 92.The total magnet weight required to sup-port 1400 lb is 54.28 lb.

Since the number of magnets is direct-ly proportional to the required support

114

CURVE

1254Se7

ES

0 0aos0.100.20aso1.002flO

N

«;

M

1 1

S

(a)

L F>I 2

S

1.00(b)

Fig. 1. Axial separation force,repulsion magnet.

Fig. 2. A simple axially stableattraction magnetic bearing.

force, which in turn is directly propor-tional to the energy storage level, thenthe relationship , E . can be used

nl = no <E-}

in determining preliminary magnet require-ments for other flywheels.

In a detailed design of the repulsionmagnet type suspension, the inclusion oftangential eccentricities arid the inherentnegative radial stiffness must also be con-sidered. This preliminary study, however,shows the repulsion type magnet suspensionto be feasible and should be practical tobuild from a manufacturing viewpoint.

NEW ATTRACTIVE AXIALLY STABLE DESIGN

The main objection to the use of theusual attractive magnetic lift is that itis unstable in the direction of suspension.A new concept has been developed to over-come this deficiency.

the familiar attraction system"bearing design utilizing the new con-

cept has a stationary member composed ofpermanent magnets and soft magnetic polepieces. The design also included a moving

member which is fabricated from a soft mag-netic material. When assembled, they forma bearing configuration with each polepiece of the stationary member facing aslot in the moving member, shown in Figure 2.

A stable axial attractive system isachieved by saturating the iron pole pieces.The saturated poles keep the total fluxconstant as the air gap distance "g" de-creases when the two members approach eachother. Saturation of the magnetic mate-rials in the moving member will furtherensure this. The geometric configurationallows two flux paths which can be dividedinto two components, one in the radialdirection and the other in the axial direc-tion. By choice, the reluctance of theradial path decreases more rapidly thanthe reduction of the axial path for a fixeddecrease in the axial distance "g". Thiscan be seen by considering positions (a)and (b) of Figure 2. More flux will divertin the radial direction, since the totalflux remains constant, this will result ina decrease of the axial flux. Therefore,the radial force will increase, and theaxial force will decrease. This is ex-actly the opposite of a normal attraction

115

system. The decrease in axial force, withthe decrease in axial distance, can beaccentuated furthermore by saturating themoving members, thereby increasing theaxial stiffness. This concept is selectedfor optimization since it presents a posi-tive stiffness in the attractive mode.

Because of the circular symmetry, thebearings will be neutral in the circumfer-ential direction; in the radial direction,it will also be neutral because the totalflux remains unchanged for any radial per-turbation, due to saturation in the softmagnetic pieces.

FINAL MAGNETIC SUSPENSION DESIGN

The permanent magnet assign problemis, in general, to specify the permanentmagnet materials in terms of defined unitproperties and to arrive at a volume for agiven configuration which most efficientlyestablishes the total field energy requiredin the given space. The bearing calcula-tions were performed by assuming a fluxpath and then calculating the permeance inorder to determine the air-gap flux. Thetotal permeance establishes the point ofoperation on the demagnetization curve.It is essential that the permeance be eval-uated very accurately in order to ensurethat the magnet operates at the optimumpoint in the demagnetization curve forminiaium weight. The optimum point of op-eration is defined as the point in the de-magnetization curve where the product BmHm,or available magnetic energy, is maximum.For the ferrite magnet, B m = 0.2 Tesla andH,,, = 150794 AT/m at the optimum point ofoperation.

CALCULATION RESULTS

The magnetic bearing dimensions arecomputed for a 10 kW-hr energy storage fly-wheel. The total weight of the flywheelis about 725 kg (1600 lb) and the maximumspeed is 12,500 rpm. A large percentage,but Jess than the total weight of the fly-wheel, has to be supported by the magneticbearing. Computations were performed forboth ferrite and Samarium-Cobalt magnets,and are presented in Table 2. Materialused for soft magnetic pieces is 65 per-molloy in the stator and Supermalloy, inthe rotating member.

The variation of axial force with airgap is shown in Figure 3. If the magneticlift force is set at any value between theclosed and open positions, the magnetic bear-ing will provide a stable system in the

Table 2. Magnetic Bearing Dimensions fora 10 kW-hr Flywheel.

Parameter

3 • Flux Densityo

H " Magnetic Field Strength

Mean aadius r

Lengch lg

Gap i;g" for open position

Gap "g" for closed position

:

Lra

C

Vertical Force for Open Position

Vertical Force for Closed Position

Total Weight of Magnet

Ferrite

0.2000

150794

30.48

0.508

1.524

0.508

1.504

0.851

4.887

725.0

532.0

3.83

Samarium-Cobalt

0.4250

337302

30.48

0.508

1.524

0.508

1.366

0.389

2.093

725.0

516.,;

1.27

Units

Telst

At/n

cm

fflffl

am

tan

cm

cm

cm

kg

kg

ke

axial direction. Stationary magneticpieces are made of 65 permalloy radialstampings separated by insulating materialof equal thickness. The radial laminationswith alternating nonconductive and non-ferrous shims, serve the following func-tions: Circumferential flux flow is pre-vented if the shaft is off vertical andthe uniform gap is not maintained, therebyreducing the non-uniform air gap fluxdensities and, hence, the unbalanced mag-netic pull; eddy currents, which might begenerated due to the non-homogeneity inthe flywheel and the insert materials, areprevented. The soft magnetic rotor fabri-cated front a solid piece of supermalloy.

Z « .6 .0 1.0 *.2 1.4 1.6 l.«

Fig. 3 Axial force versusairgap "g"

116

FLUID FILM BEARING DESIGN

The design for the journal and thrustbearings of a 10 kW-hr magnetically sup-ported flywheel includes the requirementfor low power loss and excellent rotor-dynamic performance.

Since the flywheel studied operatesvertically, special considerations are im-posed on journal bearing selection, partic-ularly with respect to subsynchronous whirl.Figure 4 schematically illustrates thevertical flywheel-bearing system.

The journal bearings are representedby short cylinders, or "pintles," whichare designed with sufficient flexibilityto provide necessary damping via a radialdamping device. A fixed-pad, tapered landthrust bearing is provided at the lower pin-tle for support of axial load.

Jt SCMWI

«' UMTHK)

30'0U.

«"TWCX

NSSWASCNMHOCT U S E M U[FCH03CNI

THUSTBUKINS «N0 OUKR

Fig. 4. Schematic of flywheel-bearing system

The entire flywheel assembly will op-erate in a substantial vacuum, so thatwindage losses will be negligible. Therotor is modeled, as shown in Table 3 byeleven stations. This model is succes-sively used for all critical speed, syn-chronous response and stability (subsyn-chronous whirl) studies.

The mathematical model given in Table3 may be used directly in the criticalspeed computer program. It is necessary,however, to compute an estimated stiffnessof the pintle supports so that theireffects may be included on the criticalspeed map.

PINTLE FLEXIBILITY CALCULATION

The pintles are treated as cantileverbeams; the stiffness, due to bending andshear deflection, can readily be estimatedand combined to yield an "effective" stiff-ness. Assuming a simple cantilever beam,the bending and shear deflections for steelare calculated to be K^ d _ 2.76 x 10

5lb/in.

and K s h e a r = 2.084 x 106 lb/j.n., resulting

in an effective stiffness: K e f f = 2.44 x105 lb/in.

Although the stiffness of the pintleshave been used consistently in the rotor-dynamics calculations, the geometry assumedmay change: for example, the length of thepintles may have to be increased due tospace requirements for the thrust bearingand damper. The stiffness, however, canbe fixed by providing the same effectivelateral stiffness. This would entail,perhaps, boring out the pintles to a speci-fied diameter or adjusting its activelength.

Table 3. Mathematical Model of Flywheel-Bearing System.

ROTOR DATASTATNO.1 0.2 0.3 0,4 0.5 .6 0.7 .8 0.9 0.10 0,11 0,

MASS(L8S>

I

!l36£«031

,136E«031

1

1

IP

0.0.C.0.

o!.39S0E»04

0.

0.0.

BEARING STATIONS2 10

(L0000

0

0000

IT LENGTHB-!N«2» (INI

.2503.0833.1333.334

.1975E«O4 3.0003.000

.1975E<04 2.0002.0001.750.250

0.000

STIFF.OIA.6.0006.0006.0006.000IS.OOQ11.0006.0006.0006.O006.O006.000

MASSDIA.6.0006.0006.A006.00030,00030.0006.0006.0006.0006.0000.000

INNER YOUNGS HOD. OENSITV SHEAR MOO.OIA. as/IN<"»2> <LB/IN«3>0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.000

•3000E+08.3000E*OB.3000E»08.3000£*08.3000E*OB.3OO0E«0B,3000E«Oa,3000E«08.3000E«08,3000E*08,3000E*OB

.283E»00

.283E*00

.283E*00

.283E*00•283E.00.283E*00,283E*00•283E*00.283E«00.283E«00.283£*00

(LB/IN»»2).110E*08.U0E •OB•110E»08.}10E»0e.uoE»oa•110E*0B.!10E*0a.110E*08•110E»08.110E*08,110F*08

117

A critical speed map was prepared forthe following cases where the attractivemagnetic bearing coefficients included andestimated at Kradlai = 0; K a n g u l a r =-6.74x10J in. lb/radian. The critical speed mapis shown in Figure 5.

With reference to the critical speedmap the observations can be made that:the flywheel rotor has to traverse throughone, and possibly two, critical speeds toreach the maximum operating speed, depend-ing on the actual bearing stiffness. Itsnould be noted that the first criticalspeed is characterized by a cylindricalwhirl of the rotor in its bearings and thesecond critical speed by a conical whirlof the rotor in its bearings.

As will be shown later, the highstiffness pedestals yielded unsatisfactoryperformance, both from a stability and asynchronous response viewpoint.

f*mm * I2S00 WW

K_ • *

)

G.T4E3-T5J-

. 5

2 9 Iff 2 8 10° Z 9BEMtO STIFFNESS, ft/In

FUIT HOMES CONSTANT FED STIFFNESS OF 2 4 E S I b / h '

Fig. 5. Critical speed map; 10 kW-hrflywheel; attractive magneticbearing; flexible pedestals.

DYNAMIC BEARING LOADS

A vertical rotor can have bearingloads resulting from misaligning the planeof the disk flywheel with respect to thespin axis. It can be shown that the dy-namic load on each bearing (for a symmetri-cal rotor) is given by:

sinif>R

and for a th in disk and small angle ij>( r ad ians ) , t h i s equation reduces to

I $R *

It should be noted that this dynamicload varies in direction as a synchronousvector with constant magnitude.

The calculations also indicate thatextreme care must be taken in the manu-facture of the flywheel, with particularattention paid to the misalignment angle<)>, between the shaft centerline and theprincipal axis of the flywheel disc.Large dynamic loads could result in bear-ing overloads.

Although the angle between the spinaxis of the flywheel and its principalinertia axis must be held to a value ofless than <|> = 0.001°, manufacturing pro-cedures and present machine tool capabil-ity are available to assure that this con-dition can be met.

PEDESTAL DAMPING FORMULATION

Pedestal damping of a translational(or radial) nature will be provided in theform of a thin film of lubricant, similarto a journal bearing except that rotationdoes not occur. The damping forces in thepedestal will limit synchronous whirl ofthe flywheel-bearing system.

From short bearing theory, and foreccentricity e =. 0, and L/D ~ 0, the damp-ing is calculated from: IT ,L. 3

"radial = 2 (C}

(Full Film). The damping can be supplied,as an example, by a single ring.

A damping value of 200 lb sec/in,obtained from a 2 in. dia x 2 in. wide x.001 thick oil film was used in the sta-bility and synchronous response calcula-tions as a first approximation for therequired level of damping.

JOURNAL BEARING SELECTION

Because of its excellent stabilitycharacteristics, the preloaded three lobebearing, as shown in Figure 6 was con-sidered in place of the elliptical bearingas the design for use on the flywheelrotor. Reference [3] provides most of thenecessary information from which thesteady-state and dynamic characteristicsmay be determined.

This design is appealing since no"tilting" is required to load the bearing;the geometry provides the load. Thus, theflywheel-bearing system may be operatedsuch that the static rotor centerline istruly vertical.

118

10* 10* There is some uncertainty, however, thatJOURNAL the geometry of the lowest loss case willRADIAL CLEARANCE yield a stable system and/or show a satis-

C~ factory synchronous response. Thus, addi-tional parametric studies are required.

DYNAMIC BEARING COEFFICIENTS

Reference [3] provides tables ofdimensionless stiffness and dampness coef-ficients for various preloaded three lobebearings operating in vertical rotors.Also, we have K = K._. = Radial Stiffness;

Fig. 6.

GROOVE

Schematic of preloaded threelobe bearing

fi«ference [3] a]30 provides dimen-sionless friction force coefficients as afunction of bearing geometry and preload.The dimensional friction force can be com-

puted from (|) uNLD (£)"{} = pNLD(•jb W> where {} is the dimensionlessquantity obtained from the appropriatetable in Reference [3].

For u = 0.75 x 10~6 reyns; N =12500/60 = 208.3 rev/sec; L = D = 0.50 in.The horsepower loss per bearing may now becomputed from

Bearing = 63025 '

where R = journal radius in inches; N =rotational speed in rpm. For N = 12500rpm and R = 0.25 in., the hp/bearing =0.0496 Ff.

Since the horsepower loss is of ut-most importance, we will naturally examinethe configuration which yields the lowestloss.

This design study indicates that thelowest power losses occur when the set-upclearances are the largest. This is to beexpected for the vertical rotor.

Kyx

- Kxy

xx yy= Cross-Coupling Stiffness;

Radial Damping; B,xx yy= Cross-Coupling Damping.

yx- B

xy

Actual (dimensional) values ofstiffness and damping are computed fromthe following equations

Ktypical

typical

(|)2 uNLD

(|)2 pNLD

when the quantities in {} are obtained fromReference [3].

STABILITY ANALYSIS

Damped natural frequencies and associ-ated log decrements were determined for anumber of cases of preloaded, three-lobebearings. Figure 7 indicates the sharp de-crease in log decrement when the radial

a«4!0

1.

N1

1 \ 1

1 V.

1 VN J 1MS1MLE : i

N ^ j

3 LCet SS. m. 1/2

vonicju. noroRPED. K-«Z5OIB/«l

inM> 0.31b

0005 0007 .001Cm,aEWNG UDULCUAIUNCE (IN)

Fig. 7. Log decrement S versusclearance Cm, three lobejournal bearing.

119

clearance became larger than 0.0005 In. Itshould also be noted that flexible pedestalswere included for all runs, with K

81250 lb/in. and Bped

ped200 lb sec/in.

Assuming that the synchronous responseis satisfactory, the final bearing designfor the 10 kW-hr flywheel is a 1/2" x 1/2"preloaded (m = 1/2) three-lobe bearing hav-ing a machined radial clearance of 0.000500in., a set-up radial clearance of 0.000250in., and a power loss of approximately 21.6watts.

SYNCHRONOUS RESPONSE

Synchronous response calculations cor-responding to both static and dynamic un-balance were made using the dynamic bearingcoefficients determined for the three-lobedbearing.

Response calculations for these bear-ings assumed unbalancss of 1 oz-in. atall planes; results of this analysis areshown in Figure 8 for both in-phase andout-of-phase unbalance. The major semi-amplitudes are plotted as a function ofrotor speed.

For 1 oz-in of dynamic unbalance,the maximum semi-amplitude of 0.252

mils occurred at bearing 1. This levelof unbalance is clearly unacceptable,as it exceeds the steady-state clearanceof 0.250 mils. However, for an unbalanceof 0.1 oz-in., the semi-amplitude atbearing 1 can be reduced to 0.0252 milswhich represents a very acceptable level(10 percent) of dynamic whirl.

THRPST BEARING DESIGN AND PERFORMANCE

Hydrodynamic fixed-pad thrust bearingcalculations were made for the flywheelrotor-bearing system with reference to theprocedure given in Reference [4].

The steady-state load was assumed tobe 10 percent of the total flywheel weight,160 lb. Calculations were based upon amaximum shaft speed of 12500 rpm, a coef-ficient of viscosity of 0.75 x 10~ reyns,and a six-pad bearing.

Results of three cases are summarized

below

1

2

3

O.D.

1.

1.1.

h. —

(in.)

0

25

50

I.D.

0.

0.

P.

• flla

(In.) h.

50

.5050

thlckaua

,C

0

0

0

•Hi)

.203

.295

.416

Alt

21

13

10

•n.6

.7

.2

tomr LOM tacta)

36

60

87

.6

.9

.8

H z

K)

J

/

1

1

^ -

^ —

^ — —

t

1

- • —

(

-

1

•

PEDESTALm • 0 3 Ibk - 812501b /in.b • 2001b «c/ in.

ItfMOOE BRG2

2nd MOD!•

• BRGI• —

CM

0 1,000 3,000 5,000 7,000 9,000 11,000 13,000 15,000ROTOR SPEED,RPM

Fig, 8. Synchronous regppns.e three lobe bearings; (m = 1/2),

120

In order to minimize power loss, itappears that as small an OD as possibleshould be selected. Manufacturingconsiderations may, however, dictate a lar-ger than optimum outer diameter. In anyevent, the thrust bearing power loss rep-resents a substantial and probably a majorportion of the total fluid-film fricticnalloss in the flywheel bearing system,

SUMMARY

A viable suspension has been estab-lished to support a 1600-lb, 10 kW-hr en-ergy storage flywheel. The flywheel designconsists of a flat disk, vertically ori-ented, with fluid film journal bearings atopposite ends of its axle. This conceptpermitted the positioning of a small thrustbearing at the lower axle and a partial mag-netic suspension system to act directly onthe wheel's lateral surface.

Based on analysis, fluid film journalbearings incorporating a three^lobe designwill provide adequate load capacity. Whencombined with the flexible pedestal con-cept employing an extended pintle and dash-pot, the bearings maintain sufficient rotorresponse control. An evaluation of rotorresponse has established that requiredlevels of rotor balance and orthogonalityDf the rotating inertia mass to its spinaxis are reasonable and achievable bypresent manufacturing techniques.

In conclusion, the successful comple-tion of the flywheel's suspension system de-sign would indicate that extension of thatdesign to a finished test vehicle would pro-vide the necessary tools to evaluate boththe basic designs and the scaling factors.It is recommended, therefore, that a rig de-sign be implemented and that constructionand testing of the suspension design, bothfor the magnetic and fluid film systems beundertaken.

REFERENCES

1. Badgley, R.H., "Standardization Manual,Sliding Surface Bearings," MTI 69TR61,December 1969 (Revised June 1970).

2. Fuller, D.D., "Theory and Practice ofLubrication for Engineers," New York:John Wiley & Sons, Inc., 1956.

3. Lund, J., Rotor-Bearing Dynamics DesignTechnology, Part VII: The Three LobeBearing and Floating Ring Bearing; Me-chanical Technology Incorporated,February, 1968.

4. Raimondi, A.A., and Boyd, J., "ApplyingBearing Theory to the Analysis and Designof Pad-Type Bearings"; Part I; ASME paperNo. 53-A-8A.

A new method of designing magneticbearings has been presented. The methoduses a concept which makes an attractionsystem with soft magnetic pieces stable inthe axial direction. The conventional at-traction system is only stable in the trans-verse directions and requires servo tech-niques for stability in the axial direction.The new design develops a passive attractionsystem which is stable in the axial direc-tion and neutral in the transverse direction.

The means for extending the 10 kw fluidfilm bearing design to flywheels of otherenergy storage levels is also presented.The extension of magnetic suspension designto other wheel sizes is also explained. Be-cause of its complexity, the totordynamicanalysis does not lend itself to parametriccurves for extending performance to otherinertia systems. The procedure for the dy-namic analysis must be employed after pre-liminary sizing is accomplished.

121

PROJECT SUMMARY

Project Title: Development of an Advanced Flywheel Bearing Performance Model

Principal Investigator: David B. Eisenhaure

Organization: The Charles Stark Draper Laboratory, Inc.555 Technology SquareCambridge, MA 02139(617) 258-1421

Project Goals: Develop a performance prediction model for ball bearings based onexperimental data combined with a theoretical model. The objectiveis to have available a design tool to assist in the optimization ofthe flywheel bearing set when other system parameters become avail-able.

Project Status: Completed

Contract Number: 07-6995

Contract Period: Oct. 6, 1977 - Sept. 30, 1978

Funding Level: $99,000

Funding Source: Sandia Laboratories, Albuquerque

123

LOW-LOSS BALL BEARINGS FOR FLYWHEEL APPLICATIONS

David B. Eisenhaure and Edward P. KingsburyThe Charles Stark Draper Laboratory, Inc.

555 Technology SquareCambridge, Massachusetts 02139

ABSTRACT

This paper summarizes work performed by The Charles Stark Draper Laboratory,Inc. during a Sandia Laboratories sponsored contract (No. 07-6996) for the analysisand test of ball bearings suitable for vehicular flywheel applications. Particularemphasis was placed on the CSDL full complement retainerless concept which priorresearch indicated had extremely low losses and long life at high speeds in a vacuumenvironment. The primary accomplishments during the first six months of this 12-month program were the procurement and modification of test bearings, the design andconstruction of a test fixture, and the development of a theoretical representation ofthe bearing losses. The primary accomplishments during the second half of the pro-gram were the development of a good data base for the test bearings and a ball bearingperformance prediction model. This performance prediction model was formulated bycombining the experimental data base with the theoretically derived model and allowsthe optimization of future ball bearing configurations for vehicular flywheel systems.Three sizes of standard bearings were modified to the full complement retainerlessconfiguration. These modified bearings and one set of standard retainer bearingswere tested during the contract. Friction losses were obtained for a variety of radialand axial loads and the results integrated into the performance prediction model.

INTRODUCTION

The Charles Stark Draper LaboratoryInc. (CSDL) is currently under contractto Sandia Laboratories for the develop-ment of an "advanced flywheel bearingperformance model" (Contract No. 07-6996). This development deals withbearings for vehicular applications andplaces particular emphasis on the CSDLretainerless bearing concept. Thispaper summarizes work performed duringthis contract and outlines possible futureresearch. The following paragraphs sum-marize the background of this problem,as well as some of the problems asso-ciated with flywheel bearings and thesolution to these problems.

It has been generally recognizedthat a high-efficiency energy-storageelement will be necessary for the con-struction of effective hybrid and electric-al vehicles. Flywheel energy storagehas many characteristics which make itparticularly well suited for this applica-tion. This flywheel storage elementwould be used as a storage node for re-generative braking energy and might beutilized for other power averaging func-tions such as hill climbing, acceleration,or other high-powered modes of operation.

The three primary characteristics offlywheels which make them particularlywell suited for this application are high-power density, reliability, and long life.The last two lead potentially to low lifecycle cost. Despite these advantages,flywheels have several unique technicalproblems associated with them whichmust be addressed. These problemsrevolve primarily around increasing theoverall efficiency of the flywheel to anacceptable level by reducing the aerody-namic, bearing, and possible seal losses.The aerodynamic losses at high flywheelspeeds require a relatively high vacuumwhich in turn demands either a low-lossbearing that will operate in a vacuumenvironment or rotating seals whichcharacteristically have substantial lossesassociated with them, especially at highspeeds. The bearing design problem iscomplicated by the reaction torques ofthe rotating wheel during vehicle dynam-ics. These torques, without carefulsystem design, can create a hostilebearing environment.

The bearing requirements for avehicular energy-storage flywheel mustinclude: stability, reliability, fail-safe

124 - I

operation, low loss at high speed, temp-erature insensitivity, and substantialcapacity for combined and shock loads atzero speed, at high speed, and in vacuum.Advanced retainerless ball bearings rep-resent an ideal candidate for this appli-cation. These advanced bearings arecharacterized by long life and low powerdissipation, and are well suited for oper-ation in a vacuum environment. Thesebearings consist of standard commerciallyavailable ball bearings, suitably modifiedfor operation without the conventional ballretainer (also called the cage or ballseparator). In conventional ball bearings,the retainer is subject to violent instabil-ities of various types which are partic-ularly troublesome in vacuum operationand typically treble the driving torquedemand of the bearing. This problem iseliminated with retainerless operation as:

a) There is no possibility ofretainer instability.

b) An optimum oil supply and scav-enging system can be provided toto ensure adequate EHD lubricantfilm thickness in all the ball racecontracts, without introducingexcess oil and unnecessary vis-cous drag.

The results are long reliable life andminimum friction loss. There are noknown disadvantages resulting from r e -tainerless operations. A prior CSDLprogram* has produced prototype retain-erless bearings designed to replace thespin axis bearings in the Skylab ControlMoment Gyroscopes which had sufferedfailures in space triggered by retainerinstability. These retainerless bearingshave successfully passed every perfor-mance test during their development, anda pair is currently on life test at morethan 8000 hours in vacuum, without diffi-culty.

DEVELOPMENT EFFORT

Summary of Work to Date

The goal of this research program isto develop a performance predictionmodel for ball bearings based on experi-mental data combined with a theoreticalmodel. The object is to have a designtool available to assist in the optimizationof the flywheel bearing set when othersystem parameters become available.The following summarize efforts duringthe first half of the program:

a) Four sets of angular contact ballbearings of various sizes wereacquired. Three sets were modifiedto the full complement retainerlessconfiguration while one mid-size setwas not modified.

b) An appropriate test fixture was des-igned and constructed.

c) An analytical model for the bearingswas formulated and a correspondingcomputer code generated.

The following work was accomplished dur-ing the second half of the program:

a) A series of tests was conductedon each of the four bearing sets toobtain the bearing losses as a func-tion of speed under various axial andradial loads.

b) A detailed performance predictionmodel for retainerless ball bearingswas developed by combining theexperimental results with the theo-retical model. A more limited per-formance prediction model for stand-ard ball bearings a lso was developed.

c) A final report is currently being pre-pared presenting the results of theresearch and recommendations forfurther research.

Test Bearing Construction

Angular contact bearings of standardinertial geometry in three sizes, spanningthe range of interest, were procured forthis program, as shown in Table 1.

The outer race dams on these bearingswere ground away to eliminate ball damagein the assembly process. They are allthus "fall aparts" in the full complementconfiguration.

Since no long-term running is contem-plated in these tests (the driving torques tobe measured in the near-term steady stateafter run up), no provision for an oil sup-ply was required.

Oil in controlled amounts was applied tothe balls by evaporation from a dilute solu-tion in Freon before assembly. Thismethod has been perfected in other pro-grams and allows sufficient running timeto make the torque measurements withoutany damage to the bearings.

125

Table 1. Angular Contact Bearings

1Size

100H

105H

108H

Geometry (inches)

Bore

0.3937

1.2910

1.9385

Width

0.3166

0.4740

0.5922

OD

1.0236

1.8504

2.5772

BallDiameter

0.1875

O.2bOO

0.3125

FillComplement

11

17

21

Bearing Loss Analytical Model

Absence of the ball retainer and ofuncontrolled bulk lubricant in the full com-plement configuration eliminates four ofthe energy sinks discussed for convention-al ball bearings. There remain threesinks to be considered in a theoreticalformulation of a retainerless energy de -mand model:

a) Shear losses in the EHD films dueto linear velocity differences (slips)between the load-carrying surfaces.

b) Shear losses in the EHD films dueto angular velocity differences(pivoting) at these contacts.

c) Hysteresis losses in the elasticdeformations of the metal parts asthey pass through the loaded con-tacts.

One additional sink, which is peculiar toretainerless operation, involves losses atthe ball-ball contacts.

The initial version of the loss modelinvolves only i tems b) and c). The ball-ball losses are expected to be smallbecause the ball-ball forces and contactareas are known to be small compared tothose at the ball-race contacts. The sliplosses are not accounted for initially be-cause slips are known to be small in thepure axial load cases. It is hoped to usea comparison of the axial and combinedload results (both test and analytical) toestimate the overall importance of slipsfor later iterations of the model.

The hysteresis and pivoting lossesdepend on specific conditions at each con-tact, which are , in general, different ateach ball and also different at the innerand outer race contacts on the same ball.An existing computer program due toJones^ is used to calculate these individualcontact conditions for the bearing sizes andoperating conditions of these tests.

The quantities required at each contacton each ball are:

Normal load

Maximum hertzpressure

Hertz major semi-axis

Hertz minor semi-axis

Contact angle

Ball orbit rate

Ball spin rate

Ball spin orientationangle

N

m a x

a

b

B

S

s

lbf

lbf/in

i n .

in .

deg

rad/s

rad/s

deg

Individual contact losses are then calcula-ted according to the equations developed inthe following sections and summed for thewhole bearing to get the theoretical energydemand.

Hysteresis

The hysteresis losses are estimatedbased on the work of Drutowski^, whofound, for a hard steel ball rollingon a similar flat, that the rollingresistance force could be representedby a power law as a function of normalload. The loss equation is

for the outer (o) and inner (i) contacts onthe jth ball.

Pivoting

The pivoting losses are estimated basedon the work of Johnson5, who found that his(and others) experimental traction resultsat high pressure could be explained as fol-lows: "at high pressure the (EDH) filmexhibits a critical shear stress which isapproximately proportional to the pressure

126

and decreases slightly with temperature,somewhat similar to a granular solid.This hypothesis leads to an approximatelyconstant 'coefficient of friction1 independ-ent of film thickness, as would be observedin dry sliding or boundary friction. " Thisis a great simplification for the purposesof the present work, as it eliminates cal-culation of the EHD film thicknesses. Onthis basis, the pivoting torque at any con-tact is

(2)

where ri is the Johnson coefficient and theintegral is taken over the contact area.The loss equation for the jth outer (inner)contact is

jLp(o,i) = (3.17xlO~2) jPmax(o,i) j f i

(o,i) (watts)where fi is the pivoting at that contact.

(3)

For the first iteration of the lossmodel, it is assumed that the ball spinvector is always normal to the line of con-tacts (<j> = o). In this case the pivotingsare

and

(4)This is equivalent to assuming a fixeddivision of the pivoting between the innerand outer races (it is kineruatically impos-sible to have zero pivoting at both racessimultaneously). In later iterations, val-ues of 4> for minimum total pivoting dissi-pation will be sought numerically.

Some hand estimates based on theseconsiderations resulted in pivoting lossesconsistently higher than the hysteresislosses. Since the pivoting losses do notdepend directly on the size of the bearing,this suggests that for a given load theremay not be a large loss reduction in goingto a smaller bearing.

CONCLUSIONS AND FUTURE RESEARCH

It is anticipated that the bearing per-formance model would be used in a futuredevelopment program, along with thecharacteristics of other flywheel compo-nents, in order to specify an optimumvehicle flywheel module. After the fly-wheel module is specified, the bearingperformance model would be utilized inthe design of an optimized bearing set.The optimized bearing design would includedetermination of bearing size, rotationalmode, internal geometry, oil supply rate.

and determination of mounting details. Thedevelopment of a specially tailored lubri-cation supply and scavenging system wouldalso be required. Although the lubricationproblem has been successfully handled onprior CSDL programs, it was not addressedduring the current program in order todirect the resources toward more extensivetesting. The technology generated duringthis contract will allow an optimized des-ign to be obtained in a very cost-effectivemanner. After the bearing is developed,it should be tested extensively under antici-pated conditions of vehicle static and dy-namic load.

To determine the specifications of theflywheel module bearings, one must startwith the vehicle and its operating environ-ment. The flywheel module would bedesigned for use in some class of automo-bile (forinstance,a 3000-pound familycar) in which the flywheel would be able tostore the energy of a couple of stop/startcycles. Past government studies on asimilarly sized vehicle have used fly-wheels with about 2/3 horsepower-hourcapacity (1/2 kWh). Parameters such assteady load and speed are determined to alarge degree by the car size and flywheelmaterial. To specify values for transientload capacity and bearing stiffness, it isnecessary first to model the vehicle dy-namics. This would involve a model ofthe vehicle suspension and allowable roadirregularities. A characterization of theamplitude and frequency of vehicle bodymotions, coupled with a knowledge ofgyroscopic forces inherent in the flywheel,leads to a calculation of loadings on thebearings. There is a trade-off involvingspeed, bearing strength, torque, etc., sothe above analysis must be done paramet-rically.

An analysis of other factors is alsonecessary before attempting to define thebearing. Losses such as windage andvacuum pumps must be modeled in orderto optimize flywheel shape and speed.Once equation^ are defined relating bearingloss, windage, flywheel shape and material,spee'', and loads, a "strawman" flywheelmodule can be designed, taking into accountgeometrical constraints and other factors.At this point, one has a set of nonlinearequations describing the vehicle and fly-wheel module systems. A computer pro-gram could then be written to optimizesome figure of merit for the flywheel mod-ule (for example, the energy storage effic-iency for a driving cycle). The outputs of

127

the computer program would be the fly-wheel operating parameters and the bear-ing design specifications.

The retainerless configuration showedconsiderably higher reliability during thetest period together with somewhat lowerlosses when compared to retainer typebearings. The original premise of the pro-gram that this type bearing was idea l lysuited for high speed flywheel applicationswas confirmed.

REFERENCES

1. Kingsbury, E. , Evaluation of Alternate Bearing Designs for the Skylab GMG FinalReport, Charles Stark Draper Laboratory Report R-1026, December 1976.

2. Townsend, D. P . , C. W. Allen, and E. F. Zaretsky, "Study of Ball BearingTorque Under Elastohydrodynamic Lubrication, " Trans ASME, JOLT, 1974.

3. Jones, A. B. , "General Theory for Elastically Constrained Ball and Radial RollerBearings Under Arbitrary Load and Speed Conditions, " Trans ASME, J . BasicEngineering, 1960.

4. Drutowski, R. C , "Energy Losses of Balls Rolling on Plates, " Trans ASMS, J. ofBasic Engineering, 1959.

5. Townsend, A. and E. F. Zaretsky, "Elastohydrodynamic Lubrication of a SpinningBall in a Non-Conforming Groove, " (Comments by R. L. Johnson), Trans ASME,JOLT, 1970.

128

Project Title:

PROJECT SUMMARY

Seals Evaluation for Advanced Flywheel Energy StorageSystems

Principal Investigator: I. Anwar

Organization: Franklin Research CenterDivision of the Franklin Institute20th and Race StreetsPhiladelphia, PA 19103(215) 448-1000

Project Goals: An evaluation of the problems associated with the use ofrotating shaft seals in this application, along with acompilation of performance data for present state-of-the-art seals and suggestions for future developmental work.

Project Status: The draft report has been approved by Sandia Laboratories andthe final version of the publication is being produced.

The report documents the performance of those seal conceptsthat are most promising for flywheel application. Particularattention is given to the so-called "dynamic" seals as beingintrinsically well suited to the problem. It is concludedthat a hybrid seal combining the screw-type and turbo-pumpgeometries offers the greatest prospect for future development.

Contract Number: 07-7141

Contract Period: Jan. 1979 - Nov. 1978

Funding Level: $22,376

Funding Source: Sandia Laboratories, Albuquerque

129

SEAL STUDIES FOR ADVANCED FLYWHEEL SYSTEMS

I. AnwarFrank)in Research Center

Division of The Franklin Institute20th and Race Streets

Philadelphia, Pennsylvania 19103

ABSTRACT

A seal application for an advanced flywheel system where housing pressure (vacuum)is equal to \0~l* torr is being investigated. Many conventional types of seals can beruled out due to such extreme pressure conditions. For a dynamic sea! to work success-fully, it should develop a compression ratio of 106 magnitude during its normal opera-tion. In view of this, the principle of molecular vacuum pumps, where pressure ratiosof this magnitude are obtainable, is being examined. It is believed that a fiywheel sealsystem based upon molecular vacuum pump principles on the vacuum side and upon viscousflow principles on the high-pressure side can be developed. For future work, a designstudy is recommended where the concepts of molecular drag and turbo pumps are combinedwith one of the conventional configurations of the viscous pumping seal.

INTRODUCTION

The present study involves a review ofvarious types of seals with the purpose ofidentifying seal concepts that could be suc-cessfully applied to an advanced flywheelsystem. Two specific applications are con-sidered. These are:

AutomobileFlywheel

Peak PowerFlywheel

Speed = 40,000(rpm) = 12,000(rpm)

Shaft dia. : 25.<» to = 76.2 to50.8(mm) 101.6(mm)1 to 2(in.) 3 to Mln.)

Vacuum ^ 10

Wheelcapacity

= 10(kW-h)

The basic criteria for the selection ofa seal are: small leakage, low energy con-sumption, and long useful service life.

The secondary criteria are: material,manufacturing cost, auxiliary requirement,failure mode, and experience.

Since the seal is a part of the vacuumsystem, a definition of the overall systemis required. At this stage one could vis-ualize two basic systems. These are:

1, A vacuum pump is employed tolower the flywheel housingpressure to \Q~h torr and thencontinues to operate at reducedcapacity in order to maintainthe inside pressure within theprescribed limits. A systemwith this approach can toleratesome leakage from the seal.

2. A vacuum pump is used to lowerthe flywheel housing pressureto 10"1* torr and is then shutoff. In this case the sealshould have zero leakage at pre-scribed pressure levels andshould act as a pump to cope withany small pressure changes inthe housing.

The second system is, of course, more de-sirable because of reduced energy require-ments.

PERTINENT DEFINITIONS

1. Mean free path is defined as an averagedistance traveled by gas molecules be-tween collisions and is given by:1>2>3

130 -

X = 2.331 x 10T20 L- (cm)

where P = pressure in torr, T = abso-lute temp, in °K, and <5 = moleculardiameter in cm.

2. Gas flow is characterized by a para-meter called Knudsen number de'ined as:

N = iK c

where A = free mean path, c = physicalchamber dimension

N., < 0.01 Continuum Flow (viscous)K

0.01 < N,. < 1 Transition Flow (slip)

N K > . Molecular Flow

Since X is a function of pressure,changes in pressure will cause changesin flow regimes.

3. An accepted terminology for the degreeof vacuum below atmospheric pressure is:

Low Vacuum 760 torr to 25 torrMedium Vacuum 25 torr to 10-3 torrHigh Vacuum 10-3 torr to 10-6 torrVery High Vacuum 10-6 to 10-9 torr

k. Gas flow in vacuum technology is de-fined as:

Q = SP

where S = speed (liter/sec), and P =pressure (torr), Q = throughput (torr-1iter/sec).

Conductance = ~ (liter/sec)

5. The ratio of velocity of flow to thevelocity of sound is known as the Machnumber.

M = u/c

where u = velocity of flow, c = velo-city of sound.

If M < 1, the flow may be consideredincompressible. In theory of lubrica-tion, compressibility or bearing numberis defined as:1*

where y = absolute viscosity of gas;to = angular velocity of moving surface;Pa = ambient pressure; ft =_outsideradius of bearing; c -nominal clear-ance of film gap. If A < 5 compressi-bility effects may be neglected. Ingeneral, the compressibility effectsreduce the increase in pressure in gasbearings and seals.

REVIEW OF SEALS

Seals can be divided into two generalcategories: static and dynamic. Staticseals include 0 ring seals, metal dia-phram type se:ils,and gaskets. For thepresent application, static seals are ofno interest. Dynamic seals can be categor-ized by their motions - rotary, oscillatory,and reciprocating. For flywheel applica-tions, only rotary dynamic seals will beconsidered.

Dynamic seals exist in many configura-tions and sizes and can be classified innumerous ways. However, their operation canbe described in terms of a few fundamentalprinciples. Zuk5 gives the following class-ification:

A. Positive (rubbing) contact: mech-anical face, circumferential, lip,soft packing;

B. Close clearance: hydrodynatnic,hydrostatic, floating bushings;

C. Fixed geometry clearance seals:fixed bushing, labyrinth;

D. Control of fluid properties:freeze, ferromagnetic;

E. Control of fluid forces: centri-fugal, screw pump, magnetic.

In view of the flywheel application,many types of dynamic seals can be elimin-ated. These include: positive rubbing typeshydrostatic fixed bushing, and labyrinth.Seals based on the principle of ferromagne-tic, magnetic,and freezing phenomena arenot to be evaluated in the present study.

The seals that were found to be of in-terest for the flywheel application are theones that have pumping capability. Theseinclude: centrifugal, spiral groove, andviscoseal (screw).

131

Since the pressure is 10-1* torr onthe vacuum side and 760 torr on the highpressure side, a seal in this applicationwill operate in three different flow re-gimes. These are: molecular, transition,and viscous. The review of seal litera-ture indicates that most work has beendone in the viscous regime with the excep-tion of a few studies which have been ex-tended in the transition regime. Therefore,the performance of these seals was evalu-ated only for viscous flow conditions.

EVALUATION OF SEAL PERFORMANCE

The evaluation of seal performance ofpumping type seals indicated that only twoconfigurations have the potential for fly-wheel application. A description of bothof these seals is given below along withthe expressions that govern their perfor-mance.

SPIRAL GROOVE**

A spiral groove face seal is shown inFig. I. A radial pressure gradient developswhen there is relative motion between theflat and the grooved surface. The direc-tion of the pressure gradient will dependon the direction of rotating member and thegroove orientation. Therefore, a spiralgroove seal could be made either inward oroutward pumping. Very often spirals do notextend across the full width of the sealbut terminate in a circumferential sealingdam. The purpose of this dam is to re-strict radial leakage during operation andat shutdown.

k - number of groovesWt = total load (N)i) = viscosity (N sec/m2)ho = groove depth (m)At = film height (m)a, = groove width (m)at = ridge width (m)

<u = angular velocity (rad/sec)H = Aj/Ai5 = At/Ao = W(l - «)y = oi/oihe » hoA = nln

The theory of spiral groove bearingsis considerably developed,and the same re-sults could be used for spiral face sealsfor preliminary analysis. Investigationscover viscous as well as transition regimes.Since the compressibility number is small( a 5 ), results of incompressible flowcan be used to estimate the performance.

Flat Thrust Bearing without Transverse(Radial) Flow for Viscous Conditions

Pressure Distribution:

3nurz

(I-X2)g (<X,H,Y)C (a,H,Y,A,k) (l)

Bearing Load:

W =2h2

Fractional Torque:

2h2(l-Xlt)g2(a,H,Y)

(2)

(3)

where gj, g2» Cj and C2 are functions ofgroove configurations and are given foroptimum geometry.1* The symbols are definedin Fig. 1.

It is noted that the pressure develop-ing capability of the spiral groove config-uration Is reduced when there is a radialflow. The performance of inward and out-ward pumping bearing is discussed in theliterature.1*

Table 1. Performance of spiral groovebearing. (Without radial flow)

outside dia. - 76.2 (mm) 3 (in.)inside dia. - 50.8 (mm) 2 (in.)axial clearance = 0.0025 (mm) 0.001 (in.)li.. = l.8x 10"5.(N sec/m2)Mair

1.8 x 10 ( c/m)?.62 x 10-9 (lbf - sec/in

2)

Fig. 1. Spiral groove configuration

Speed(rpm)

20,000

40,000

(increasein pressure)

12.4 x I03 (M/m2)

1.8 (psl)

24.7 x 103 (N/m2)

3.6 (psi)

PowerLoss(watts)

6.1

24.3

132

The effect of gas rarefication hasbeen considered.6 The analysis was carriedout for Knudsen number 0.01 - 1.0. Theresults indicate substantial reduction inthe performance of spiral groove bearings.

VISCOSEAL (SCREW TYPE)7'8

The viscoseal is a rotating devicewhich will develop an axial pressure grad-ient in the fluid annulus around a shaftby means of a helical groove located eitheron the shaft or sleeve. Figure 2 shows thebasic arrangement of the viscoseal.

O'U-y)-r0tona

a * Helix angleB « (h + c)/cY - b/(a + b)

Fig. 2. Visco seal.

Considerable theoretical and experimentalwork has been done for viscous flow condi-tions, while limited investigations havebeen carried out for rarefied gas conditions.The performance of a seal is generally ex-pressed in terms of a sealing coefficientdefined as:

6UUL

c2Ap(4)

where JJ = viscosity, U = linear velocity,L - axial length, c - rad. clearance, andAp = increase in pressure

For viscous flow:

A"» f (a, B, Y» tan o, Rec)

For transition flow*

A » f (a, B, Y» tan a, l O

(5)

(6)

where a, B, Y. and tan a are seal configu-ration parameters and are shown in Fig. 2.Rec (Reynolds number) * pUc/u, NK (Knudsennumber » A/c, p « density.

The power loss is given by:

q , [$] ^MDLU2

(7)

[$] is a dissipation function and is f(Rec).The values of A and <j> can be found inRef. 7.8.

Sample calculations were carried outfor the present application where a sealoperates in viscous regime. Table 2 showsthese results.

Table 2. Performance of viscoseal (optimumgeometry.

0 - 50.8 (mm) 2 (in.)

L - 2.45 (cm) I (in.)

c - 0.0025 (mm) 0.001 (in.)

U . = 1.8 x 10"5 (N sec/m2)31 r

2

Speed(rpm)

20,000

40,000

.62 x 10-9

A forZero Flow

12

11.4

(lbf

18

2.

39

5.

- sec/in2)

Ap

.8x10*(N/m2)

74 (psi)

.2xlO<l(N/ni2)

7 (psi)

PowerLoss(watts)

5.95

19.9

For transition flow conditions, it has beenshown that degradation in performance of aseal occurs.9*10

REVIEW OF MOLECULAR VACUUM PUMPS

For a seal to work successfully for theflywheel application, it should develop com-pression ratios of the order of 106 duringits normal operation. A preliminary reviewhas indicated that compression ratios ofthat magnitude are achieved by molecularvacuum pumps. This led us to review theprinciples and constructions of varioustypes of molecular pumps.

MOLECULAR DRAG PUMP3.11

In 1912 Gaede introduced a type ofmechanical pump which works on the princi-ple of imparting momentum to gas molecules

133

preferential)y in the direction of thedesired flow. In the molecular drag pumpthere is an open passage from the inlet tothe outlet, between which a pressure dif-ferential is maintained by the high-velocitymotion of one side of the passage relativeto the housing of the pump in which theinlet and outlet are located.

(a) Principal of the molecular drag pump.

(b) Plane representation of themolecular drag pump.

Fig. 3. Gaede molecular pump.

In Fig. 3 the principle of the molecu-lar-drag pump is illustrated. A cylindri-cal member rotates within a casing witha radial clearance h between them. At thetop of the cylinder the clearance space isblocked by a projection of the cylinderwall which reduces the clearance locallyto essentially zero. At either side of theprojection the clearance passage opens intoa closed volume. If there are no leaksin the system the total amount of gas inthe system remains constant, but some gasis shifted by the motion of the rotor witha reduction of the pressure Pj and increasein the pressure P2 . The equilibrium re-lationship between P\ and P2 depends uponthe rotational velocity of the rotor andupon the pressure regime in which thepump is operating, i.e., whether viscousor molecular flow is involved in the pro-cess.

The following gives governing relationsin two different regimes.

Viscous Flow

P2 - Pj =

Molecular Flow

Pa _ S h ) (9)

where K =5 x IP'1* (rf9.71

p - absolute viscosity of gasM = molecular weight of gasT = temperature

For air at 20°C K = 1.62 x 10"5

Based on the simple design describedabove, Gaede built a multistage molecularvacuum pump where compression ratios of theorder of I06 were achieved. The degree ofvacuum produced is significantly affectedby the speed of rotation and fore-pumppressure. Table 3 summarizes these effectsfor the Gaede pump.2

Table 3. Effect of speed of rotation ondegree of vacuum obtained with Gaede molec-ular pump.

Speed ofRotation

(rpm)

12.00012.00012.00012.0006.0002.5008.2008.2008.2006.2006.2004,0004.000

Rough-Pump a

Pressure,P,,,m

0.051

10200.050.050.11

100.11.0I.I1

Pressure on *Fine Side,

P

0.00000030.0000050.000030.00030.0000200003

Not measurable0.0COO20.00050.000010.000050.000030.0003

(8)

mm Hg

A number of alternative designs forthe molecular drag pumps have been devisedwith two considerations In mind:

1. To ensure a low conductance leak-age path from outlet to inletthrough the running clearance ofthe pump.

2. To vary the depth of the pumpingchannel in order to maintain themolecular flow conditions overthe compression range.

The following describes two such pumps.

134

Hoiweek Pump. Figure k shows the main ele-ment of the Holweck pump.12 The innercylinder, A, is made of duralumin and issmooth. The housing, B, is made of bronzeand has a spiral groove cut into its sur-face. The clearance between the cylinderand the housing is made not more than 0.05mm (2 mils). Two spiral grooves are used,one right handed and the other left handed,and they meet at the inlet port C, whichis at the center and connects with thehigh vacuum system. The.depth of thegrooves increases from the ends where itis 0.5 mm (20 mils) to 5 mm (200 mils) atthe center where thegrooves join and con-nect to the inlet port C. In operationthe vacuum pump is first pumped by a suit-able forepump down to a pressure of a fewmicrons. The molecular pump is then turnedon, being backed by a forepump. TheHolweck pump operates at speeds of about5000 rpm and produces pressure down to 10"6

torr.

To vacuum system

Ball bearingsChannel

| To fore Vacuum

Induction motor

Fig. k. Holweck molecular drag pumprotor dia. = 152 mm (6 in.)

Siegbahn Pump.2'3 Figure 5 shows a cross-section of the Siegbahn pump where pumpingchannels in the form of Archimedes' spiralsare cut in the two flat sides of the hous-ing, within which a disk rotates at highrotational velocity. The clearance betweenthe disk surface and the flat section ofthe end plate between the adjacent spiralsis made as small as practical for freerotation. The inlet is at the peripheryof the disk and the discharge at the hub.In the figure shown three spiral grooves(22 x 22 mm at inlet and 22 x 1 mm center)are cut in parallel starting 120° apart,providing three times the pumping speed ofa single channel.

Rotatingdisk

Housing

Fig. 5. Molecular drag pump of Siegbahn.Disk dia. = 5*»0 mm (21.26 in.)

All the pumps described above use airgaps of 0.02 to 0.05 mm maximum (1 or 2milt). Nonuniform thermal expansion, thepresence of small foreign particles of theorder of tenths of millimeter (a few mils)in diameter, or a sudden air shock canresult in seizure of the rotor. A recentdesign intended to overcome this type ofdifficulty is described below.

TURBO OR AXIAL FLOW MOLECULAR PUMP

Figure 6 shows the general arrangementof a turbo pump which was first describedby Becker.3'13 Rotating disks all mountedon the central shaft are disposed alter-nately with stationary plates mounted inthe housing. The disks and plates are cutwith slots set at an angle so that gasmolecules caught in the slots of the movingdisk are projected preferentially in the di-rection of the slots in the stationaryplates. The running clearances between therotating and stationary plates generally areof the order of 1 mm (4o mils), which is anorder of magnitude greater than the permis-sible clearances in a conventional type ofmolecular pump. The rotational speed for apump having a rotor diameter of about 17 ~m(6.5 in.) is 16,000 rpm, giving a peripheralspeed of 1.56 x 104 cm/sec (6.14 x 103 in/sec), about one-third average velocity forair molecules at room temperature.

135

Vacuum SystemRotating \ stationaryDisk \ f

To ForeVacuum

Gas

Rotor Disk Rotor Disk

X-Section of Disks

Fig. 6. Turbomolecular pump, rotordia. = 165 mm (6.5 in.) (Becker)

10

Fig. 7. Observed dependence of inletpressure on outlet pressure for H2air, Freon-I2.

The observed dependence of the inletpressure Pj on the outlet pressure P2 isshown graphically in Fig. 7 for hydrogen,air, and the refrigerant Freon-12. A com-pression ratio P2/P1 of the order of 107 isobtained for air when P2 is equal to 0.1torr, but the value drops off rapidly withincreasing pressure to about 10 when P2 isequal to 1 torr. The compression ratio forhydrogen is significantly smaller for thesame outlet pressure.

According to Becker, the turbomolecularpump is like a Gaede pump except that itoperates with large clearance between thecooperating drag surfaces, and with thesepumping drag surfaces fed by a series of in-clined slots in such a way as to greatlyimprove the pumping speed over that of theGaede pump.

Kruger and Shapiro1**'15 produced atheory of turbomolecular pump performancebased on the kinetic theory of gases. Thebasic concept of the theory, which is appli-cable in the free molecule range only, isthe transmission probability; that is, thefraction of molecules entering a rotor orstator from one side which ultimately leavethat same rotor or stator from the otherside. The theory was validated by experi-ments on a single test rotor. The followingsummarizes some of their important conclu-sions.

1.

3.

if.

Pressure ratio developed per stageis dependent on the ratio of bladespeed/velocity of gas molecules.

A pressure ratio 5:1 (per blade)can be achieved for air with bladespeed = 1000 ft/sec. This wouldgive a pressure ratio per stageof 25:1.

Performance begins to fall whenthe mean free path is about fivetimes as large as the minimumblading dimensions.

For a single rotor the geometricvariables are the blade angle andratio of blade spacing (circum-ferential ly)/blade chord length.

HYBRID MOLECULAR PUMP

Alcatel16 has recently developed a newmolecular pump called the "hybrid molecularpump." The pump has two stages. The firststage uses the turbomolecular structure(open blading), while the second stage

136

has a drag pump construction. The arm ofthis concept is to combine the high com-pression ratio of a drag pump with thehigh pumping speed of a turbomolecularpump. Figure 8 shows the pump in detail.The turbo part has three stages on therotor and four for the starter. The dragpump is of multigroove type and the groovesare in the stator. There are five grooveson the low pressure side and 15 on thehigh pressure side. The following givessome of the pump parameters.

Turbomolecular Stage

Rotational speed = 24000 rpm

Pump diameter = 200 mm (= 7.8 in.)Pumping speed = 460 H/sCompression ratio at zeroflow (Nitrogen) = 150

Drag Stage

Pump diameter = 138 mm (= 5-4 in.)Cylinder length = 80 mm (» 3.15 in.)Depth of groove varyingPumping speed - 45 H/sCompression ratio at zero flow (forNitrogen) = 109

Fig. 8. Cutaway of Alcatel "Hybrid Pump"Rotor dia. = 200 mm (7.874 in.) turbo.

= 138 mm (5.433 in.) drag.

The pump uses a dynamic seal which hasunique design features. The followingdescribes the seal in detail.

Dynamic Seal. Alcatel17 developed ascrew-type seal to withstand a pressuredifference from 1 x 10"3 to 760 torr for50 mm (2 in.) diameter shaft at 24,000 rpm.Figure 9 shows the seal. The seal isspecifically designed to operate in threedifferent flow regimes, i.e., molecular,transition, and viscous. For each of theflow conditions efforts were made to opti-mize the geometry of the threads.

-THREADS

Fig. 9. Dynamic Seal for Alcatel Hybrid PumpRotor Dia. = 50 mm (1.968 in.)Threaded Shaft Length - 50 mm (1.968 in.)Radial Clearance - .01 mm (.0004 in.)

to .015 (.0006 in.)

Figure 10 shows the pressure vs. flowrate. It is interesting to note that aseal designed to withstand a pressuredifference between 760 and 1 x 10"3 torrwii? leak when it withstands a bigger pres-sure difference. On the contrary, if thepressure difference is smaller the sealwill pump. This property is unique asthe dynamic seal could be used as a backuppump.

to

19

4

10

-

SPIHDLE 40*.(.ei. lo.0. /H» 24,000 rpm 1

1/PRESSURE ,»rr

Fig. 10. Vacuum pressure vs.Fore pressure = 760 torr(Alcatel dynamic seal)

10- '

flow.

137

CONCLUSIONS

1. Many conventional types of seals may beeliminated because of high rotationalspeed and extreme pressure conditions.These include face seals, lip seals,and packings.

2. Close clearance seals such as capillary,face, were evaluated. These were foundnot to be suitable due to high leakagerates.

3. Seals of interest were found to be thosewhich have the pumping capability.These are: centrifugal, spiral groove,and viscoseal (screwj. Since thepressure is lO""* torr on the vacuumside and 760 torr on the high pressureside, a seal in this application willoperate in three different flow regimes,i.e., molecular flow, transition, andviscous. The review of literature in-dicated that most of the work has beendone in viscous flow regime with the ex-ception of a few studies which have beenextended in the transition flow regime.However, evaluations were carried out atviscous flow conditions to get some ideaof their relative performance. Theresults indicate:

a. The visco seal has the lowest powerlosses-to-pressure sealed ratio.

b. The performance of all the seals de-teriorates when the flow conditionschange from viscous to transition.

c. It was also concluded that a seal de-sign based on the continuum fluiddynamics will show further deteriora-tion in its performance whenapplied in molecular flow regime.Therefore, theories developed forthese seals cannot be applied to de-sign a seal for the present applica-tion.

To study the pressure building abilityof gases in molecular flow conditions,theory of molecular pumps was reviewed.Basically there are two types of molecu-lar pumps; drag and turbo, both workingessentially on the same principle, buttheir construction is entirely different.Both types of pump can achieve compres-sion ratio of the order of 107 providedthe gas is in molecular flow condition.A drag pump achieves higher compression

ratios for a given length and runs atsmall radial clearance (0.05 mm = 2mils), whereas turbo is a multistagepump with high pumping speed andoperates at much larger radial clear-ance (0.5 mm = 20 mils).

5. The performance of molecular pumps,drag and turbo, is significantlyaffected by the fore or downstreampressure. Typically a molecular pumpwith fore pressure less than 0.01torr, can develop a compression ratioabout I0G; but for higher values offore pressure compression ratio fallsoff drastically.

6. To k^ep the gas flow in molecular flowregime, physical dimensions of thepump are varied along the length ofthe rotor. Generally one tries tokeep the values of minimum-pump di-mensions to about 1/4 to 1/5 of meanfree path of gas.

7. The review of fluid seals and molecu-lar pumps Indicates the principle ofmolecular pumps can be combined withthe principle of viscous pumping sealsin order to develop a seal for anadvanced flywheel system. The analysiswill require defining the three flowzones along the length of the seal.Once the flow zones are defined,theories of molecular pump, slip flow,and viscous flow can be used to pre-dict the performance.

8. Since molecular pumps are of two con-structions, two seal configurationsare possible. The advantage of dragtype construction is that it is simpleand develops high compression ratioas compared to the turbo type. However,the disadvantage of the constructionis that it runs at fairly small radialclearance (1 to 2 mils).

RECOMMENDATIONS

1. Investigate the screw types of sealconcepts where the design on the vacuumside is based on the principle ofmolecular drag pump and continuum fluiddynamics on the high pressure side.For analysis, existmg works could beused, and efforts should include ex-tensive parametric studies prior toselecting the final configuration.

2. Investigate the seal concept which usesthe construction of a molecular turbo

138

pump. For analysis, existing workscould be used, and efforts should includeoptimizing the rotor and stator designat each stage.

3. For both seal concepts (1 and 2 above),investigate a design for control ofleakage at start and shutdown.

4. Evaluate both designs and select one formanufacturing, testing, and final evalu-ation.

REFERENCES

^.J. Santeler, D.W. Jones, O.H. Holkeboer,and F. Pagano, Vacuum Technology and SpaaeSimulation (NASA SP-1O5, 1966).

2S. Dushman and J.M. Lafferty, ScientificFoundations of Vacuum Technique, 2nd ed.(John Wiley 6 Sons, 1962).

3C.M. Van Atta, Vacuum Science and Engi-neering (McGraw-Hill, New York, 1965)."*E.A. Muijderman, Analysis and Design ofSpiral-Groove Bearings, J. Lub. Tech.(July 1967), pp. 291-306.

5J. Zuk, Dynamic Sealing Principles, (NASATMX-71851, April 1976).

6F.C. Hsing, S.B. Malanoski, Mean FreePath Effect in Spiral-Grooved Thrust Bear-ing, Paper No. 68-Lubs-17, presented atLubrication Symposium, Las Vegas ASME(1968).

7E.F. Boon and S.E. Tal, UydrodynamisaheDichtung fur Rotiersnde Wallen, Chemie-Ing-Technik Vol. 31, No. 3 (Jan. 1959),p. 202.

8W.K. Stair and R.H. Hale, The TurbulentVisco Seal-Theory and Experiment, ThirdInternational Conference on Fluid Sealing,Cambridge, England, Paper H2 (April 1967).

9M.W. Milligan and H.J. Wilkerson, Theo-retical Performance of Rarefied-Gas Vis-coseals, ASLE Transactions 13 (1970),pp. 296-303.

10M.W. Milligan and H.J. Wilkerson, Visoo-seal Performance for Rarefied Gas Sealant,5th International Conference on FluidSealing, Warwick, England, Paper Bl(1971).

1]W. Gaede, The Molecular Air Pump, Ann.Physik, 41 (1913), pp. 337-380.

12A. Guthrie, Vacuum Technology (John Wiley6 Sons, 1963).

13W. Becker, Zurtheorie der Turbo-Molekular-Pumpe, Vakuum-Technik 7 (1961), pp. 199-204.

llfC.H. Kruger and A.H. Shapiro, VacuumPumping with Bladed Axial Flow Turbo-machine, Trans, of 7th Symposium on

Vacuum Technology (Peramon Press, 1961),

pp. 6-12.15G.E. Osterstrom and A.H. Shapiro, ImprovedTurbomoleeular Pump, J. Vac. Sci. Tech.,Vol. 9, No. 1 (1972).

16L. Maurice, A New Molecular Pump, Proc.6th International Vacuum Congr. 1974,Japan, J. Appl. Phys. Suppl. 2, Part 1(1974).

17L. Maurice, Dynamic Seals, Proc. 6thInternational Vacuum Congr. 1974, Japan,J. Appl. Phys. Suppl. 2, Part 2 (1974).

139

PROJECT SUMMARY

Project Title: A Composite Flywheel for Vehicle Use

Principal Investigator: F. C. Younger

Organization:

Project Goals:

Project Status:

William M. Brobeck & Assoc.1235 Tenth StreetBerkeley, CA 94710415/524-8664

Design and fabrication of fiber-composite flywheelswith a usable energy in the range of 1 to 5 kWh with atotal energy density in excess of 80 Wh/kg at themaximum operating speed. The design power output rateis 37 kW. The maximum diameter is 0.6 m and themaximum thickness is 0.2 m.

A preliminary design of the flywheel has been completed.This design uses a biannulate rim of S2 fiberglass/epoxy overwrapped with Kevlar 49/epoxy. This rim issupported from an aluminum hub by polar catenaryspokes filament wound with Kevlar 29/epoxy. The com-bination of materials and shapes yields a compatibleset of stress/strain relationships for all of thefiber-composite components to insure that centrifugallygenerated stresses are principally in the direction ofthe various fiber orientations with minimum stressestransverse to the fibers. The geometrical parametershave been selected to provide that the various fibersare all stressed to approximately the same percentof their ultimate strength to assure that the maximumenergy density will be achieved. Design calculationsshow that project energy and energy density goals willbe satisfied.

Design drawings and specifications for the biannulaterim have been prepared and sent to fabricators forquotations. It is expected that orders will beplaced by the end of October 1978.

Design work is presently concentrated for optimum spokeand hub parameters. Adequate rigidity of the spokesis required to insure dynanic stability. However,excessive rigidity will give undesirable transversestresses and a reduction in maximum energy density.Stability criteria have been established, and designswhich satisfy these criteria are being subjected tostress analysis. Winding forms for the spokes arebeing designed, and spoke samples for testing willsoon be fabricated.

The design of assembly fixture has not yet started,although preliminary concepts and methods have beenbriefly studied to identify the problem areas andindicate probable solution methods.

141

Contract Number: 13-0291

Contract Period: The contract covers only the first phase of atwo-phase program. The first phase contract periodis nine months and will be completed by the endof April 19.79.

Funding Level: $99,000

Funding Source: Sandia Laboratories, Albuquerque

142

A COMPOSITE FLYWHEEL FOR VEHICLE USE

Francis C. YoungerWilliam M. Brobeck & Associates

1235 Tenth StreetBerkeley, California 94710

ABSTRACT

A program for the design and fabrication of a fiber-composite flywheel is described.The objective of the program is a usable energy in the range of 1 to 5 kWh with a totalenergy density in excess of 80 Wh/kg at the maximum operating speed. The design outputpower is 37 kw and the maximum dimensions are 0.6 m diameter and 0.2 m thickness. Thedesign uses a biannulate rim of S2 fiberglass/epoxy overwrapped with Kevlar 49/epoxy.This rim is supported from an aluminum hub by polar catenary spokes filament wound withKevlar 29/epoxy. The Kevlar 29/epoxy spokes have a lower tensile modulus than the rimmaterials so compatible strain levels can be achieved without excessive stress in thespokes. The stress analysis for the biannulate rim shows nearly optimum tangentialstress levels in each of the two materials and a very reasonable distribution of radialstresses. Compressive radial stress occurs at the interface between the S2 fiberglass/epoxy and the Kevlar 49/epoxy. The spoke design concept permits a careful balance ofspoke flexibility to assure that the spoke stiffness is adequate to maintain concentri-city of the rim and hub to satisfy dynamic stability requirements while at the same timebeing flexible enough to allow dilation of the rim due to centrifugal loading withoutimposing excessive radial loads at the rim-to-spoke attachment points.

INTRODUCTION

A design and fabrication effort isunderway to (1) produce an automotive typeflywheel having suitable mechanical char-acteristics of energy density, volume, andcycle life; and (2) to establish the prob-able cost of production of such flywheels.The approach to the development of a fly-wheel to satisfy specific performancespecifications is outlined. A preliminarydesign of a flywheel to meet the perform-ance specifications is presented.

During the course of the work, thispreliminary design is being completelyanalyzed and reviewed as required to assurea flywheel which is economically and tech-nically feasible. Material and geometrictrade-offs are being examined to maximizethe probability of achieving a flywheelwhich satisfies the program objectives.Our primary effort is directed toward theobjective of achieving a maximum totalenergy density (Wh/kg). However, we alsorecognize that the hub-wheel interface isa critical problem area and that a success-ful solution of hub-to-wheel attachmentmay require some compromise in maximumenergy density as may the requirement toprovide a high energy within a limitedvolume (Wh/m3).

The program is to be divided intotwo phases:

Phasa I: Design and fabricatespecimen wheels togiven specifications.

Phase II: Testing and character-ization of specimenwheels.

This paper covers the first phaseonly. Upon completion of this firstphase, at least two specimen flywheels willhave been fabricated and the requireddocumentation will have been delivered.

BACKGROUND

The effort required for the design,fabrication, and documentation of at leasttwo specimen flywheels must satisfy thefollowing specifications:

WHEEL SPECIFICATIONS v

(1) The usable energy, defined as thedifference in kinetic energy betweenthat stored in the wheel at itsmaximum design operating speed and at33% of that speed, will be in therange of 1 to 5 kWh.

143 —

(2) As i design objective, the total energydensity will be at least 80 Wh/kg atthe maximum operating speed. For thepurposes of this calculation, theweight of the wheel will be taken asthe weight of the flywheel itself,including its hub and whatever por-tion of a shaft is necessary to coupleto it. The weight of additionalshafting, mechanical drive components,etc., are not to be charged to thewheel nor is their kinetic energy tobe credited to it.

(3) The overall diameter of the rotatingassembly will not exceed 0.6 m, and itsthickness must be less than 0.2 m.

(4) The wheel shall be designed to run for1000 total hours at its maximum designoperating speed without failure. Itis to be designed to be cycled from33% to 100% of that speed, 10,000times without failure. During test-ing it will be required to demonstrate100 such cycles without failure.During tests, the wheel will be spundown at an average power output rateof 37 kW.

(5) Choice of materials is at the discre-tion of the contractor, although theobjective of the program is to producea wheel which is economically as wellas technically feasible.

The required documentation will includebut not be limited to the following:

(1) Complete construction detailsas adequate to allow the wheelto be duplicated by othersskilled in the necessaryarts. These details willinclude such things as inform-ation regarding specialmaterials and any unusualfabrication techniques suchas the use of specializedwinding facilities.

(2) An analysis of the stressesinduced in the wheel duringfabrication and operation,with an evaluation of fatigueeffects and of the scalinglaws involved in changingthe size of the wheel.

(3) An analysis of the dynamicsof the wheel with particularemphasis on critical speeds and

the effects of materialcreep upon balance.

(4) An estimate of the costs toproduce similar wheels inquantities of 100, 1000,and 10,000. Note will betaken of trends in the costof any exotic materials.

PROJECT DESCRIPTION

High strength-to-weight ratio fibercomposite materials seem to have thegreatest promise for satisfying the ob-jectives of an automotive energy-storageflywheel. However, the orthotropic natureof these materials presents some seriousdesign problems. Their low transverse .strengths make it difficult to transferloads from one radius to another. A simplethin rim spinning about its axis can storea great amount of energy per unit weight.However, finding a suitable support for thering is difficult because the support mustnot impose large radial loads. Increasingthe rim thickness can beneficially lowerthe attachment loading by reducing theradius at which the rim support is to beattached; however, increasing the thick-ness imposes difficult radial stresses inthe rim.

A design concept to avoid these prob-lems employs a rim supported by a system oftension-balanced spokes which minimize theradial forces upon the rim. The designfeatures a biannulate rim composed of twoconcentric components of different materialtightly bonded at their interface. Thisbiannulate arrangement permits efficientutilization of the available space, allowsuse of material with differing specificmoduli of elasticity and reduces theradial span of the tension-balanced spokes.Figure i shows the design.

An energy density in excess of 84Wh/kg appears to be possible with such adesign using available technology. Thisarrangement is dynamically stable and willpermit more than adequate power extraction.

RIM DESIGN

The design objective for very hightotal energy density requires that much ofthe material be highly stressed by centri-fugal loading. A radially thin ring spin-ning about its axis is a well-known exampleof a geometric arrangement where all thematerial is uniformly stressed in pure

144

tension (in the limit of infinitesimalthickness). For such a geometry, it may beshown that the energy density is related tothe strength-to-weight ratio of the materialby the equation:

w " 2 p u ;

where E = the energy in Joulesw = the weight in kilograms0= the strength in Pascalsp = the density in kilograms/metre3

The factor of one half can be consideredas a shape factor for a very thin ring. Asthe radial thickness of the ring is in-creased, this value drops off slowly. Forexample, bringing the radial thickness to11 percent of the outer radius would dropthe shape factor from .5 to .465. In-creasing the radial thickness of a ringincreases the total energy stored in theavailable space, while lowering the totalenergy density (Wh/kg).

The requirement for usable energy inthe range of 1 to 5 kWh from a flywheelnot to exceed 0.6 metres in diameter witha thickness less than 0.2 meters indicatesthat the fiber composite must occupy afairly large fraction of the availablespace. Economic feasibility and vehicleintegration also demand an efficientutilization of space. From our analysis,it appears that a thin ring would not pro-vide the most efficient utilization ofavailable space and that a thick ring orcombination of rings is necessary.

Of course, there are also disk-typeflywheels utilizing radial fibers and cross-ply composite. We recognize that theseother type flywheels also may provide anefficient utilization of space (Refs. 1,2, 3, 4). However, we prefer a designsimilar to those which we have tested thatshow good promise of success.

The stresses in a thick ring may becalculated by equations in Ref. 5. Theequations in Ref. 5 are for isotropicrings rather than for orthotropic rings.However, comparisons of calculations ofthe stresses for isotropic and orthotropicmaterials (Ref. 6) shows that for ringswith radial thicknesses less than 20 per-cent, the simpler isotropic equations arequite adequate. For centrifugal loading,the radial and tangential stresses arefound by:

and the dilation is found from:

The maximum radial stress is found as:

•_ awhere a = r-

» •

a = inner radiusb = outer radiusu = angular velocityr = radius at which stress is

calculatedR = outer radiusw = weight densityE = modulus of elasticityg = gravitational constanta = unit stressu = Poisson's ratio

Thus, increasing the radial thicknesscauses a large increase in radial tensilestress which could lead to failure becausefilament-wound fiber-composite materialshave very low transverse tensile strength.There are several ways to circumvent thedifficulty. The one we will use is tofabricate a biannul ate ring with twoconcentric elements having materials withdiffering densities and tensile moduli.*These two rings can be wrapped one on topof another so that they will bond on curing.Under the action of centrifugal loading,the dilation of the innermost ring will beresisted by the outer ring and a pressureat the interface will be produced.

The interfacial pressure can be com-puted from the deflection equations andthe requirements for continuity. Reference5 gives the stresses and deformation ofisotropic thick rings subjected to exter-nal and internal pressure.

*This design is quite similar to one iden-tified by Reedy and Gerstle (Ref. 7).This build-up gives a combined radialthickness great enough to give excellentspace utilization and to reduce the spanof the spokes to an acceptable value.

145

For two concentric rings in intimatecontact, the external pressure of one equalsthe internal pressure of the other and con-tinuity requires that the total deformationof each at the interface must be equal—Thedeformation being that due to the combinedaction of the interface pressure and thecentrifugal loading. Superposition isused for finding the total deformation; thus,the deformation of the interface edge of theinner rings can be represented by the follow-ing equations:

Pal,2 = \Pa2,2 = h

(6)

(7)

where P = interface pressure<*•. i = dilation of outer edge of inner

' ring due to centrifugal force atunit angular velocity

a2 , = dilation of inner edge of outer' ring due to centrifugal force at

unit angular velocity .'a, « = dilation of inner ring due to

' unit pressuredilation of outer ring due tounit pressure

= dilation at the interface

a2 2

1si = dilation at the interface

By setting 6j = 62, the value of P canbe found as a function of u> as below.

P = (8)

By using these relationships, a two-part rim using an inner ring of S-2 fiber-glass and an outer ring of Kevlar 49 hasbeen designed. This arrangement allows anoverall radial thickness of slightly over25 percent. Thus, slightly more than 45percent of the available volume is used forthe rim. The centrifugal force field atthe spoke-to-rim attachment point is abouthalf that at the rim outer edge.

Figure 2 shows the stresses in theproposed two-part rim at 32,725 rpm, inthe absence of creep, thermal stresses andfabrication stress. All the values inFig. 2 vary with the square of rotationalspeed. Thus, all stresses are zero atzero speed.

TENSION-BALANCED POLAR CATENARY SPOKES

The means of connecting the rim tothe hub is at least as important as the

design of the rim itself. We expect thisto be a difficult engineering problem forany developer of flywheel technology.Allowing rim dilation while maintainingconcentricity is a difficult task. Webelieve that tension-balanced polar-cate-nary spokes provide an acceptable solutionto this problem. They will provide ade-quate rigidity and torque-carrying capa-city.

The rim-to-hub attachment methodemploys a system of tension-balanced polar-catenary spokes as shown in Fig. 1. Apolar catenary is the naturally-assumedcurved shape of a flexible element undertension in a centrifugal force field.These spokes will be pre-formed on aspecial template to obtain the desiredpolar-catenary shape to insure pure ten-sion loading with minimum radial loadingof the rims at the points of attachment.

Continuity requires that the averagestress level in the spoke be proportionalto the stress in the rim at the point ofattachment, with the proportionalityfactor being equal to the ratio of tensilemodulus of the spoke material to that ofthe rim. It is desirable to use a spokematerial with a modulus lower than that ofthe inner portion of the rim in order toallow spoke stresses to be lower than rimstresses.

Figure 3 shows the polar-catenaryshape and the force balance on an element.The derivation of the shape is given inRef. 8. The amount of curvature of thespoke is determined by its tension and den-sity. Increasing the tension or loweringthe spoke density gives less curvature.The amount of curvature limits the radialspan which can be bridged by a practicalspoke. For the proposed design, the hub-to-rim radius ratio is almost 3 to 1. Thisratio appears to be too great for an S-2fiberglass spoke but appears to be accept-able for a Kevlar 29 composite spoke. TheKevlar 29 is an ideal candidate materialfor the spokes because of its low densityand because its modulus of elasticity isjust slightly less than that for S-2 fiber-glass. This modulus is low enough toinsure reasonable stresses while at thesame time is high enough to provide ade-quate rigidity for the hub-to-rim connec-tion . Inadequate rigidity would lead todynamic instability of the ring support.

The spokes will provide a sufficientlyaccurate and rigid hub-to-rim connection

146

to insure balance and dynamic stability.The concentricity of the hub and rim willnot be absolutely perfect. Some smallinitial or residual eccentricity will exist.Thus, at high speed a centrifugal force pro-portional to the square of the angular ve-locity and the first power of the offsetdistance between the center of mass of therim and its center of rotation will begenerated.

The combined stiffness of all thespokes oppose this force but some increasein the offset will occur. This increaseis inversely proportional to the combinedspring constant of the spoke system andmay be expressed as:

Ax = F/K (9)

where "K" is the spring constant of thespoke system.

Even though tha initial eccentricityof the rim may be very small, there willbe a force which will cause the eccentricityto increase which will in turn permit theforce to further increase. In order tohave elastic stability, the force associatedwith the spoke stiffness at a given offsetmust be greater than the centrifugal forceassociated with that offset. This requiresthat:

K > Moo2 (10)

Thus, there will be a critical speed u

(ID

The curved spokes appear to be quiteflexible. A force applied at the end ofthe curve spoke tends to straighten thespoke and also acts to stretch the spokematerial. Thus, there are two componentsof the elasticity of the spoke. Thestretching of the spoke material is direct-ly related to the modulus of elasticity.The straightening of the spoke is morecomplicated and is found to be dependentupon the centrifugal loading on the spoke(Ref. 8). The overall spring constant,Kg can be expressed in terms of the com-bination of these two components.

from:The stretch of the spoke is found

* _ £k61 EA

where F = the force on a spoke elementL = the length of spokeE = modulus of elasticityA = cross-sectional area

The displacement associated withstraightening the spoke is found from(Ref. 8 ) :

(13)

where w - density of spoke materialR = hub radiusu> = angular velocityL = curved length of spokex = length of chord of curved spoke

Since the component 6% is dependent upon u2

and the density of the spoke, the centri-fugal force on the spoke acts to increaseits stiffness.

The requirement for the criticalspeed to be in excess of the operatingspeed sets a minimum value of the overallspring constant, which in turn sets aminimum cross sectional area for the spokes.This minimum cross sectional area can befound using Eq. (12) and (13).

In addition to the stability ofradial offset (eccentricity), there willbe a requirement that the spoke system pro-vide adequate rigidity to prevent instabilityof out-of-plane motion. What is importanthere is that the principal axis of the rimand the hub will not be absolutely parallelwith the axis of rotation; consequently,there will be a gyroscopic torque actingbetween the rim and hub tending to changethe alignment of their axes. This changein alignment will be resisted by the stiff-ness of the spoke system.

The out-of-plane whirl modes are de-pendent upon the rotational frequency andthe ratio of the polar moments of inertiato the diametral moments for both the huband rim. The moment of inertia of the rimis so much greater than that of the hub thatthe frequency of this mode may be approxi-mately calculated from the moments of thehub alone. The approximate relationshipat low speed is:

(12) -VF (14)

14 7

where K is the spring constant defined asthe restoring torque due to a unit of angularrotation of the hub about its diameter.

At high rotational speed, the gyros-copic moment of the hub becomes signifi-cant and the frequency for this modeincreases to approach:

= An (15)

where a is the rotational speed of the fly-wheel and A is the ratio of polar momentof inertia of the hub to its diametralmoment.

For a thin disk A equals 2 but for thethick hub of the flywheel:

A = 2r2/(r2 + V ) (16)

where r = radiusa = axial height

At very high speeds, the modal fre-quency will be less than the rotationalfrequency. Figure 4 shows the shape of thecurve of whirl mode frequency as a functionof rotational speed. The point where themodal frequency equals the rotational fre-quency is shown at point P. This criticalspeed for self-excitation of whirl is al-ways greater than the value from Eq. (14).Operating above the critical speed is tobe avoided as dynamic stability cannotbe assumed. By raising the stiffness ofthe spoke system, the critical speed canbe kept above the maximum operating speed.

The relationship for dynamic stabilityfor the whirl or wobble motion is similarto that for the radial offset so that acritical stiffness must be exceeded. Thiswill require axial spacing of an array ofspokes.

PRELIMINARY DESIGN

The preliminary design for the fly-wheel is shown in Fig. 1. Its character-istic? are summarized in Table 1. Theavailable energy of 4.60 kWh is withinthe desired range and its energy densityof 84.3 Wh/kg is above the desired minimum.In order to finalize this preliminary design,trade-off studies, additional stress cal-culations, and further dynamic analysisare required. The effects of dimensionaltolerances and material variations arebeing considered. Preliminary engineeringdesign data from the Lawrence Livermore

148

Laboratory are used as a guide for fiber-composite properties. Where additionaldata are required, specimens will be pre-pared and tested.

The main features of the design willbe retained as it is believed that thisconcept provides promise for solving theproblems faced by filament-wound fiber-composite flywheels.

The stresses associated with acceler-ation and deceleration are very small athigh speed. A preliminary examinationshows that the added spoke tension due todeceleration is a few pounds compared toabout 20,000 pounds for centrifugal load-ing.

REFERENCES

1. Hatch, B. D., "Alpha Cross-Ply Com-posite Flywheel Development," 1977Flywheel Technology Symposium, CONF-771053, March 1978.

2. Lewis, A. F. and Gupta, P. W., "Opti-mization of Hoop/Disk CompositeFlywheel Rotor Designs," 1977 FlywheelTechnology Symposium, CONF-771053,March 1978.

3. Rabenhorst, D. W., McGuire, D. P.,and Lewis, A. F., "Composite FlywheelDisk/Hub Attachment Through Elasto-meric Interlayers," 1977 FlywheelTechnology Symposium, CONF-771053,March 1978.

4. Garber, A. M., "Polar Weave CompositeFlywheels," Proceedings of the 1975Flywheel Technology Symposium, ERDA76-85.

5. Timoshenko, S., "Strength of Materials,Part II," Third Edition, D. VanNostrand Company, Inc., 1956.

6. Morganthaler, G. F., and Bonk, S. P.,"Composite Flywheel Stress Analysisand Material Study," Society ofAerospace Material and ProcessEngineers, Vol. 12, 1967.

7. Reedy, F D., Jr. and Gerstle, F. P.,Jr., "Dfc-ign of Spoked-Rim CompositeFlywheels," 1977 Flywheel TechnologySymposium, CONF-771053, March 1978.

8. Younger, F. C , "Tension-BalancedSpokes for Fiber-Composite FlywheelRims," 1977 Flywheel TechnologySymposium, CONF-771053, March 1978.

FIBER - COMPOSITE

Fig. 1. Tension balanced catenary spoke flywheel.

149

091

Tangential Stress x 10 psi

O

1—*

oo

I—»ino

rooo

ino

Radial Stress x 10 psi

Interface

H + AH

V + AV T + AT

Fig. 3. Polar catenary force diagram.

Fig. 4. Whirl mode frequency as a function of rotational speed.

152

Table 1. Flywheel Characteristics

Outside Dia.Inside Dia.HeightMaterialFiber FractionDensityElastic ModulusPoisson's RatioWeightMoment of Inertia

Hub

Outer Ring

23.52 in19.70 in7.85 in

Kevlar 49/epoxy60% ,

.050 lbs/inJ

11.76 x 106

.3150.9 lbs „15.4 Ib-in-sec

Inner Ring

19.70 in16.50 in7.85 in

S-2 Glass/epoxy60% ,

.075 lbs/inJ

8.15 x 106

.28253.6 lbs 9

11.4 lb-in-sec^

Spj

TypeLoadi ng

MaterialFiber FractionDensityModulusNo. of SpokesTotal WeightTotal Momentof Inertia

ikes

Polar-catenaryTension-balancedw/ added weightsKevlar 29/epoxy65% ,.05 ibs/iV

5.85 x 10616 x 44.64 lbs

.41 lb-in-sec^

Summary of Characteristics

DiameterHeightMaterialDensityWeightMoment of Inertia

6 in7.85 inAlum.alloy 7075.1 lbs/in3

22.2 lbs ,.26 Ib-in-sec

Loading Weights

MaterialNo. of WeightsWeight EachTotal WeightMoment of Inertia

Aluminum8.50 lbs

3.98 lbs.63 lb-in-sec

Total WeightTotal Moment of InertiaRotational Speed

MaxMaxMaxMaxMaxMax

135.3 lbs ,28.09 lb-in-sec

32,725 rpmu 3,427 rad/secStored Energy 5.18 kWhEnergy Density 84.3 Wh/kgAvailable Energy 4.60 kWhMax. Hoop Stress in Kevlar/epoxy 216,736 psi

Radial Stress in Kevlar/epoxy 3,013 psi compressionRadial Tension in Kevlar/epoxy 1,144 psiHoop Stress in S-2/epoxyRadial Stress in S-2/epoxyRadial Tension in S-2/epoxyTension in Spokes Kevlar 29/

epoxyCritical Speed for RadialStability

189,349 psi3,013 psi compression1,004 psi

163,057 psi64,000 rpm

153

PROJECT SUMMARY

Project T i t le : Prototype Development - Composite Flywheel Having Nominally40 Watt-hrs/lb Energy Density

Principal Investigator: P. W. Hil l

Organization: Hercules Aerospace DivisionHercules IncorporatedAllegany Ballistics LaboratoryP. 0. Box 210Cumberland, MD 21502(304) 726-4500

Project Goals: The development of hardware that wi l l demonstrate the very latesttechnology relating to composite wheels. This project is oneof four that explore different approaches to producing a wheelhaving the highest practical energy density.

Project Status: This concept is unique in that i t incorporates a solidcircumferentially-wrapped composite wheel with an elongatedhour glass cross-section. Such a profi le places most of themass near the periphery of the wheel ~ as distinguished fromthe more common modified Stodole disk shape which concentratesthe mass at a relatively snail radius. Considerable attentionis being given to the disk-hub interface, where the use of anelastomer as a grading medium is being explored.

Sandia's work in aerodynamic heating of composite wheels is beingincorporated at the design stage in the interest of managingthermal stresses.

At present, design and optimization of the disk contour arecomplete. Aerodynamic heating analysis is under way, and dynamicanalyses by a subcontractor (Rockwell) are beginning.

Contract Number:

Contract Period:

Funding Level:

Funding Source:

13-0292

July 1978 - June 1979

$86,000

Sandia Laboratories, Albuquerque

155

PROGRESS IN COMPOSITE FLYWHEEL DEVELOPMENT

P. W. HillT. C. WhiteD. G. DrewryW. B. Stewart

Hercules Aerospace DivisionHercules Incorporated

Allegany Ballistics LaboratoryP. 0. Box 210

Cumberland, Maryland 21502

ABSTRACT

Development of filament reinforced polymer (FRP) flywheels for energy storage invehicles has been continuing at ABL with emphasis on material selection and design syn-thesis based on the filament winding process. Three species of designs have been studied:

1. Helically wound shells;2. Circumferentially wound rims with radial spokes;3. Circumferentially wound contoured disks.

The objective has been to evaluate the potential of design concepts with commercialmerit. A new approach to the type 3 flywheel offers competitive performance in a wheelconfiguration that is moire compact than other types and is also attractive from a massproduction viewpoint. Combining optimum contour with programmed winding tension in aprocess that fixes the winding tension as the wheel is fabricated results in rated per-formance ar~rnr...'u.La& 30 W-hr/lb and 1 W-hr/cu in. for a rotor storing 1270 W-hr at thedesign operating speed. Design and fabrication are in progress under contract from DOE/Sandia. A description of the design concept, the basic theory of the design, and asummary of material properties are provided.

INTRODUCTION

Circumferentially wound composite diskflywheels offer the advantage of a low costmanufacturing process. However, in manyinstances their performance is unsatisfac-tory as a result of radial tensile stresseswhich exceed material strengths. Somemechanisms to reduce the radial stressesinclude:

1. Increasing radial compliance ofthe material.

2. Introducing bands or rings ofanother more compliant material.

3. Using concentric but separaterings with mechanical coupling.

A. Using concentric bands of differentfibers which provide a modulus gradientincreasing toward the rim and/or a massgradient decreasing toward the rim.

5. Contouring the faces of the disk.

6. Applying a beneficial (compres-sive) prestress field by press fit orshrink fit assembly of individual ringsor rings and hub (loses cost advantage).

7. Applying a beneficial prestressby successive winding and curing of concen-tric rings (loses cost advantage).

8. Applying a beneficial prestressby modifying the winding tension programand cure process during fabrication.

All of these have been studied and reportedwith differing degrees of optimism, butwith little experimental success, unlessthe depth (Ro-Ri) or radius ratio (Ri/Ro)was limited so that the "disk" was morenearly a "rim."

Hercules has studied combinations ofthe above mechanisms to identify benefitsthat might be gained thereby and whichmight offer acceptable performance. Em-phasis was placed on methods 4 through 8.Results have indicated that combining

156 -

methods 5 and 8 is most beneficial and canachieve useful performance without furthercomplication. The following discussionwill outline the basis for this observation.

BACKGROUND

Toland1 and Hill2 have described theuse of a transfer matrix approach forcomputing the stresses in a rotatingorthotropic disk for the purpose of find-ing the optimum disk contour. An extensionof that approach was prepared to calculatethe stress distribution in a stationarydisk resulting from an arbitrary programof winding tension. The two solutions aresuperimposed in a computer program thatseeks the optimum contour for the prescribedwinding tension program.

Reuter3 has shown that the residualstresses associated with conventional fila-ment winding and cure processes are detri-mental. Fig. 1 illustrates the effect foran example case with uniform winding ten-sion. These residual stresses can causefailures by and of themselves. The maximumpermissible AT of 163°F in the example ofFig. 1 is inadequate to cover the servicetemperature range (+200° to -50°F) plusadequate cure temperature increment. Re-duced cure temperatures and winding ten-sions which increase toward the outer radiusare beneficial, but are generally unaccept-able because:

1. Low temperature cures are incompat-ible with elevated service temperatures(ca. 200 F).

2. Winding tension programs whichincrease outward cause wrinkling of thefibers near the inner boundary.

Some form of unconventional processing isneeded.

Staged cure of sequential, concentricwindings is one solution. However, thecost penalty is severe, and the repeatedthermal cycles make a difficult problemto analyze, since some "annealing", orloss of pretension can be expected witheach heating and softening of the insiderings.

Laakso1* has described a process thatsupplied the desired qualities; that is, aquick fixing of the fiber to lock in theapplied tension and prevent wrinkling ofthe inside fiber, while at the same timeeliminating the need for a general elevated

temperature cure cycle. This latter prop-erty permits all of the beneficial (compres-sive) radial prestress to be applied tocounteract the spin stresses, thereby im-proving energy storage rather than merelycombatting manufacturing residual thermalstresses.

RADIAL STRESS. °r

15-j J / X / 60

a 10 II \ / 40V HOOP STRESS. *£ \ /

20 u.

- 20

-40

-60

-80

Fig. 1. Thermal stresses due tocooldown from cure.

The concept is based on the use ofcarbon fiber preimpregnated with a poly-sulfone thermoplastic matrix. The prop-erties of this material are compared withcarbon/epoxy and aramid/epoxy compositesin Tables 1-4. Note that the propertiesare equivalent to conventional epoxy(thermosetting) matrix composites at tem-peratures to 350 F. The material has beenthoroughly evaluated and qualified foraerospace application under Air Forcesponsorship, and properties are reportedin Reference 5. In addition, the materialexhibits very low creep and improvedtoughness when compared to epoxy matrixcomposites. Having found a material andprocess that appear to satisfy the require-ments for a prestressed contoured diskflywheel, the next step is to examine thepotential performance.

An extensive matrix of designs meet-ing the requirements for the Sandia vehicleflywheel was studied to identify optimum

FLA1Ro =

Ri =E<TEr =v&ra9

• DISK11.8 IN2.0 IN16.5 x 106 psi1.2 x 106 psi= 0.34= -.01 x 10~6 / °F= 15 x 10~6 / °F

157

contours, stress distributions, andgeometrical correlations. The first stepwas to establish the residual stressdistribution resulting from "freezing"the winding tension as a function of(Ri/Ro) and the tension program. (Note:In all cases the O.D. was fixed at 23.6inches.)

Table 1. Properties of candidatefibers and resins.

Table 4. Design allowables.(150°F, 104 cycles, 103 hours)

Fibers:

Property

Tensile Modulus <106 psl)Tensile Strength (ksi)Elongation (?)Density Ub/in.3)Specific Modulus (108 In.)Specific Strength (106 in.)

Resins:

Property

Initial Tensile Modulus(10° psi)

Tensile Strength @ Yield (psi)Ultimate Strength (psl)Elongation (%)Density (lb/ln.3)

Type AS-6 Carbon Kevlar-49

34.0440.01.4

0.0654.9

6.9

Polysulfone

0.36

10,20010,200SO to 1000.0448

19.0SOO.O

2.6

0.052

3.6

9.6

Epon 828/Jeffamlne T403

Epoxy

0.45

8,00013,5007.0

0.0441

Table 2. Nominal room temperaturecomposite material property data.

7i) enType AS/3l)U<. Type AS/Epoxy Kevlar-49/Epoxy

Property (Vf - 57iQ (VF - 603? (VF • 60S)

Longitudinal Tensile Hodu- 16.5 18.0 10.5lus. En HO6 psl)

Transverso Tensile Modulus, 1.2 1.0 0.50

E22 (106 pal)

Major Poisson's Ratio. Vl7 O.34 0.2? 0.34

In-Plane Shear Modulus, 0.56 0.B5 0.30Gi2 (106 psl)Longitudinal Tensile 190.0 210.0 225.0Strencth, F'n

r <kai>

Longitudinal Cotnpresslue 102.0 160.0 45.0Strenpth, F u

c (ksl>

Transverse Tensile Strength, 5.5 6.0 2.5F;2

T Cfcsl)

Transverse Coopressive 18.9 18.0 14.0Strength, F2 ; c (lcsl>

ln-Pl.me Shear Strenirth, 16.0 12.0 3.0F12 Cksl)

Short Beam Shear Strength 11.6 B.O 5.0(test)

CoeffUienr of ThermalExpirsion

U-nf-itudinal (10"6 -0.006 0.00 -^.Gin./in./°F)

Transverse (10"6 27.0 15.0 20.0in./in./°F)

(l)Typical for an epoxy reain such as Epon 828/JcfFaalne T-403.

p

LongIcud!n

LongitudlnaStrength, i

Transverse

Transverse^Strength, F

In-Plane Sh

F2 3 <ksi)

opertv

1 Tensile Str

1 ConpresslveHc (Wai)

Tensile Stren

Compressive

ear Strength,

ngth.

t h ,

Type AS/Epon-82B/T403

155.0

94.0

2.7

11.5

3.5

Type AS/3004

142.0

60.0

Z.9

12.0

7.3

5.3

Kevlar-49/

101.0

26.3

0.50

8.9

1.0

Fig. 2 illustrates the result for anexample case. Note that the winding ten-sion was limited by the compressive trans-verse (design) strength of the composite,indicating the maximum benefit of thiseffect was achieved. This benefit hasthe effect of increasing the allowableradial stress due to spinning by a factorof five. The effect of the hoop stressdue to winding is slightly detrimental inthe case illustrated, but is slightlybeneficial for thin rim type rotors.

TRANSVERSETENSILE

STRENGTH LIMIT

"?_ -0.5b

-3.0

-4.0

9 pE, • 1.2*10*p«a,' -2M0pi iOg». 142.000 MiU - IMCnd / lK

TRANSVERSECOMPRESSIVE

STRENGTHLIMIT-

Fig. 2. Stresses in an orthotropicdisk flywheel.

158

Table 3. Material degradation to design environment.

Material

Type AS/Epon 826/T403TensionCompressionShear

Kevlar-49/Epon 826/T403TensionCompressionShear

Type AS/3004TensionCompressionShear

Material Variation-3X) (%)

Lone.

15

35<1)

,5(6)

(l)Hercules Data for Carbon &

(2)Based on Data from

(3)Estimated.

(4)Based on Data from

(5)Based on Data from

(6)Based on Data from

Hofer,

Chiao,

DuPont,

Trans.

25* '

20 ( i )

(1)25* '

30 < 6 )3°(6)

Shear

30 ( 1 )

25(D

30(6)

150°F TemperatureDegradation

(% of Ultimate)

Long.

io">

io»>

0(6)

Trans.

o(6'

Shear

15(1)

2O(3)

10 ( 3 )

Kevlar-49 Composites.

K. E., et al., Reference 6.

T. T., et al., Reference 7.

Reference 8.

Reference 5.

Degradation Dueto Static Fatigue(% of Ultimate)

Long.

3<2>

30 ( 4 )

3(6)

Trans.

5(2)

35(3)

5(6)

Shear

5(3)

3O<3)

5(3)

Degradation Dueto Dynamic Fatigue(% of Ultimate)

Long.

10")

15(5)

10(6)

Trans.

2O ( 2 )

20 ( 3 )

20 ( 6 )

Shear

20<3)

20 ( 3 )

20(3>

In. Fig. 3, the effect of the combinedprestress and spin stress on the failurecondition throughout the disk is shown.Although the critical condition is stillthe radial stress, the hoop stress is nowelevated to 65% of the allowable fiberstrength. Further improvement of the flatdisk, is possible by adjusting the windingtension program so that the peak negativeprestress coincides with the peak positivespin stress and by better definition ofthe maximum prestress allowable.

To balance the design so that radial(transverse) and hoop (fiber) failure condi-tions occur simultaneously, the sides ofthe disk are contoured. Fig. 4 illustratesthe relative benefits from prestress andcontouring for carbon fiber/polysulfonematrix composite flywheels as computed bythe above methods of contour optimization.Substantial deviations from these plotsmight be expected for other materials andwinding tension programs. However, thesignificant trends are valid:

1. Prestress has useful value forrelatively thin rims (a between 0.5 and 0.8).

2. The effect of contouring increasesfor disks with small holes (a below 0.4)

and admits simplification of hub designand high volumetric efficiency as well asweight efficiency.

r- SAFE OESIGN LIMIT

«/»„

Fig. 3. Failure condition in spinning diskafter prestress by winding tension.

159

CONTOOtEfWltHltKTKS!

Table 6. Expected performance of thepreliminary design.

OJ 0.4 O.« O.» 1.0

RAOftJS tATIO. • • • *

Fig. 4. Effect of contour and prestresson the performance of orthotroplc diskflywheels.

PROJECT DESCRIPTION

Hercules Aerospace Division/AlleganyBallistics Laboratory Is under contract toDOA/Sandia to design and fabricate two testwheels based on the above approach. Energydensity targets were established as shownIn Table 5. The expected performance ofthe preliminary design Is given In Table 6.The final design will be supported by com-plete quasi-static stress analysis andnormal modes dynamic analysis. Thedynamics analysis and planning for all spintesting will be performed by the RpcketdyneDivision of Rockwell International undersubcontract. The project is just underway and in the design phase.

Table 5. Practical energy density targetsfor composite flywheels in vehicles.

Cbaraetarletlc Prosertv

Teoalle strength, P UT (HI)

Denalty (lb/ls.3)O.S ffu^g), (tfc/lb)1

birlroraental Xnclidoira Factor, ( I ) 2

Oltlaete Energy Denalty, Holt OHi/lb)General Safety FactorIdeal Operating Energy Denaity, no(Ub/lb)Faraaitlc Valgbt AUo>ance (I)Deelgn Inergy Deaelty Target, twWWlb) B

Kevler-49/fnoTt

223

0.05070.6SO

35.31.2528.3

13

24.0

MaterialAS Carbon/

tooicr

210

0.05463.926

47.31.2S37.>

15

32.1

AS Carbon/PolTaulfona

190

0.05455.225

41.41.2533.1

*3

2S.2

Bunt 5pccd (Aftar Service Life Knockdown)

Dealgn Operating Speed (DOS)

Stored Energy at DOS

Delivered Energy at DOS

Weight, Dl»k

Weight, Hub

Weight, Total

Stored Energy Denalty

Maxima Hoop Streit at DOS

Matimm Mdlal Streaa at DOS

31,560

28,230

1,270

1,131

41.3

4.0

45.3

28.0

113,600

2,320

rp«

rp»

Uh

Wh

al b

lb

Wh/lb

pal

pal

Ideal ahape factor for coapoalta Mtarlala - 0.5. Sae" - - " " " I . 1»7S).taferance 9 (CeratU • Ugga

2. lee Tabu 3.

REFERENCES

Poland, R. H., and Alper, J., "TransferMatrix for Analysis of Composite Fly-wheels," Journal of Composite Materials,Vol. 10, July 1976, page 258.

2Hill, P. W., et al., "Advanced FlywheelDevelopment," 1978 Flywheel TechnologyConference, San Francisco, CA, October1978.

3Reuter, R. C., Jr., "Fabrication and Ther-mal Stress in Composite Flywheels," Pro-ceedings of the 1975 Flywheel TechnologySymposium, Lawrence Ball of Science,Berkeley, CA, November 10-12, 1975.

'•Laakso, J. H., "Potential Merits ofThermoplastic Composite Materials forModular Rim Flywheels," Proceedings ofthe 1975 Flywheel Technology Symposium,Lawrence Hall of Science, Berkeley, CA,November 10-12, 1975, page 164.

5Anon., "Mechanical Property Data for AS/3004 Graphite/Polysulfone Composite,"Issued by AFML and Prepared by Universityof Dayton Research Institute under ContractF33615-75-C-5085, November 1976.

6Hoefer, K. E., et al., "Development ofEngineering Data on the MechanicalProperties of Advanced Composite Mate-rials," IIT Research Institute, TechnicalReport AFML-RT-74-266, February 1975.

7Chlao, T. T., et al., "Lifetimes of FiberComposites Under Sustained Tensile Load-ing," Lawrence Livermore Laboratory,University of California, Contract No.W-7405-Eng.-48, 1976.

8Anon., "Kevlar-49 Data Manual".

160

9Gerstle, F. P., and Biggs, F., "On OptimalShapes for Anisotropic Rotating Disks."Proceedings 12th Annual Meeting, Societyof Engineering Science, Inc., Universityof Texas, Austin, TX, October 20-22, 1975,see alsoGerstle, F. P., and Biggs, F., "On Effec-tive Use of Filamentary Composites in Fly-wheels," Proceedings of the 1975 FlywheelTechnology Symposium, Lawrence Hall ofScience, Berkeley, CA, November 10-12,1975, page 146.

161

PROJECT SUMMARY

Project T i t le :

Principal Investigator: Dr. D. E. Davis

Prototype Development - Composite Flywheel Having Nominally40 Watt-hrs/lb Energy Density

Organization: Rocketdyne DivisionRockwell International6633 Canoga AvenueCanoga Park, CA 91304(213) 884-3075

Project Goals: The development of hardware that will demonstrate the very latesttechnology relating to composite wheels. This project is oneof four that explore different approaches to producing a wheelhaving the highest practical energy density.

Project Status: The concept here is a radially-overwrapped circutoferentially-wound wheel. This design is a follow-on to a wheel that had beendeveloped for attitude control and energy storage in spacecraft.The earlier design utilized an elastomeric buffer at the interfacebetween the windings; this has been identified as a source ofmechanical problems and will be omitted in future designs.

At present, the first iteration of stress calculations has beencompleted. Materials characterization tests are now beingperformed. When completed, these and the stress computationswill be incorporated into a detailed design.

Contract Number: 07-6955

Contract Period: July 1978 - May 1979

Funding Level: $149,562

Funding Source: Sandia Laboratories, Albuquerque

163

ADVANCED COMPOSITE FLYWHEEL FOR VEHICLE APPLICATION

Dr. D. E. DavisRocketdyne Division

Rockwell International6633 Canoga Ave.

Canoga Park, California 91304

ABSTRACT

Rockwell's high energy density composite wheel concept is based upon work done aspart of a program in spacecraft power and attitude control. Experience and analyticcapabilities which evolved during the early program will be utilized in the course ofproducing next-generation hardware. This paper will describe the design and presenttypical results of computer analyses treating operation and fabrication stresses. Theadvantages of the concept (viz., great axial stiffness and high volume efficiency) willbe discussed. Earlier design problems will be briefly touched upon.

INTRODUCTION

Energy wheel development is a highpriority Rockwell International techno-logy effort consisting of both contractwork and related company research.*>2»3 Arecently completed research program, par-ticularly applicable to the DOE compositeflywheel programs, is the Integrated Powerand Attitude Control System (IPACS). TheIPACS program resulted in the design, fab-rication and test of a Rockwell proprie-tary Kevlar/epoxy tape-wound compositeflywheel. The IPACS data is being uti-lized in the current Sandia LaboratoriesContract No. 07-6955 issued under DOEPrime Contract No. AF(29-l)-789.

BACKGROUND

ROCKWELL RESEARCH PROGRAM

Rockwell International, in previouscontract and research efforts, has comple-ted composite materials and wheel confi-guration evaluation studies which will beutilized in the existing flywheel pro-gram. More importantly, these studiesenable the effort to be focused on arelatively narrow set of alternativematerials and a specific wheel concept.Rather than studying alternative wheelconfigurations, the effort can bedirected toward the evaluation of designrefinements that will optimize theselected design concept.

The original Rockwell-evolved conceptis shown in the sketch of Fig. 1. This

design utilized a metallic shaft uponwhich a composite core was circumferen-tially wound. An elastomer matrix wasused for the core. A transition liner ofelastomer was used between the core andthe radial overwrap of Kevlar/epoxy. Theimportant features of this design conceptare:

• High energy density,

• Good volumetric efficiency,

• Uniformly distributed loads,

• Rugged configuration,

• Conventional mfg. technology.

KEVLAR/EPOXYOVER WRAP

RUBBER LAYER

9 IN.RADIUS

CIRCUMFERENTIALLYWOUNDE GLASS CORE

TITANIUMSHAFT

( ROTATION

Fig. 1. Original Rockwell researchcomposite energy wheel

t>

Rockwell carried out a research pro-gram in 1975 to fabricate and test a modelof the rotor concept shown in Fig. 1. Therotor was designed and fabricated and isshown in Fig. 2. Testing was completedutilizing the test setup shown in Fig. 3.

Fig. 2. Rockwell research rotor

Fig. 3. Rockwell research rotor beinginstalled in spin testfacility

This research program contributed signi-ficantly to the present composite wheeldevelopment project. Fabrication methodswere demonstrated, including evaluation

of tooling concepts, winding, wrapping,curing techniques, and machining pro-cesses. Application of the elastomericinterface material was demonstrated onthis research program. Following fabri-cation, the model wheel was tested toevaluate actual performance in relationto design predictions. A test to des-truction provided significant insightinto failure mechanisms and the variousforms of energy dissipation (Fig. 4 ) .This research activity was completedduring 1975 at the Rockwell InternationalSpace Division Facility.

Fig. 4. Rockwell research rotor -posttest

The Kevlar/epoxy tape-wound researchcomposite test wheel was designed to havea storage capacity of from 30 to 32 watt-hours per pound. The first test wheelmeasured approximately 18 inches in dia-meter and had a nominal design operatingspeed of 35,000 rpm. Burst speed wascalculated to be approximately 44,000rpm. Studies indicated that wrappedcores made of S-glass/elastomer andKevlar/elastomer would give higher speci-fic energies (15 to 20 percent and 35 to40 percent, respectively) than the E-glass/elastomer core. The workingallowables for the composite materialused in the study were taken at approxi-mately 65 percent of the ultimatestrengths. Significant test data con-cerning rotordynamics, resonant frequen-cies and failure modes and effects ana-lysis were obtained although premature

165

failure of the first test rotor occurredearly in the test program.

'IPACS RESEARCH COMPOSITE ROTOR

Under a NASA Contract NAS1-13008, afive phase Integrated Power and AttitudeControl System (IPACS) flywheel programwas completed. Under Phases IV and V ofthe program (Phase V having been com-pleted in February 1978) composite testflywheels were designed, fabricated andtested.

A cross-section drawing of the IPACSrotating assembly Is shown in Fig* 5,The rotor was fabricated from compositesbuilt upon a titanium hub. The rotorcore was Kevlar in an elastomer matrixand the core was overwrapped with Kevlarin an epoxy matrix.

PBELISCREW

PRELOADHOUSING' PRELOAD

PISTONPRELOAD SPRING

VACUUM ENCLOSURE

COMPOSITE ROTOR

MO STATOR HOUSING

M G 5TATOR

END COVER

FIXED BEARINGHOUSING

.TACH PICKUP

SENSOR

HOUSING

OILER

SENSOR

ANGULAR

CONTACT BALL BEARING

ILSLINGER

WASTE OIL RETAINER

MOTOR GENERATOR ROTOR

Fig. 5. IPACS composite rotatingassembly cross-section

Two identical motor-generator (M-G)units were designed to be incorporated atboth ends of the rotor and together toprovide 2500-watt motor or generatorcapacity. Losses were minimized undernormal operating conditions with magneticefficiency projections greater than 96percent.

Fabrication Techniques. The IPACS compo-site flywheel consisted of: (1) an iso-tropic shaft/hub fabricated from 6A1-4V

titanium alloy, (2) a Kevlar/nitrile corewrapped circumferentially around thetitanium hub, (3) a nitrile rubber inter-face liner, and (4) an overwrap appliedradially over the interface liner andcore.

The Kevlar/nitrile core consisted ofa series of 6 rings circumferentiallyapplied over the shaft hub. For applica-tion of the circumferential plies, a spe-cial tensioning device and special sidetooling plates were used. The tensioningdevice allowed incremental tape tensionreduction as the winding of each ringproceeded. Consolidation of the core foreach ring was accomplished, using six in-dividual, concentric stepped rings, byapplying lateral pressure to both annularside plates. As the material was cured,the inward movement of the side platesapplied the required hydraulic pressureto the prepreg tape to give the desiredwrinkle-free and minimum void structure.In addition, the pressure was controlledto give a 58 percent to 60 percent fibervolume composite. •

Machining of the core to the designcontour required special tooling and pro-cedures as determined from previous com-posite wheel fabrication experience. Thekey to the machining operation is tochill the wheel to a low enough tempera-ture for efficient cutting and yet not toa point where the nitrile matrix wouldcraze.

The final stage in the wheel fabri-cation (after the rubber liner was ap-plied) was the application of the over-wrap. This task was subcontracted to theBrunswick Corporation and consisted ofradially wrapping the core/liner assemblywith a Kevlar/epoxy composite system.The outerwrap was applied in a sevenpointed star pattern using a numericallycontrolled winding machine.

The IPACS wheel was machine wrappedwith 9 layers of 0.0203 cm (0.008 inch)roving (Kevlar 49) impregnated with aBrunswick formulation of DER 332 matrixmaterial. A tension of 13.3N (3 lbs) perroving was applied and held constantduring the winding operations.

Figure 6 is an edge view of thewheel showing the tapered sides and the

166

shaft stubs with balance discs and thebearings mounted on the journals. Figure7 shows a side view of the wheel to illus-trate the radial wrap pattern.

Fig. 6. Complete IPACS test wheel -edge view. Rotor weight is59.9 pounds

Fig. 7. Complete IPACS test wheel -side view

Test Program. A series of tests were runon the IPACS composite wheel. Thesetests were performed at the Rocketdynespin test facility using a speciallydesigned dual bearing support housingfixture as shown in Fig. 8.

Fig. 8. IPACS test wheel installed inspin pit using dual bearingsupport housing

This housing which was attached tothe bottom side of the spin pit lidallowed versatility in the test setupand supported the wheel in case the drivequill shaft failed. The wheel could be(1) suspended freely on the turbine quillwith nylon burnout blocks at the upperand lower shaft journals to help containthe wheel in case of failure, or (2) thewheel could be supported in the rig withbearings on the upper and lower shaftjournals. In either case, the turbinequill could be attached to the wheeldirectly or through a flexible coupling.

A series of 36 tests were completedinvestigating dynamic stability, thermaland radial growth, wheel balance shiftsand axial stiffness. Although the designgoal was 61.7 watt hours/Kgm (28 watthours/pound) the design goal was neverreached due to dynamic instability.

Following the test program, thewheel was inspected thoroughly fromvisual observations and through the useof NDT techniques. No crazing, delamina-tion, or other failures were noted onclose visual examination of the wheel.NDT inspections including X-ray, ultra-sonic and neutron radiography, revealedno gross defects in the wheel such asvoids, core ply separations, fiber buck-ling or fractures. There were some smallareas shown in the X-rays which could notbe explained. These areas appear to be

167

located at the bond surfaces between thecore, rubber interlayer and overwrap.Development work is required to adapt theX-ray and ultrasonic techniques to thicksection structures such as the compositewheel.

Posttest Analysis. The primary problemwith the earlier IPACS composite rotordesign is dynamic out-of-balance. Reviewof the stress analysis, detail fabricationprocedures, and test results has led tothe conclusion that imbalance resultedfrom eccentric shifts in core materialthat varied with rotational speed. It isknown that minor amounts of air weretrapped on the core during wrapping andwere undoubtedly retained after cure eventhough the core was evacuated and exter-nally pressurized under heat during cure.Further, repeated cure cycles on innerwraps would tend to ca .se variation inmaterial properties.

An even more significant problemwith the core was the compression ofinner wraps by the buildup of outer mate-rial which was tensioned during wrapping.This effect, in conjunction with externalpressure during cure, produced wrinklesand local buckling of inner wraps. Thetest program also indicated that permanentshifts in balance occurred. These shiftscould be attributed to creep which wasnoted in the structure. Another reasoncould be circumferential delamination ofcore material from defects, insufficientwetting, overstress, etc., which wouldthus allow local expansion and produceimbalance. Upon reduction of speed theselocal cracks would close, preventing de-tection by NDT methods.

IMPROVED ROTOR SYSTEM

Utilizing the technology that hasbeen obtained through Rockwell's previousradial wrap rotor system programs, seve-ral improved stress analysis, design andfabrication approaches are being imple-mented for the present DOE flywheel pro-gram.

ANALYTICAL ANALYSIS

The finite element method is used todetermine the displacement, stresses, andstrains in axisymmetric and planar solids.The programs available for this methodare:

• APSA - Finite element axisym-metric and planar struc-tural analysis with or-thotropic temperature-dependent material pro-perties;

• APSAC - Finite element axisym-metric and planar struc-tural analysis withloading and creep dutycycles;

• APSAPLOT - Plotting program forAPSA and APSAC;

• APSADUMP - Data recovery programfor APSA and APSAC.

The IPACS analysis utilized the APSAand APSAPLOT programs.

These programs allow for orthotropic,temperature-dependent material propertiesunder thermal and mechanical loads. Themechanical loads can be surface pressures,surface shears, and nodal point forces aswell as an axial acceleration and angularvelocity. The continuous solid is re-placed by a system of ring or planar ele-ments with quadrilateral cross-section.Accordingly, the method is valid forsolids which are composed of many differ-ent materials and which have complex geo-metry.

The APSA program analyzes elastic orelastic-plastic problems with a single setof loads. The APSAC program analyzes pro-blems where plasticity and creep must beconsidered during a duty cycle representedby a series of load increments. Eitherisotropic or kinematic strain hardeningcan be specified. The APSAPLOT program isused to furnish a visual display of thefinite element results. Contour plots,grid plots, and graphical plots ofstresses and strains can be obtained. TheAPSADUMP program is used to recover infor-mation stored on the data tape generatedby the APSA or APSAC program.

As an example of the APSA program,the undeformed mathematical model of thewheel used in the APSA program is shown inFig. 9, and the deformed model is shown inFig. 10. An axisymmetric rotating elasticbody is assumed. The deformed model (Fig.10) shows a large displacement of theradial rubber liner between the outerradial wrap and the hub. The cause of thelarge displacement is assumed to be the

168

program analysis, which did not representthe wrap accurately in this area, but suc-cessive iterations can accurately repre-sent this area.

32 p

s 2 0

1 1 6

ITOf 12

8

10.000

z

- oc

5.000

4

oL0 5.0000

Z. DIMENSION (INI

4 0 4 8Z, DIMENSION (CM)

12

Fig. 9. Composite energy wheel -undeformed structure

10.000 -

zre

5.000 -

TYPICAL ELEMENTSFOR CORE SHEARSTRESSES AND STRAINS

SPEED = 32,500 RPMMAXIMUM RADIAL GROWTH =

0.396 CM( 0 . 1 5 6 IN.)

I0 5.0000

Z, DIMENSION UN.)1 I I I

4 0 4 8

Z. DIMENSION (CM)

12

Fig. 10. Composite energy wheel -deformed structure

MODIFIED DESIGN

An improved design wheel will bebased on many factors of materials, pre-dictable stress versus allowable stress,and fabrication processing. The APSAstress analysis program is being updated

and detail wheel dimensions are beingtraded off with properties to optimizeenergy density while providing adequatestress/strain margins.

In the outerwrap, the basic convexcontour and design is being retained forease of fabrication and structural loading.An improved matrix is being used havingmore elongation. This matrix materialwill result from the materials and pro-cesses studies now being carried out inorder to identify and define propertiesfor stress analysis and factors in fabri-cation.

The elastomer liner between the metalhub and outerwrap which serves to distri-bute loads between the harder compositematrix and metal will be retained. How-ever, the rubber liner ovar the remainderof the wheel will probably be deleted toprovide better strain compatibilitybetween the core and outerwrap.

The circumferentiall}' wrapped coredesign will be retained in concept, butotherwise will be substantially changed toprovide a stiffer structure, free fromcreep and with an improved strain matchbetween core and outerwrap. For increasedstiffness, the core matrix will be changed;polyurethane and blended epoxies will beconsidered. The wrap from inner to outerlayers will be modified to vary modulusand core density to provide improvedstrain compatibility and increased energydensity by increasing stresses of theinner core.

MODIFIED FABRICATION TECHNIQUES

Major fabrication technique improve-ments are possible in the core whichshould substantially improve the perfor-mance of the radial wrap wheel. Of para-mount importance, the circumferentiallywrapped fibers will be maintained straightand concentric during layup by two possi-ble means: (1) computer control of wind-ing to automatically maintain proper *-<:n-sion and concentricity and (2) singlefilament wrapping with wet layup. Layerswill be cured during the wind process orin intervals of 0.63 cm (0.25 inch) orless to preclude compressing inner fibers.These techniques will hopefully avoid en-trapping significant bubbles in the matrixand prevent buckling. At prescribed in-tervals, the core will be machined asrequired and balanced. This will be ac-complished by wrapping variable density

169

tape around the circumference to obtainbalance and not change concentricity.

The outerwrap may be applied instages to allow dynamic balancing of thewheel by applying patches as above withentrapment by subsequent layers of fibers.In either case, it will be the purpose ofthis procedure to arrive at a finishedfabricated wheel that is substantiallyin dynamic balance.

REFERENCES

1. SD77-AP-OO38, Fabrication and Test ofa Composite Rotor for Integration Withan IPACS Rotating Assembly, FinalReport, Rockwell International SpaceDivision, July 1977.

2. SD76-SA-OO14, Design Report for theFabrication of a Composite Material .Rotor and Integration With an IPACSRotating Assembly, Rockwell Inter-national Space Division, September1976.

3. Davis, D. E., C. Heise and D. Hodson,Rocketdyne's High-Energy-StorageFlywheel Module for the U. S. Army,Rockwell International RocketdyneDivision, October 1977

170

PROJECT SUMMARY

Project Title: Prototype Development - Composite Flywheel Having Nominally40 Watt-hrs/lb Energy Density

Principal Investigator: David L. Satchwell

Organization: Garrett-AiResearch2525 W. 190th StreetTorrance, CA 90509(213) 323-9500

Project Goals: The development of hardware that will demonstrate the very latesttechnology relating to composite wheels. This project is one offour that explore different approaches to producing a wheelhaving the highest practical energy density.

Project Status: This is a rim-and-spoke design based on a wheel produced for the ArmyMobility Equipment Research and Development Command ("MERADCOM").That wheel, which incorporated a brute force hub and spoke con-figuration intended only as an expedient for rim testing, hasalready been successfully spun to an energy density of 26 watt-hrs/lb. The new version will use a similar rim, but incorporatesa sophisticated carbon-fiber hub and spoke design.

At present, fatigue and ultimate strength tests have been performedupon samples of the new design spokes. Hub fabrication has begun.

Contract Number: 13-0293

Contract Period: July 1978 - June 1979.

Funding Level: $267,759

Funding Source: Sandia Laboratories, Albuquerque

171

HIGH-ENERGY-DENSITY FLYWHEEL

David L. SatchwellGarrett-AiResearch Manufacturing Company of California

2525 W. 190th St., Torrance, California 90509

ABSTRACT

This paper describes the design and fabrication of a flywheel rotor with an energydensity of 80 w-hr/kg. the design features a multiring, S-glass and Kevlar compositematerial rim that is mounted on a graphite composite spoked hub. Composite materialflywheels have been constructed with identical rim design and with aluminum spoked hubs.These flywheels have been successfully tested to energy densities of 53 w-hr/kg. Theuse of graphite composite material in place of the aluminum material significantlyreduces the weight of the rotor assembly and thereby increases the energy density. Thispaper will present the progress to date on this program and describe the program plannedto complete the fabrication of a test specimen.

INTRODUCTION

A flywheel is being developed byGarrett-AiResearch to demonstrate a highenergy density of 80 w-hr/kg in the rotor,which represents a significant increase inflywheel energy density. Because rotorswith high energy density are smaller andlighter, this 80 w-hr/kg rotor promisesto substantially reduce cost and weight ofthe installed unit.

The rotor design consists of a multi-ring, subcircular, S-glass and Kevlarcomposite material rim that is mounted ona star hub of graphite composite material(see Fig. 1). The lightweight Kevlar com-posite is very strong for its weight andtherefore has a high energy storage capa-city.

Because of its multiring design, therim can accommodate the radial stressescaused by centrifugal forces. Also, therim was given a subcircular shape so thatwhen placed on an oversized star hub, theresulting rim forces hold it onto the hub.This eliminates the stress associated withattaching the rim to the hub by ordinaryattachment methods.

The graphite/epoxy composite materialhub has a small cross section and is veryliqht, yet can withstand the forces imposedon it. The combination of high-energy-density storage in the rim and a verylight, strong hub yields an exceptionallyhigh-energy-density flywheel rotor.

INNER S-2 GLASS EPOXY COMPOSITEKEVLAR 29 EPOXY COMPOSITEKEVLAR kS EPOXY COMPOSITE

RINGS

-GRAPHITE COMPOSITECRUCIFORM HUt

SUBCIRCULARRIM

S-3M50

Fig. ?. Flywheel rotor

SCOPE OF WORK

Two rotors will be designed and fabri-cated for evaluation—an effort that com-prises six principal tasks.

Task 1, Materials Selection and HubFabrication Process Proof. A new rim epoxyis to be selected because the epoxy used onthe previous AiResearch flywheels is nolonger commercially available. A replace-ment epoxy has been recommended by T.T.Chiao of Lawrence Livermore Laboratories

on the basis of composite material andneat epoxy sample tests. (The epoxy isDow DER 332, the diluent is Ciba GlegyRD-2, and the hardener Is UnlRoyal Tonox60-40.J The contract provides for thefabrication and test of flywheel ringsmade from a composite of S-2 fiberglass,Kevlar, and the new epoxy. These testswill be run to confirm the acceptabilityof the new epoxy. The test ring Is shownin Fig. 2.

I—T = APPROXIMATELY0.080 IN.

0.938 IN.S30448

Fig. 2. Test ring

The hub, to be made of a graphite-epoxy composite, requires proof of fabri-cation process and material propertiesacceptability. The hub Is being fabri-cated by Structural Composites IndustriesInc. (SCI) of Azusa, California. Graphitecomposite is chosen for this applicationbecause of its modulus of elasticity of18 x 106 psi In composite, because of itslight weight of 0.05 Ib/cu in., and itsultimate flexural strength of .210,000 psi.The combination of these properties allowsthe reduction of hub weight from 20.5 Ibfor an aluminum hub to 6 Ib for the graph-ite epoxy composite hub. Thus, thealuminum-hub rotor yields 53 w-hr/kg Inthe rotor and the graphite-hub rotor yields80 w-hr/kg in the rotor. Task 1 willInclude mechanical testing of hub samples.

Hub testing will be conducted onthree hub sections of 1-ln. axial lengthand will consist of compression testingacross one opposite set of spokes to learnthe location, mode, and load at failure.One hub will be cycled 100 cycles underload to a stress that wilI give an accel-erated, full-life test. The third hub

will be utiIIzed as needed to confirmthese tests.

Task 2. Design Analysis. Design analysisIncludes an evaluation of the stressesInduced In the flywheel rotor during fab-rication and operation, with an evaluationof fatigue effects and scaling lawsInvolved In changing the size of the fly-wheel. The analysis treats the following:

• Stress in the wheel caused bywinding, curing, thermal effectsand assembly.

• Stress during operation, bothsteady-state and transient.

• Comparison of trlaxial materialproperties to the stress state,with particular emphasis uponsource and reliability of mater-ial properties data.

• Special stress conditions suchas those at rim-hub interface.

• Critical speeds and the effectsof possible material creep uponoperation and balance.

Stresses caused by fabrication andassembly can be accurately defined, and theresults of such an analysis will be usedfor final assembly design refinements.The calculation of the triaxial stressstate within the hub will be accomplishedas a guide to the tests of Task 1. Theresults of Task 1 wilI be used to refinecalculations for the full axial length ofthe hub.

Task 3, Hub and Rim Fabrication. Twoassembled rotors, shown in Fig. 3, will becomp I eted.

Flywheel rotor assembly

173

The hub will be fabricated by SCI,using the methods developed and provedduring Task 1. The hub consists of multi-ple slats, alternately bonded one uponanother to form a four-spoke hub with arigid cruciform shape. (See Fig. 4 ) . Thetwo-module rim will be similar to the onequalified by AiResearch on the DOE'S Near-Term Electric Vehicle contract.

19CRUCIFORM HUB MADEOF 69 SLATS

SLAT

Fig. 4. Composite Hub

Task 4, Test Planning. A spin test of thethe rotor will be planned for Phase II. Inthis test, the flywheel will be suspendedon a quill shaft in a spin pit evacuatedto approximately 1 micron of Hg. Theflywheel will be cycled 100 times betweenmaximum and 33-percent speed. Decelerationwill be accomplished at a shaft power out-put rate of 37 kw by means of a reverse-thrust nozzle on the air turbine. Thewheel will be examined after the test forsigns of fatigue fracture, material pro-perty changes, and any deterioration ordamage.

Task 5, ProducibiIity and Cost Analysis.The costs and processes involved in thefull-scale production of flywheels willbe analyzed to provide data for accuratecost tradeoffs of potential flywheel appli-cations. Cost-effectiveness is affectedmainly by the design configuration, byseal ing for economic use of materials, andby effective tool ing and production tech-niques. Depending on the application,whether vehicular or stationary, or whetherlarge or small, material tradeoffs willproduce the greatest economy. Rotor speed,and therefore rotor size, not only affectthe cost of the rotor, but very stronglyaffect the cost of housing and installa-tion. These relationships between size,speed, and cost will be presented in thistask. ProducibiIity and tooling of uni-axial composite products are areas thatrequire a great deal of development.Certain cost-effective processes are beingappIi ed to the fIywheeI rotor i n Phase I toallow closer evaluation of producibiIity.

Task 6, Final Report and Program Review.All work accomplished will be described ina final report which will include resultsof (1) material evaluations, (2) materialtests, (3) stress analyses, (4) flywheeldynamic studies (critical speeds, creep vsDalance, etc.), (5) producibiIity studies,and (6) Phase II test plan. Fabricationtechniques and equipment will be describedand flywheel outline drawings will beincluded. Photos will supplement thedescriptions.

ACCOMPLISHMENTS TO DATE

The epoxy matrix material has beenqualified as a satisfactory replacement forthe previously used epoxy by hydrostatictesting of flywheel rings. Fig. 5 showsthe hydrostatic test fixture used to deter-mine modulus of elasticity from tests ofring stretch vs internal pressure. Ulti-mate strength also was obtained on thetest rings by bursting them in the fixtureand recording the burst pressure. Place-ment of the test ring in the fixture isshown in Fig. 6. Table 1 compares theresults of the current and previous epoxymatrixes in Kevlar 49.

174

Table 1. Comparison of Current and PreviousEpoxy Matrixes in Kevlar 49.

Fig. 5. Hydrostatic test fixture

TEST SPECIMEN

Parameter

Modul us ofelasticity, psi

Ultimatestrength, psi

Current

14.23 x 106

157,000

Previous

13.37 x 106

163,000

Fig. 6. Hydrostatic test section

Graphite hub slats and the hubassembly mold have been tested and evalu-ated for their ability to produce partswith Iimited stress concentrations, smoothfaying surfaces, and good dimensional uni-formity. Inspection of preliminary samplesproduced by that tooling is under way.

The design analysis has been com-pleted, and the results of that analysishave been applied to the hub tool ing.Specifically, the spoke/rim interfacedesign has been analyzed to ensure thatradial forces result in manageable trans-verse stresses in the rim. The analysisof mechanical and thermal stresses associ-ated with fabrication and assembly willa I low the composite hub to be mated withthe composite rim to compensate for theoperational deformations.

Detailed analyses of stress concen-trations at the hub root have been made asguides to spin testing where actualstresses must be empirically determined.

FUTURE WORK

Upon the completion of the three 1-in.hub test specimens, they will be testedstatically and cyclically to determine thebasic slat quality and the quality of theslat-to-slat bond integrity. These testsresults will be used to validate the analy-sis. Final adjustments, if needed, willbe made to tool ing and processes in pre-paration for the fabrication of the 4-in.-wide hubs.

175

SESSION IV: SOLAR MECHANICAL

177

PROJECT SUMMARY

Project Title: Solar Mechanical Energy Storage Project

Principal Investigators: H. M. Dodd and B. C. Caskey

Organization: Sandia Laboratories, Division 4716Albuquerque, NM 87185(505) 264-8835

Project Goals: The goal of this project is to identify and develop mechanicalenergy storage technologies for eventual commercialization, with theemphasis on solar (sun and wind) sources. The project utilizesresults of systems analysis, special studies, and detailed designstudies to define areas for hardware demonstration projects. Theresulting cost and performance data will be available for considera-tion by industry. The project involves both Sandia in-house analysisand contracts with universities and industries.

Project Status: System analysis has identified flywheel energy storage as the mostpromising of the mechanical energy storage technologies for resi-dential applications. Therefore, ongoing special study and de-tailed design study contracts (three with universities, two withindustries) address various aspects of stationary flywheel energystorage. All will be completed late in FY 1979 when a decisionon construction of a demonstration flywheel system will be made.

Additionally, pneumatic (compressed air) industrial energy storagesystems are being studied, along with some advanced concepts.

Systems analysis, in addition to evaluation of proposed systems,is being used to investigate the relationship of various storageusage strategies with factors such as time-of-use pricing andenergy sell-back.

Contract Number: AT (29-D-789

Contract Period: Oct. 1977 - Sept. 1982

Funding Level: FY 1978-$285,000; FY 1979-$553,000

Funding Source: DOE, Division of Energy Storage, Advanced PhysicalMethods Branch

179

SOLAR MECHANICAL ENERGY STORAGE PROJECT I

B. C. CaskeySandia Laboratories

Systems Analysis Division 4716Albuquerque/ New Mexico 87185

ABSTRACT

Plans for the Solar Mechanical Energy Storage Project are presented,and its current status is described. Systems analyses have identifiedflywheel energy storage systems as the most promising technology forsmall to intermediate load applications. Detailed design studies andspecial studies are under way to investigate this storage mode. Addi-tional analyses and pneumatic system studies are planned for FY79 as wellas a decision on whether to construct a prototype flywheel energy storagesystem.

INTRODUCTION

In FY77 Sandia Laboratorieswas assigned primary responsibilityfor the Solar Mechanical EnergyStorage Project by the US Depart-ment of Energy (DOE, formerly ERDA).Activities for that year includeddevelopment of an optimizing sys-tems analysis computer programcalled ENERA and subsequent analy-ses utilizing ENERA. H. M. Dodd,et al presented a paper in Septem-ber 1977 titled "An Assessment ofMechanical Energy Storage for SolarSystems."1 This paper and laterresults showed that flywheel sys-tems are the most promising mechan-ical energy storage devices forresidential applications. There-fore, the primary emphasis hasbeen on development of low-coststationary flywheel technology.Activities for FY79 will include amore detailed investigation ofpneumatic energy storage systemsfor industrial and residentialapplications, storage for agricul-tural use, and an evaluation ofadvanced concepts.

PROJECT DESCRIPTION

Described below are the organ-

ization, scope, and managementphilosophy of the Solar MechanicalEnergy Project, followed by an ac-count of the status and plans ofeach of the three project tasks.

ORGANIZATION

The DOE's Advanced PhysicalMethods Branch funds this project.This Branch, directed by Dr. G. C.Chang, is in the Energy StorageSystems Division of Energy Tech-nology. Figure 1 shows the SandiaLaboratories project organizationand identifies the principals withtheir functions. Sandia will de-vote -2.5 man-years to this projectin FY79.

J. H. SCOTTDIRECTOR OF

ENERGY PROGRAMS

XG.E. BRANDVOLD

SOLAR ENERGY PROJECTS| DEPARTMENT

MANAGEMENTH. H. DODD

SYSTEMS ANALYSISDIVISION

C. CASKEYEM^ ANAl Y M

[ H 7 E. SCK1LDKNECHTTECHNICAL CONTRACT

MANAGEMENTSTAFF

Fig. 1. Solar Mechanical Energy Stor-age Project Organization at SandiaLaboratories, Albuquerque, NM.

Sandia Laboratories is a US Department of Energy (DOE) facility. Thiswork was supported by the Division of Energy Technology, US DOE, underContract AT(29-1)-789.

180

SCOPE

This project is concerned withmechanical energy storage modesthat may be used in conjunctionwith solar (including wind) inputssupplying small to intermediateloads. Since, for central stationutility applications, energy stor-age would be tied to the gridrather than to a solar energysource, this project does not ad-dress central station applications.Figure 2 lists the sources, storagemodes, and applications to be con-sidered.

| PROJECT PHILOSOPHY |

[PROJECT SCOPE IFLYWHEELS .

SOLAR

WIND

RESIDENTIAL

INTERMEDIATE

ADVANCED'

I SOURCE| |MODE| | APPLICATION!

Fig. 2. Solar Mechanical EnergyStorage Project Scope.

MANAGEMENT PHILOSOPHY

The goal of this project is toidenti fy and develop appropriatemechanical energy storage technolo-gies for eventual industrial com-mercial uses. Figure 3 reveals theunderlying relationship among tasks.Based on existing technology, fly-wheel energy storage has beenidentified as the most promisingresidential application; detaileddesign studies are under way togenerate credible cost and per-formance data* During the fourthquarter of FY79, additional systemsanalyses will use these data toform the basis for a decision onconstructing and testing a demon-stration flywheel energy storagesystem. These detailed designstudies, in conjunction with sys-tems analysis, may identify collec-tor-to-storage interfaces that re-quire detailed development.

MOUSING 'NW U C W O N S J

DEMONSTRATION > „PROJECTS J

DETAILEDDESIGNSTUDIES

TECHNOLOGY TRANSFER TO INDUSTRY

Fig. 3. Solar Mechanical EnergyStorage Project Management Philos-ophy Leading to Mechanical EnergyStorage Commercialization.

This information flow cycle isexpected to continue for other stor-age modes and applications untilthey have been either demonstratedor rejected.

Additional advanced mechanicalenergy storage concepts are ex-pected to be proposed, studied andentered into the development cycleas appropriate.

Figure 4, which presents theproject milestone chart for thenext three fiscal years, reflectsthis philosophy. Positive deci-sions are assumed on both flywheeland pneumatic system demonstrationprojects. Development of an inter-face between a vertical-axis windturbine (VAWT) and a flywheel energystorage system (FESS) is also shownas a likely candidate.

TASK I. RESEARCH ON ADVANCEDCONCEPTS

Systems Analysis. During FY78, thesystems analysis code ENERA wasused to develop representative sup-ply and demand data for use in thedetailed design studies for fly-wheel energy storage systems. Bothphotovoltaic2 and wind3 energysources were used. Improvements tothe code are planned in three areas:

1. Time-of-Day Pricing—utilitypower costs will be a function

181

00

ro

Task I. Advanced Concepts

A. Systems Analysis (SLA)

B. Special Studies(Contracts)

1. Advanced PneumaticStorage

2. Cellulosic Flywheel3. Variable Inertia Fly-

wheel4. Flexible Flywheel5. Industrial Compressed

Air Storage6. Advanced Concepts

Task II. Prototype Develop-ment

A. Detailed Design Studies

1. Contract Management2. Residential PV/FESS3. Residential Wind/FESS4. Intermediate-size FESS5. Advanced Concepts

B. Storage/Collector Inter-face

1. VAWT-FESS Interface

C. Construct and Test

1. FESS (10 kWh)2. Pneumatic ESS3. FESS (100 kWh)

Task III. Headquarters Suppt

Note: V StartA Finish

FY79O N D J F M A M J J A S

-A

FY80O N D J F M A M J J A S

.RFP.

RFP,

.RFP

FY81O N D J F M A M J J A S

-A.RFP.

Fig. 4. Solar Mechanical Energy Storage Project Milestone Chart

of the time of day. Many utili-ties are proposing or adoptingthis technique to reflect theirproduction costs. The effect oftime-of-day prices on storagedesirability will be studied.

2. Utility Sellback—energy sell-back to the utility at somefraction of the utility's sell-ing price may affect the use-fulness of storage, since thegrid essentially providesstorage capability.

3. Storage Logic—with the im-provements listed above, thedisposition of energy from thesource as a function of time,state of storage, anticipateduse, and predicted supply,poses a significant modelingchallenge. Several storagestrategies will be tried to de-termine the appropriate levelof sophistication.

<In addition to refinements in

the flywheel system analyses, thefollowing areas will be exploredsubject to manpower constraints:

1. Pumped Hydro/Wells—undergroundaquifers have been proposed forstorage by utilizing existingwells and adding a downholepump/generator and an above-ground pond.

2. Industrial Compressed Air—JalarAssociates is studying the in-dustrial use of compressed airas a power source as an appro-priate use of solar energy(wind turbine) and storage.

3. Agricultural Applications—theBDM Corporation, under contractto Sandia's photovoltaic (PV)project, is identifying agri-cultural applications for PVenergy. One application thatwould require daily storage isa dairy farm with two dailypeak demands, just before andafter PV energy is collected,which requires that essentiallyall the collected energy flowthrough storage.

4. Other Advanced Concepts—newstorage or applications con-

cepts are periodically forth-coming from individuals, uni-versities, and industry. Eachconcept must be examined toassure that no good idea isoverlooked. All specialstudies funded under this pro-ject resulted from unsolicitedproposals judged to havepotential payoff.

Special Studies. The followingspecial studies will be describedin detail in subsequent presenta-tions at this conference:

1. Variable-Inertia Flywheel—David G. Ullman from UnionCollege in Sch'enectady, NY.

2. Cellulosic Flywheels—ArthurG. Erdman from the Universityof Minnesota in Minneapolis,MN.

3. Flexible Flywheels—John M.Vance from Texas A& M inCollege Station, TX.

There are two additional specialstudies:

1. Industrial Compressed Air—Jalar Associates is investi-gating the feasibility ofutilizing solar sources togenerate compressed air forstorage and use by industriespresently consuming significantquantities of compressed airin their operations. The6-month study began in Septem-ber and will be reported indetail at the next contractor'sconference.

2. Pneumatic Energy Storage Sys-tem—Sandia has been negotia-ting for several months toarrive at satisfactory termsfor placement of a contractfor a 12-month study of pneu-matic storage systems and com-ponents. This study will alsobe reported in detail at thenext conference.

TASK II. PROTOTYPE DEVELOPMENT

Detailed Design Studies. Two de-sign studies are under way forresidential flywheel energy-storage

183

systems and will be described atthis conference. They are:

1. Residential Flywheel withPhotovoltaic Array Supply—Francis C. Younger from Wil-liam M. Brobeck & Associatesin Berkeley, CA.

2. Residential Flywheel with WindTurbine Supply—Theodore W.Place from Garrett-AiResearchin Torrance, CA.

Depending on the results ofthese contracts and additionalanlyses, a request for proposal(RFP) may be issued late in FY79.This RFP will request study of anintermediate-sized system, perhapsfor an agricultural application.

Storage/Collector Interface. Basedon ongoing studies and analyses,it is anticipated that an RFP willbe issued during the fourth quarterof FY79 to develop the specialhardware required to interfacestorage to the collector, A VAWTinterface to a flywheel energystorage system is a likely candi-date. Another possibility is awind-turbine interface to an aircompressor.

Construct and Test. Assuming fav-orable results from either or boththe flywheel and pneumatic energystorage system studies now underway, RFPs will be issued to con-struct and test demonstrationsystems. This decision milestonewill occur during the fourthquarter of FY79.

TASK III. HEADQUARTERS SUPPORT

Project personnel support DOEHeadquarters by acting as techni-cal monitors for DOE contracts andby attending meetings and con-ferences.

SUMMARY

The Solar Mechanical EnergyStorage Project is committed todeveloping appropriate mechanical-energy storage technologies foreventual commercialization by in-dustry. In the initial phase ofthis project, small-scale flywheel

and pneumatic systems are subjectedto detailed scrutiny. Advancedconcepts and unique applicationsare constantly being evaluated forfurther study.

A major decision milestonelate in FY79 will provide for con-structing and testing one or moreenergy storage systems.

REFERENCES

1. H. M. Dodd, R. E. D. Stewart,et al, "An Assessment of Mechan-ical Energy Storage for SolarSystems," 12th IntersocietyEnergy Conversion EngineeringConference Proceedings, Wash-ington, DC, August 28-September2, 1977, Vol II, pp 1174-1180.

2. B. C. Caskey, Supply and DemandData for the Conceptual Designof a Residential FlywheelEnergy Storage System, Photo-voltaic Array Supply, SandiaLaboratories, June 1978.

3. B. C. Caskey, Supply and DemandData for the Conceptual Designof a Residential FlywheelEnergy Storage System, WindTurbine Supply, Sandia Labora-tories, July 1978.

184

PROJECT SUMMARY

Project T i t l e : Invest igat ion o f the Potent ia l of a Variable I n e r t i a FlywheelEnergy Storage System

Principal Invest igator : D. G. Ullman

Organizat ion:

Project Goals:

Project Status:

Union CollegeMechanical Engineering Dept.Schenectady, NY 12308(518) 370-6264

Study o f dynamics of a Band Type Variable Ine r t i a Flywheel (BVIF)with fixed-ratio power recirculation for rotational rate control.From these studies, determine i f the BVIF concept offers potentialadvantages in the areas of economy, re l iab i l i ty , and efficiency.

The dynamic modeling of the BVIF less i t s fixed-ratio power re-circulation system for controlled inertia change was completedand documented. Computer simulation studies were ini t iatedto analyze the dynamic characteristics of this truncated configu-ration. Expansion of the dynamic model to include the powerrecirculation system was started. Parts fabrication/procurementfor a small scale proof-of-concept BVIF was completed andassembly work in i t iated.

Contract Number: 07-3662

Contract Period: June 1978 - June 1979

Funding Level: $25,325

Funding Source: Sandia Laboratories, Albuquerque

185

THE BAND TYPE VARIABLE INERTIA FLYWHEEL AND FIXED RATIO

POWER RECIRCULATION APPLIED TO IT

C. G. UllmanMechanical Engineering Department

Union CollegeSchenectady, New York 12308

ABSTRACT

A flywheel with variable moment of inertia created by thin bands winding about acentral hub in a hollow casing is introduced. This mechanism combines the functions ofenergy storage and power control. Configurations with the moment of inertia changeinternally controlled through fixed gearing are introduced.

INTRODUCTION

Due to its very simplicity, theconcept of storing energy in the rotatingmass of a flywheel seems very attractive.A closer look, however, reveals subtletieswhich significantly complicate the imple-mentation of an operational system.Primary among these complicating factorsis the difficulty of transferring theenergy to and from the energy storage fly-wheel .

The governing energy relation for aflywheel is

E = %Iw2 (1)

where E is the energy content of the fly-wheel, I is the moment of inertia of theflywheel about the axis of rotation and u)is its angular rate. Thus in order toretrieve stored energy from a flywheel therotational rate must decrease. Thisdecrease in rotational rate is not desir-able as systems being powered by flywheelsusually require a rotational rate which isconstant or even increasing. This mismatchin rotational rate is usually compensatedfor by the use of a variable ratiotransmission such as a traction drive, ahydraulic pump and motor, or an electricgenerator and motor. Each of thesesystems has drawbacks either in terms ofcost, reliability, or most importantly,efficiency. Specifically, the electricand hydraulic transmissions (the mosttechnically developed) may be inefficientand costly enough to hamper the practicaldevelopment of commercial flywheel energystorage systems.

If the equation for stored energy isreconsidered, another approach to theproblem becomes evident. It may bepossible to alter the moment of inertia ofthe flywheel to gain added control of therelease of stored energy. The resultingmechanism is called a Variable InertiaFlywheel (VIF). This mechanism is anenergy storage device whose output can becontrolled to meet the load requirements.

THE BVIF

Potential Variable Inertia FlywheelDesigns fall into two classes, fluid andmechanical1'2. The configuration whichseems to have the qualities necessary fora successful system is the coiled bandvariable inertia flywheel BVIF which isshown in Fig. 1. This flywheel iscomposed of a hollow outer casing and aseparate central hub. Connecting these isa band of some flexible material mountedas the mainspring in a watch. Centrifugalforce due to rotation pushes the band tothe outer edge of the hollow casing. Asthe central hub is rotated relative to theouter casing the band is wound onto thehub lowering the moment of inertia of thesystem. What controls the state of theband is a balance between the centrifugalforce on the band, the angular rate dif-ference between the central hub and thecasing and the torque flow through themechanism supplemented by the torquevariation caused by the change in state ofthe band itself. Work on this concept hasappeared since an initial patent in 19653.Much of this work has been in the USSR

1S6-1-53

with two patents and one publishedarticle in English6. However the authorshave been unable to find any dynamicanalysis of such a mechanism in the openliterature. The dynamic balance oftorques, forces, and momentum is surelycomplex. Understanding of the dynamics isrequired to fully understand the potentialof the device.

FIGURE 1

BAND TWE VARIABLE INERTIA FLYWKEEL

The major appeal of this configurationover others presented in References 1 and2 is in its high energy density comparedto other designs and that the moment ofinertia change is accomplished in a rota-tional manner. This latter aspect of theBVIF is very important as all other mech-anical configurations considered1 requirelinear motion to change moment of inertiasuch as moving a mass on a radial spoke.The transmission of powe~ from a high rpmrotating source through a mechanism tocause a small distance translation motionis very inefficient.

FIXED RATIO POWER RECIRCULATION

In References 1 and 2 it was shown thatfor any variable inertia flywheel rotatingat constant rpm it takes one unit of powerto change the moment of inertia for eachtwo units of power exchanged between theVIF and an external system. Thus if a VIFgives up two units of power, one must goto change the moment of inertia while oneto power the load or in an alternate caseone unit of power can come from an outsidesource to change moment of inertia andthen two units can go to the load.

A VIF with the power for inertiachange coming from an outside source ismuch like a mechanical power amplifier.This configuration is the primary consid-eration of the Russian patents*5 and isnot discussed further in this paper.

When one unit of power is fed back tochange the moment of inertia, the poweris said to recirculate within the VIF. Itwould be most desirable to have thisrecirculating power flow for inertiachange to be completely controllableallowing for differing inertia changerequirements. Unfortunately, completecontrollability of the recirculating powercan only be had with a variable ratiotransmission such as that which the VIF isproposed to eliminate. Also, it was shownthat with constant rotational rate loadingthe amount of power to be recirculatedthrough the variable ratio transmissionwould be the same as that output. Thusit appears that the VIF with a variableratio transmission taking power from theoutput shaft to change moment of inertiawould have no better efficiency than afixed inertia flywheel with the variableratio transmission on the output shaft.

However, complete control of therecirculating power may not be requiredfor most tasks as the nature of the load-ing is usually reasonably well known andslight deviations from the desired areusually tolerated; thus it may be possibleto create a fixed ratio power recircula-tion path to meet the requirements of aspecific load. Thus use of the term"fixed ratio power recirculation" impliesthat the mechanism carrying power tocreate the inertia change is of fixedgeometry or at a maximum countably fewvariations in geometry (eg: a fixed ratiotransmission). The result of this VIFwith a fixed path, is a flywheel withfixed properties which differ from thoseof a standard fixed inertia flywheel.Another way of stating this is that nowthe moment of inertia is a function ofangular rate so that the torque is nowtotally a function of angular rate butdifferent from that of a standard fixedinertia flywheel.

PROJECT DESCRIPTION

The two concepts of the Band TypeVariable Inertia Flywheel and Fixed Rat*?Power Recirculation are not independentas, the potential results of Fixed RatioPower Recirculation are dependent on the

187

characteristics of the BVIF considered.Thus the project is organized to firstdetermine the dynamics of the BVIF andthen apply fixed ratio power recirculationto it.

The work on this project is beingapproached by first modeling the dynamicsof the system and then building and test-ing a proof-of-concept model to check theadequacy of the mathematics. Thus thefour resulting tasks are:

1. Dynamic Modeling" of the BandType VIF (BVIF).

2. Construction and Testing of aProof-of-Concept Band Type VIF.

3. Dynamic Modeling of Fixed RatioPower Recirculation as appliedto the BVIF.

4. Construction and Testing of aProof-of-Coneept BVIF with FixedRatio Power Recirculation.

These tasks constitute a program whichis,at this writing, less than one halfcomplete. The first task is finished andits results are reported in this paper.The other tasks are scheduled forcompletion by May 1979.

BAND TYPE VIF DYNAMIC EQUATIONS

Consider the Band Type VIF composedof a band(s) in an inner casing connectedto a central hub as shown in Fig. 2.

The inner radius of the casing is r,.The inner radius of the band materialpressed by centrifugal forces against thewall of the casing is r,. The outerradius of the material wound about the

BAND MATERIAL

inner hub is and the radius of theinner hub is r... It is assumed that onlya small amount of material is floatingbetween the inner and outer portions.There is experimental evidence to supportthis assumption.

The outer casing is rotating atangular rate (i)o and the inner casing atrate U). in the same direction. Thedifference between w. and to causes thematerial to be moved from the inner tothe outer portion or vice-versa thuschanging radii r. and r,. Radii r, andr^ are constant, fixed by the solidgeometry of the hub and casing (exceptfor expansion due to the forces of rota-tion).

FIGURE 2

BVIF PLAN VIEW

Each of the n bands is ofh, width b, and density p.

thickness

The moments of inertia of thissystem are considered in two parts, thatfor the outer casing and the band materialrotating at u , hereafter called I , andthat for the inner hub and material rotat-ing at angular rate oij, hereafter calledI.. Included in these moments of inertiamist be the non-variable, fixed moments ofinertia of the outer casing I and of thehub I. °f

Thus

and

h =

The dependent variables, the momentsof inertia, are functions of the radii r.and r.. These two radii are not indepen-dent with their relation fixed by thetotal amount of material in the casing.With n bands of length L and thickness h,the total volume of material in the VIFis

V = nLbh. (A)

This volume is distributed as twohollow cylinders, one against the outercasing and one around the inner hub or

188

(5)

The small amount of material betweenradii r2 and r3 which is not accounted foris insignificant when compared to thetotal volume of material. Equating (4)and (5) and writing r, as a function of

then

(6)

Thus with Eq. (2), (3) and (6) themoments of inertia of the system can bewritten as a function of radius r^> But,r_ is a function of the time history oftne angular rates u and w..

Although the mass of the materialbetween the inner and outer band masseshas been ignored it does act to transferpower between the material wound aroundthe hub and that against the outer casing.The torque which transmits the power, T ,tries to accelerate the inner hub andretard the outer casing. This torque isgiven by,*

With the external torque on the inner hubas T. and that on the outer casing is Tthe equations of motion for the BVIF are:

and

T = 2—t dt

T - T = ^^o t dt

(8)

(9)

To complete the description of theBVIF system, the characteristics of theexternal systems which relate T , T., w ,and to. must be derived. These external0

systems determine the exact configurationof the BVIF as it is those items outsidethe BVIF itself which cause the BVIF asa power amplifier or a flywheel withFixed Ratio Power Recirculation or in someother manner.

There are many potential ways ofconnecting the two shafts to an externalsystem (1,3,5). Presented here are 2types of loading techniques. The firstof these configurations is what is con-sidered the most basic loading of the

Derivation to appear in a forthcomingpaper.

BVIF. The second of these utilizes FixedRatio Power Recirculation and is one ofmany fixed ratio systems being studied.

The VIF as modeled has effectivelytwo shafts leaving the system. One shaftis connected to the outer race, is rotat-ing u>Q and has an external torque on it ofT (positive in the direction of rotation).Tne other shaft is connected to the innerhub, is rotating at angular rate u>. andhas an external torque on it of T.(positive in the direction of rotation).The loads on these two shafts are indepen-dent.

The simplest loading (called a Type Isystem) which will be considered is onewhere the band begins at time zero in thenatural state of being pressed entirelyin the outer casing which is rotating atangular rate w and has no load torque onit (T = 0) an8 the inner hub is, at timezero clutched to a load. In other wordsthe outer casing is free and the shaftfrom the inner hub is connected tc theload.

The load torque is represented by

where

= -T ±,

(10)

(11)

Ty is dry friction, I is the moment ofinertia of the load and T^ is an aero-dynamic drag coefficient. This config-uration is shown in Fig. 1.

A type of BVIF with Fixed Ratio PowerRecirculation has the load again attachedto the shaft from the inner hub, but nowalso has the outer casing connected tothe inner hub by the band and by a gearset. Thus in this configuration, here-after referred to as Type II, the angularrate of the inner hub shaft and the outercasing are related in some fixed manner.A sketch of this configuration is shownin Fig. 3 where the gear set is anepicyclic. This form of gearing has beenchosen as the ratio between w. and 0)needs to be close to unity. The exactvalue of the gear ratio depends on thenature of the loading and the BVIF config-uration as a gear ratio close to unitywill cause the band to transfer slowlyand a ratio far from unity will cause arapid transfer creating a higher torqueat a given rotational rate. For this

189

configuration with the gear ratio, g, andthe load rotational rate equal to theinner shaft rate then

Also T.L-T.-To/g

(12)

(13)

where T is as defined in Equation 10.LJ

These equations along with those describingthe dependence of I , I , and T on theband state and angular rate historyconstitute a set of nonlinear differentialequations which are stepwise solved on thecomputer.

Input for each run of the simulationmust be:

EPICYCLIC GEAR

SET

FIGURE 3

FIXED RATIO POWER RECIRCULAIIOH BAND

TYPE VARIABLE INERTIA FLYWHEEL

Unfortunately whichever type of BVIF/load system is chosen the resulting differ-ential equations are highly nonlinear.Attempts to linearize them have provenunacceptable in that the characteristicsof the system have been lost. Thus theequations have been programmed on adigital computer and step wise integratedusing a Runge-Kutta technique.

BVIF SYSTEM SIMULATION

To demonstrate the gross nonlinearityof the equations representing a BVIFsystem the Type I system will be analyzed.With T = 0 Equations 8 and 9 reduce to

and

T i + T t

_x = —t dt

—dt

where from Equations 10 and 11

(14)

(15)

(16)

Material Density,Material Width,Material Thickness,Number of Bands,Length of Each Band,Inner Hub Radius,Outer Casing Inner Radius,Fixed Inner Moment of Inertia,Fixed Outer Moment of Inertia,Initial Inner Radius of Band Material,Initial Outer Angular Rate,Load Friction Torque,Load Aero Torque Coefficient, andLoad Moment of Inertia.

For each run to be discussed in this paperthe band material is steel, density =.00073 lb sec2/in, it is 1 in wide, .005in thick, and there are 2 bands eachoptimally 324 ft long.* The inner hub has.5 inch radius and the outer casing has5 inch internal radius. The abovegeometry is that which is used in anexperimental apparatus being constructed.In these simulations the mechanism has anestimated fixed inner moment of inertiaof .01 lb-sec^-in and an estimated outerfixed moment of inertia of .1 lb-sec^-in.

The initial conditions which must beset for each run are the initial innerangular rate, the initial outer angularrate, and the initial position of the bandmaterial. This latter condition is givenin terms of the initial value of theradius of the band material wrapped on theinner hub. Thus .5 inch would imply nomaterial is around the hub and all thematerial is in the outer casing. This isthe assumed usual initial band condition.

* "optimally" infers that the moment ofinertia of material in the BVIF is suchthat the difference between the maximummoment of inertia and the minimum momentof inertia is as large as is physicallypossible. The derivation of this resultwill appear in a forthcoming paper.

190

As an example of the performance ofthe BVIF the outer casing initial angularrate was set at 100 rad/sec (995 rpm) andthe initial inner hub angular rate wasvaried below these outer rates. Theseinitial conditions result in initialstored energy of about 3200 in 1b (267 ftlb, .1 watt hr). The load conditions wereset with T = 0 (aero drag torque) and Tp

(load constant friction torque) and I(load inertia) as variables. If the aerodrag torque were not zero it would combinewith the inertia to represent an essential-ly constant torque throughout accelerationand steady state operation. Thus varyingonly the constant load torque and itsinertia are sufficient to understand theoperation of the Type I BVIF. The base-line values used for the load variablesare T^ = 1 in-lb and I = 1 lb-sec^-in.If the baseline load were connected to afixed inertia flywheel of the same initialinertia and energy content as the EVIFand rotating at 100 rad/secs the resultantangular rate time history of the loadwould be:

0) = 100 - .66t(rad/sec),

a constant deceleration with u = 0 at 150seconds. This behavior can be compared tothat of the BVIF operating with differentinner hub initial angular rates, w. . Thecases to be considered are: o

history Fig. 5, shows the band transfer-ring from the outer casing to the innerhub in 11 seconds ending the run. Duringthis time the BVIF only releases about 25percent of its energy to the load as shownin the energy time history plot, Fig. 6.The torque provided by the BVIF to theload increases steadily with time as shownin Fig. 7.

(rad/uc)

FIGURE 4

Load angular rate time history with

Initial outer casing angular rate of 100

rad/sec and variation of Initial load

(inner hub) angular rate. T f = .108 N-m

(1 in-lb), I. = .108 N'm-sec2 (1 1n-lb-sec*).

Case (rad/sec)>

t (sec)

2040608095

1115.557

120132

ND

P05ITI0H

Note that the last column is the totaltime it took the band to reach one of itslimits in winding or unwinding. In alldata plots which follow, time is normalizedwith respect to this total run time. Thismakes visualizing the effects of parameterchanges somewhat easier.

For Case 1, the initial angular rateof the load is only 20% of that of theouter casing containing the band material.As can be seen in Figure 4 the angularrate remains essentially constant through-out the first 2/3 of the run beforeincreasing at the end. The band material

.4 .6HOUWLIZED TIME t/t»

FIGURE 5

Band Position time History with initial

outer casing angular rate of 100 rad/sec

and variation of Initial load (inner hub)

angular rate. T p = .108 N-m (1 in-lb),

IL = .108 N-m-sec2 (1 in-lb-sec2)

191

(«tt-hr)

0 .2 .« .«

FIGURE 6BVIF Energy content time hi-story with

In i t ia l outer casing angular rate of 100

rad/sec and variation of in i t ia l load

(inner hub) angular rate. Tp • .108 N-m

(1 in - lb ) , I L ' .108 N-m-sec2 (1 in-lb-sec2)

Physically, what is happening incase 1 is that the torque load isapproximately equal the torque transfer-red by the bands, T of Eq. 7. Thistorque is produced By the reduction inmoment of inertia of the band material inthe outer casing. Thus the slower innerhub produces the torque T by causing thematerial to wind on the hub. This torquealso balances the loss in angular momentunof the outer casing which is primarilydue to change in moment of inertia whileangular rate holding near constant. Inother words the torque demand by the loadcauses the inner hub to resist accelera-tion resulting in a mass transfer to thehub. This transfer increases the momentumof the hub and decreases that of thecasing. Angular rate changes occur tokeep the changes in momentum balancedwith the torques.

If the initial angular rate isincreased to 40 rad/sec (case 2) there isvery little change in the behavior of theBVIF. The run now takes 15.5 sec and 53%of the energy is given up (most at theend of the run). However, if the initialangular rate is raised to 60 rad/sec asin case 3 the behavior becomes drasticallydifferent.

to•0

vtT0

sD

'l

(•••]

Normalized Time

FIGURE 7

Torque to'load time history with In i t ia l

outer casing angular rate of 100 rad/sec

and variation of In i t ia l load (inner hub)

angular rate. TF • .108 N-m (1 1n-lb),

I, - .108 N-m-sec2 (1 1n-1b-sec2).

In case 3 the angular rate s tar t sto decrease then increases and againdecreases (Fig. 4) . What i s happening canbe explained by examining Equ. 14. Whenthe angular rate slope changes from apositive to negative or vice-versa thenI.(I). = 0 so that Equ. 14 can be rewrittenwxtfe T. = -T T as

i IJ

(17)

From equation 7

TTt

Thus

2 2 2 2-(positive terms)(r, w -r to ).(18)

t(positive terms) ?] -T ] (19)1 L

The first reflex in the curve iscaused by a balance of the terms in Equ.19. The second reflex is more interestingas the terms in Equ. 19 cancel leavingI. = 0. This can be seen in Fig. 5 wheretfte band stops winding onto the inner huband reverses itself winding back intothe outer casing. What is occurring isthat the combination of the

192

2 2r 0) and T terms have become greater than2 i L2 2

r U . With r~ increasing at a rate great-er ?han r_ and UK near the value of tothis reversal can occur.

After the change in the windingdirection of the band the BVIF continuesto give up energy to the load (Fig. 6)while the angular rate slowly decreases.

It must also be noted in comparingCase 1 and Case 3 that the outer casingangular rate increases in Case 1 butdecreases in Case 3. This accounts forthe higher energy release from Case 3where wo final is only 26 rad/sec comparedto 139 rad/sec for Case 1. The reason forthis is that in Case 1 the torque from theband, Tt is higher than in Case 3. InCase 1 this torque is higher than theangular momentum variation caused byinertia variation thus the angular rate iscaused to increase to make up the differ-ence.

As can be seen, explaining the causesof the variations within the BVIF isquite difficult with the continuousbalance of momentum change along withenergy conservation determining the timehistory. If the initial inner hub angularrate is raised to 80 rad/sec, Case 4, thenthe angular rate decreases steadily asshown in Fig. 4. While this is happeningthe band material starts by winding in,reverses itself and starts by unwinding,then reverses itself again and continuesto wind in. In this case the relationbetween the band torque, T , and the loadtorque, T , is such that they change I.twice in Equ. 19. This configuration x

releases most of its energy to the load asdoes Case 5 with an initial angular rateof 95 rad/sec. For both of these casesthe output is essentially that whichwould be of a regular fixed inertia fly-wheel. Namely, with a regular fixedinertia flywheel the equation of motionwould be

10,. = -TL = -T (20)

With I as the total initial inertia of theBVIF, the solution to this differentialequation is

CO.

Thus if i

- .62t (21)

100 rad/sec then it

sec to decay. Cases 4 and 5 take 120 and132 sec respectively. Thus it can beconcluded for the loading used that athigh initial angular rates of the hub theBVIF acts like a standard flywheel. Withlower initial angular rate the band windson the hub faster resulting in lower over-all run time, tn.

KAs can be readily seen the behavior

of the BVIF is variable depending on theinitial status. Simulations have alsobeen performed with the load conditions asvariables. In these simulations certainloads result in a BVIF which gives upenergy to the load while accelerating it.This implies that future work should leadto a BVIF with truly usable characteris-tics.

FUTURE WORK

The above results are a part of acontinuing study and in themselvesincomplete. Simulations are continuingon various fixed-ratio-power-recirculationsystems such as the Type II configurationmentioned earlier. It is anticipated thatfurther simulation of these systems willenhance the understanding of BVIF dynamics

As the proof of the viability of anyconcept is in hardware a BVIF is beingconstructed with the dimensions used inthe computer simulation presented. Thisproof-of-concept model is designed to beconfigured as many different types ofsystems and can be inertially or friction-ally loaded.

REFERENCES

1. Ullman, D.G., "A Variable Inertia Fly-wheel As an Energy Storage System,"Ph.D. Dissertation, Ohio State Univer-sity, March 1978.

2. Ullman, D.G. and Velkoff, H.R., "TheVariable Inertia Flywheel (VIF), AnIntroduction to its Potential," 1977Flywheel Technology Symposium Proceed-ings, Oct. 5-7, 1977, San Francisco,California.

3. Durouchoux, 0., U.S. Patent #3,208,303,1965, U.S. Patent Office.

4. Gulia, N.V., "Centrifugal Accumulator,"USSR Patent #1131894/25-28,November1969.

would take the fixed inertia flywheel 161

193

5. Gulia, N.V., "Variable Moment-of-Inertla Flywheel," USSR Patent#1182724/24-27, July 1969.

6. Gulia, N.V., "A Coiled Band Mechanismfor the recovery of a Vehicle'sMechanical Energy," Jnl. Mechanisms,vol. 3, pp. 113, Pergamon Press.

194

PROJECT SUMMARY

Project Title: Development of Cellulosic Flywheel System

Principal Investigator: Arthur G. Erdman

Organization: University of MinnesotaDept. of Mechanical Engineering125 Mechanical Engineering111 Church Street, SEMinneapolis, MN 55455

Project Goals:

Project Status:

Contract Number:

Contract Period:

Funding Level:

Funding Source:

1. To measure the long-term design strengths and fatigueproperties of flywheel and rotors under vacuum conditions.

2. To develop self centering methods to assemble and balancerotors.

3. To develop a flywheel demonstration model.

The first quarter of the 12-month period project has been completed;experimental apparatus is being designed and built.

07-9072

Aug. 27, 1978 - Aug. 24, 1979

$41,239

Sandia Laboratories, Albuquerque

195

CELLULOSIC FLYWHEELS

A. G. ErdmanAssociate Professor of Mechanical Engineering

D. L. HagenGraduate Research Assistant

D. A. FrohribProfessor of Mechanical Engineering

W. L. GarrardAssociate Professor of Aerrrpace Engineering and Mechanics

University of Minnesota111 Church St. S.E.

Minneapolis, Minnesota 55455

ABSTRACT

Natural cellulosic materials have been shown to have moderately high tensilestrengths, low densities and thus reasonably high specific energies. Thus they appearto be technically feasible and economically competitive for stationary flywheel energystorage applications where volume and total mass are not in question. An overview ofprevious and current research on cellulose flywheels is given. It has been found thatone to two ton birch plywood rotors could be readily assembled and eight ton kraft paperrotors are available. Even larger rotors can be made using existing technology. Thebirch plywood rotors can be made at a cost of 4 wh/$ and the kraft paper rotors areavailable at 11 wh/$, with further improvements in costs predicted. Intrinsic specificenergy of 294 kJ/kG (37 wh/lb) has been demonstrated for kraft paper. The objectivesof current research are to establish the working strengths of cellulose in a vacuum asa function of moisture content and the cyclic fatigue expected in the rotor; to developmethods to attach hubs and balance cellulose rotors; and. to study the fatigue of repre-sentative rotors and hubs under actual cyclic operating conditions. Other objectivesinclude the vibrational problems during operation, and the design and implementation ofcontrols for the flywheel system. The efficiencies and use of continuously variabletraction drives are also being studied.

INTRODUCTION

Natural cellulose fibers have veryhigh tensile strengths on the order of1.0 GPa (150,000 psi). Although the ac-tual strengths of clear wood are onlyabout 160 MPa (23,000 psi), when combinedwith a low density of around 700 kg/m3these strengths result in a surprisinglyhigh intrinsic specific energy of around230 kJ/kg (29 wh/lb). The natural abun-dance of wood and the well establishedindustries produce comparatively lowprices. These result in intrinsic energystorage per cost values of natural cellu-losic materials competitive with highperformance expensive materials such asKevlar, fiberglass, and graphite. (See

Table 1) The major difficulty with natu-ral materials is the distribution ofdefects and variation in properties fromone sample to another. By laminatingmultiple layers of veneers to make ply-wood, these effects can be minimized andaveraged out to retain the high strength.The effectiveness of this can be seen inthe widespread use of glued laminatedbeams and plywood "I" beams for structu-ral use (replacing steel). Similar bene-fits are obtained by separating the fibersand reforming them as paper or hardboard.

The possibility of using cellulosicmaterials to make economically competi-tive flywheel rotors was first recognizedby Rabenhorst in 1972.(1) Cellulosic

196-

rotors appear to be suited to stationaryapplications where economics rather thanmass and volume are the governing factors.The possibility has been mentioned inseveral papers since then.(2) Tests ofwood rods and plywood disks spun to de-struction at the Applied Physics Labora-tory have confirmed the potential of usingcellulose in rotors.(3) Some similartests are being performed at the TechnicalInstitute of iurin. (4) A review of theliterature revealed no work had been doneto assemble and test full scale rotorsfrom wood. (2)

Research was therefore begun under aseed grant from the Minnesota EnergyAgency to assemble a small demonstrationmodel of a flywheel energy storage systemusing a plywood rotor. (5) Figure 1shows a schematic of the demonstrationmodel which houses a 14 inch diameter(14 inch length) flywheel that rotates ata maximum of 5,000 rpm. This pilot studyis being used to gain familiarity with thepractical problems of assembling andbalancing wooden rotors. Considerationwas also given to the task of supportingthe rotor and the associated vibrationalproblems anticipated in the operatingregions. Both electronic (Parajust) andmechanical (a Zero-Max) continuously vari-able speed controls are used to transmitthe power to and from the rotor. The ply-wood rotor weighs 28 kg (62 lbs.). Adescription of the system has been pre-sented elsewhere. (6)

CELLULOSE ROTOR

A more detailed review of the liter-ature was made to assess in greater depththe potentials and difficulties of manu-facturing large changes with the moisturecontent. (7) Very little data exists onthe performance of cellulose under thevacuum conditions required for high speedrotors. Extrapolations from existingdata indicate that wood becomes morebrittle, and the strength and toughnessapparently decrease as it dries out.Elevated temperatures in conventionaloven drying or kiln drying of wooddecrease the strength somewhat. Conser-vative estimates of assembly costs con-tinued to indicate that plywood rotorswould be competitive with Kevlar or fiber-glass rotors in stationary applications.(7)

The questions of the strength ofwood at low moisture content and the

practical problems of assembly andbalancing raised by the initial researchare beina pursued in current researchunder continued funding from the Minne-sota Energy Agency. We propose to scaleup to a 50 kg rotor mounted on a verticalaxis. Some vacuum freeze drying equip-ment is being renovated to compare thestrengths of oven and vacuum freeze driedmaterial using conventional methods.Rotors under actual operating conditionsmay become even dryer. A chamber isbeing connected to a high vacuum pump todry samples out under higher vacuum con-ditions to see if this has any furthereffect on the strength.

The key question is still what isthe design strength of cellulose underthe sustained stress and cyclic loadingconditions in a vacuum experienced by aflywheel rotor. Studies of the basicstress strain phenomena are being pre-pared under funding from Sandia Labora-tories. Measurements of the rate of plas-tic deformation and the acoustic emissionswill be made as functions of the stressand strain under dry conditions.Attempts will be made to correlate thesewith the duration of load effects. Vis-coelastic effects compound the problemand require high resolution measurements.A PDP11-03 with a 14 bit (4% digit) ana-log to digital converter has been obtainedfor these tests. The acoustic emissionsup to 100 kHz will be recorded on tapeand digitized at slower speeds for anal-ysis. Data in the literature indicatethat the rate of plastic deformation andacoustic emissions begin to increaserapidly around the 50% of ultimate stresslevel which seems to correlate with thelong term strength. Measurements of thefatigue strength in tension under dryconditions are also planned.

An elastomeric interface between themetal hubs and the plywood disks has beenfound necessary to avoid shear failurebetween them under operation. (3) Acyclic rotary fatigue test bed is beingdesigned to test the durability of thehub to rotor attachment as well as thatof the rotor itself under actual opera-ting conditions and fatigue regimes. Acontinuously variable drive will be usedto cycle the energy between two identicaldisks.

197

COSTS

The large commodity markets of ply-wood and paper suggest that large flywheelrotors could be made at low cost. Furtherresearch was conducted this summer (sup-ported by the Applied Physics Laboratory)on the economics of manufacturing flywheelrotors. Discussions with several plywoodmanufacturers indicate that cylindricalrotors can easily be assembled from ply-wood disks using existing equipment. Aone ton flywheel 4 feet in both diameterand height would cost $470 to assembleone hundred units, assuming 30% for over-head and profit. The material cost ofbirch plywood is $650 (including 23%waste). This results in a storage coston the order of 15.8J/$(4.3 wh/$). Theseactual quotes on assembly costs are halfthose previously assumed. (7) The birchplywood is available in 1.5 m x 1.5 m (5ft x 5 ft) sheets indicating that a twoton flywheel 5 ft. in diameter and heightcould readily be assembled by any plywoodassembler having this size press. Reduc-tions in assembly costs are possible withhigher production runs.

Preliminary studies of the potentialof using kraft paper to assemble woundrotors also looks promising. Four tonrolls are readily available at most mills,and one mill makes 8 ton rolls 315 in.long. Costs of lltf/lb to 17<£/lb in milllot quantities suggest that storage costsof 11 wh/$ in full scale cylindrical

rotors are presently possible using com-mercially available materials. It isprojected that higher strengths can beobtained by minor modifications in opera-ting conditions. Leboratory tests ofhighly oriented paper have demonstratedintrinsic specific energies of up to 294kJ/kg (37 wh/lb). (8) Methods would needto be developed to be able to approachthese values in commercial quantities.The actual design strengths for flywheelrotors again are unknown and need to beestablished by experiment. Of particularconcern is the higher creep rates exhib-ited by composite cellulose materials.Details of wrapping and supporting suchrotors and the balancing and vibrationproblems would also need to be worked out.

When costs of assembled rotors arecompared, cellulose rotors appear to besurprisingly competitive with high per-formance filament rotors, even in largevolumes. The prices of graphite fibersare predicted to drop down to $23/kg by1985. The glass and Kevlar filaments areexpected to increase with inflationunless large unforseen markets develop.Costs are for both the bare filamertwound rotors being developed byApplied Physics Laboratory and for thecomposite rotors being developed byothers. Current costs of paperboard areobviously competitive with all the fila-mentary rotors shown. Substantial in-creases in the performance of paper arepotentially possible, resulting in

Table 1. Energy Storage Potentials and Costs

Material

Birch PlywoodPaperboardSuperpaperE-glass/epoxyS-glass/epoxyKevlar 49/epoxyGraphite/epoxy

DensityP r.

kg/nT

700100011002100200013201520

Strength0

MPa

120753361100175018001500

FatigueFactor^

f

0.60.40.40.30.30.60.5

SpecificEnergy bkJ/kg

17.67.433.38.564.3

200.120.

MaterialCosts$/kq

0.65c

0.220.22.1.10°4- 5 0d20.00570.00

LaborCostse

$/kg

.30-.45

.03-.10,

.05-.30T

1 .-10.1.-20.1 .-10.7.-10-

StorageCosts$/MJ

54-6334-438-1555-340986-2579

105-1609

590-680s

Estimated for diurnal cycling for 30years (10,00,0 cycles).

baKfS/p: K = Shape Factor = 0.35, S =Safety Factor = 0.70.

cIncludes 23% waste.dCost estimated by P. Ward Hill, graphiteprices to drop 65% by 1985, all others torise with inflation.

eAssembly costs of winding rotors are forbare filament rotors and for epoxy resin-filament composite rotors.

Not commercially available.9Epoxy resins estimated at $2.00/kg.

198

significant increases in overall storagecosts. The major drawback of celluloserotors is their low density and largevolume. This can be used to a benefit whenthey are used as cores or hubs for com-posite rotors using higher performancematerials.

CONTINUOUSLY VARIABLE TRANSMISSION

An efficient, low-cost means of apply-ing energy to and retrieving energy froma flywheel presents a major hurdle thatmust be overcome in the development of apractical flywheel energy storage system.In order to input energy to a flywheel,the rotational speed of an input shaft,which must apply torque to accelerate theflywheel, must be matched to the flywheelshaft speed. This problem is magnifiedwhen the input shaft is driven by a powersource such as a wind turbine, which in-herently possesses large speed fluctua-tions. A parallel problem occurs at theoutput. The most useful final output ofa flywheel energy storage device is 60 Hz,AC electric power. Since constant ACfrequency, variable shaft speed genera-tors are not yet available, the varyingspeed of the flywheel shaft must be buf-fered to drive a constant shaft speed ACgenerator. Therefore, an electric, hy-draulic, or mechanical transmission capa-ble of fulfilling these restraints is anecessary component of such flywheel ener-gy storage systems.

The current flywheel demonstrationmodel at the University of Minnesota(Figure 1) uses a mechanical CVT manu-factured by the Zero-Max Corporation inits output drive train. This CVT hasexhibited promising performance, but itis not presently available in sizes overTs HP. A larger flywheel test stationnow being designed will tentatively incor-porate a 12 HP CVT manufactured by EatonCorporation. This particular transmis-sion, known specifically as the "Cleve-land speed variator", is probably the mostpopular all-metal traction drive systempresently available in the United States.The Cleveland variator has especiallygreat potential as a flywheel transmissionfor two reasons: first, the variatorfeatures a top speed of 5,400 rpm, whichexceeds the planned maximum speed of acellulose rotor; <=-"*ond, the variator hasan overall input-o > put ratio of 9:1 mak-ing it directly a, ,v icable to the expectedoperating speed range of a cellulose fly-wheel. These features raise the possibi-

lity of connecting the variator directlybetween the flywheel and the input powersource or output power generator withoutshaft intermediate step up-step down cou-plings.

One of the goals of the future teststation will be to determine the suitabi-lity of the Cleveland variator as the cou-pling element between the flywheel andexternal components. Tests are planned todetermine 1) the efficiency of the varia-tor under typical operating conditions,and 2) the durability of the variator un-der the power, speed, and cyclic loadingconditions of a flywheel energy storagesystem. If the Cleveland variator provessuitable, a prototype control system toautomatically adjust the CVT to match realoperating conditions will be developed.If the CVT exhibits problems in the fly-wheel application, the feasibility of mod-ifying the CVT to operate in this special-ized application or employing alternatemechanical CVT systems, such as variable-pitch-pulley or free ball geometries, willbe examined.

FLYWHEEL CONTROL SYSTEM

The control system for the flywheelhas been designed to accomplish two objec-tives. These are:

1. Maintain the output angular vel-ocity of the continuously variable trans-mission (CVT) constant as the input velo-city from the flywheel varies.

2. Control the input velocity to theflywheel so as to simulate a typical dutycycle.

The first object is accomplished by arelatively simple feedback control system.The control system consists of a tachome-ter, a DC motor (See Figure 2 ) , and asso-ciated electronics. The control system isshown in functional block diagram form inFigure 2. The output angular velocity ofthe CVT is measured by an analog tacho-meter. The output of this tachometer is avoltage proportional to the angular velo-city output of the CVT. This signal iscompared to a DC voltage proportional tothe desired output angular velocity. Thedifference in these two voltages is theerror signal which activates a small DCservomotor. This servomotor drives ascrew which cont ils the gear ratio of theCVT. For example if the error signal ispositive (the output velocity is less thanits desired value) the motor drives thescrew in such a way as to increase the

199

gear ratio. As the gear ratio increases,the output velocity increases, the errorsignal goes to zero, the motor stops andthe system operates at its desired outputspeed.

In practice, of course, the inputvelocity to the CVT will be constantlychanging; therefore, the output controlsystem will operate constantly to regulatethe output velocity. The control systemincludes a variable gain and variabledead zone. One objective of the experi-ments will be to determine if this controlsystem can maintain the output speed con-stant within acceptable tolerances (^.5%).

An analysis of the simplified systemindicates theoretically that this systemwill perform as desired. The gear ratioof the CVT is defined as

in

the error signal is

u = Kerror t vWdesired " wout'

(1)

(2)

and the angular velocity of the motor is

wmotor = t e r r o r = KmKt ((i)desired " %uz^

(3)The gear ratio is

N = N(°) + V o V t ^desired " wout>d '

(4)

° r S = W t ("desired " wout> < 5 >

where K = gain constant of the CVToperation screw, unit changesin gear ratio/radian of opera-ting screw rotation

K_ = motor gain constant, radians/™ volt

Kt = tachometer gain constant,volt/radian

When Uout = w d e s l > e d, dN = 0_ If U()ut <

desired > 0 and if u) o u t > desired ^< 0. Thus the gear ratio always changesin such a way to reduce the error in angu-lar velocity to zero. Since the system isfirst order, no overshoots will exist.

Of course in the real system unmodeleddynamics will increase system order andnonlinearities further complicate the pro-blem; therefore, it may be necessary toredesign the feedback control logic ifdifficulties such as hunting oscillationsdevelop in actual operation.

Microswitches are mounted at eitherend of the operating screw on the CVT sothat the system can be disconnected whenno further changes in gear ratio are pos-sible.

The flywheel is driven by an AC in-duction motor. The speed of this motor iscontrolled by a PARA-JUST controller. Thiscontroller is a solid state device whichregulates the speed of an induction motorby changing the frequency of the inputvoltage to the motor. It maintains con-tact torque output from the motor up torated speed (power varies linearly withspeed). The PARA-JUST automaticallyaccelerates the motor following a ramp in-put ranging from 15 seconds to 150 seconds.It can also be adjusted to follow any com-mand input voltage and has been modifiedfor manual control. In simulating theduty cycle a variable voltage will be in-put to the PARA-JUST which will then drivethe motor and the flywheel so as to approx-imate a typical input from an alternativeenergy source as a wind energy generator.

ROTOR VIBRATIONS

A preliminary analysis of criticalspeeds of the flywheel system mathematicalmodel indicates a fundamental natural fre-quency in the vicinity of 2400 rpm, midwayin the originally planned operating speedrange. The rationale in this analysis isthat the natural frequency of the rotor-elastic support system is a valid indica-tor of critical speeds, as support asym-metries, damping and bearing tolerancesare small. The lowest critical speed con-dition is assumed governed by the rotorand its flexible support; the criticalspeed range is likely to be narrow, closeto that of the transverse natural frequencyof the assemblage.

Refinement of this analysis was con-ducted to include the effects of the iner-tia of the hubs supporting the rotorthrough elastomeric disc-like pads sand-wiched in the hub-flywheel interface.

The model used in that analysisis shown in Figure 3.

200

As the lowest natural frequency is asso-ciated with essentially a quasi-staticdistortion mode, a distortion was usedderived from statics. An energy balancethen provided the prediction of criticalspeed:

2k. 1+ ke/k3

1 + 2 Q (l1ke/k3

k kwhere k = .

e K 1 +The effects of elastomeric pad stiffnessin relationship to k , and hub-to-rotormass ratio m/M, can be demonstratedgraphically with the three-dimensionalmanifold (see Figure 4).

This demonstrates the sensitivityof the lowest critical speed to hubinertia and pad stiffness. For large padstiffness (K /K = 0) and sma1! hub mass(jjj- = 0) the critical speed is gov-erned by the shaft and bearing supportingthe flywheel inertia. As these ratiosincrease from zero, a pronounced reduc-tion in critical speed occurs, and hencelow ratios are desired if critical speedmaximization is desired.

In the prototype system,Ke/K3 = .18. Hence, un/|2ke/MThe subsequen' design trade-offs areassociated with the selection of a padelastomer with inherent complex dampingmodulus sufficient to attenuate criticalspeeds, while simultaneously loweringthese speeds due to the attendant elasticmodulus of the pads. Similarly, the hubinertia needed to provide a sufficientlystiff hub to load the elastomer in a con-trolled manner also reduces the criticalspeed. Refinement of the design mustheed these tradeoffs; it is likely thatthe critical tradeoff will involve thepad damping — pad stiffness effect ifother sources of damping in the systemare minimal.

CONCLUSIONS

These studies continue to indicatethat rotors could be assembled from

commercially available cellulose materialswith energy storage costs competitive jwith proposed rotors made from fiberglassor Kevlar. Large multiton rotors can bereadily assembled now in commercial plantsfrom plywood or paperboard. Such rotorsshow potential for stationary applicationswhere cost is the primary criteria. Ply-wood also appears to be a promising mate-rial from which to construct inexpensiverotor hubs.

Data on the general trends affectingthe long term strength is available.Experiments are in progress to measurethe actual performance of the materialsand assembled rotors under vacuum condi-tions. The practical details of assemblyand balancing are being studied. The fac-tors of outgassing and sustained highvacuum are major unknown factors.

Analytic expressions for criticalspeeds are being refined and verifiedexperimentally. Continuously variabletraction drives appear to be an efficientmeans of transferring power to and fromthe rotor. Control algorithms for theoperation of the flywheel system are beingdeveloped and implemented.

ACKNOWLEDGEMENTS

We would gratefully acknowledge thefinancial support for the different areasof research as provided by the MinnesotaEnergy Agency, the Applied Physics Labor-atory, and the Sandia Laboratory. Wewould also acknowledge the assistance ofthe large number of students who have con-tributed so much time and effort to theproject. Thanks are due also to theEaton, Zero-Max, Onan, 3M, and LenderinkCompanies, and to the Forestry, Physicsand Chemical Engineering Departments whohave donated, or made available equipmentand materials, withcv <hich much of theprogress would not have been possible.

REFERENCES

1. Rabenhorst, D. W., "The Applicabilityof Wood Technology to Kinetic EnergyStorage." Johns Hopkins University,Applied Physics Laboratory, APL/JHUTechnical Digest, Vol. 11. No. 5, May/June 1972, pp. 2-12.

2. Hagen, D. L. and Erdman, A. G., "Fly-wheels for Energy Storage: A Reviewwith Bibliography," Design Engineering

201

Technical Conference, Montreal, ASMEPaper No. 76-DET-96, Sept. 26-29,1976.

3. Rabenhorst, D. W., "Composite FlywheelDevelopment Program - Final Report,"Johns Hopkins University, AppliedPhysics Laboratory: APL/JHU SD0-4616ANSF Grant No. AER 75-20607, May 1978.

4. Genta, G., Institute Delia Motorizza-zione, Politecnico Di Torino, Torino,Italy, Private Communication, Febru-ary 1978.

5. Erdman, A. G., Frohrib, D. A.,Garrard,W. L., Hagen, D. L., Carlson, T. P.,"The Development of a Cellulose Fly-wheel System for Rural Wind EnergyStorage: Final Report," Universityof Minnesota, for the Minnesota Ener-gy Agency, July 1978.

6. Erdman, A. G., Frohrib, D. A., Carl-son, T. P., Hagen, D. L., and Garrard,W. L., "The Design of a Wind EnergyStorage System with a Cellulosic Fly-wheel," Proceedings of the 1977 Fly-wheel Technology Symposium, October5-7, 1977.

7. Hagen, D. L., "The Properties ofNatural Cellulosic Materials Pertain-ing to Flywheel Kinetic Energy Stor-age Applications," Proc. 1977 FlywheelTechnology Symposium, October 5-7,1977, San Francisco.

8. Stockman, V. E., "How Strong CanPaper Be Made?" TAPPI, Vol. 59,No. 3, March 1976, pp. 97-101.

9. Chiao, T. T., "Fiber Composite Mate-rials and Their Application to Energy-Storage Flywheels," Lawrence Liver-more Laboratories, 1978.

10. Hill, P. Ward, Hercules Inc., Cumber-land, Maryland, Private Communication,October 1978.

202

LEGENDA = Flywheel HousingB = Drive Motor (BelowC = Clutch-CouplingD = Vacuum Pump (Below)E = Zero-Max DriveF = D.C. Control MotorG = A.C. Generator (Below)

Fig. 1. Flywheel Demonstration Model

w\ \ \\\ \ \\ \ w wwwwwwwFig. 3. Vibration Model

/ * / 7 i 4 2 /J

Fig. 4. Natural Frequency Variations.

CVT

TransmissiontotioControl

radiometer

Motor

Mou.t

out

ErrorFig. 2 . Control System Schematic

'Desired

203

Project Title:

PROJECT SUMMARY

Research and Development f o r I n e r t i a l Energy Storage Based on aF l e x i b l e Flywheel

Pr inc ipa l I n v e s t i g a t o r : J . M. Vance

Organ iza t ion :

Pro jec t Goals:

Project Status:

Texas ASM UniversityCollege of EngineeringCollege Station, TX 77843(713) 845-6225

Experimentally and theoretically analyze the rotational charac-teristics of proof-of-concept flexible flywheel configurationsand devise methods for suppressing any undesirable whirl modes.Use the results of these analyses to develop the conceptualdesign of a flexible-flywheel energy storage system suitablefor interfacing with a small size solar energy source.

The flexible flywheel support structure and associated equip-ment were moved from the University of Florida and installedin the newly established flywheel experimentation area at ASM.Installation of dynamic measurements instrumentation on thestructure was begun. A 20-inch diameter flexible flywheelfrom the University of Flo.ida was spun on the structure andmovies were taken (at both normal and high speed) of i t s non-synchronous whirl motion. The movies have not yet beenscreened for study.

A concept has been developed for gimbal mounting the motor/generator in a flexible flywheel energy storage system for thepurpose of suppressing nonsynchronous whir l . The mathematicalanalysis of this concept has begun. Sandia has been requestedto authorize the design and construction of a small table topmodel to explore this gimbal concept.

Contract Number: 07-3693

Contract Period: Sept. 1978 - Sept. 1979

Funding Level: $39,708

Funding Source: Sandia Laboratories, Albuquerque

205

A CONCEPT FOR SUPPRESSION OF NONSYNCHRONOUS WHIRL IN FLEXIBLE FLYWHEELS

Dr. John M. VanceMechanical Engineering Department

Texas A&M UniversityCollege Station, Texas 77843

ABSTRACT

An energy-storage flywheel sized for home or farm use is being developed at TexasA&M University. A unique feature of this "flexible flywheel" is its construction fromflexible fibers (such as synthetic rope) with no bonding agent. The potential advantagesare low cost, self balancing, and increased safety in operation. The one current dis-advantage, a whirling instability, is the major subject of the present contract. Aciesigp concept to suppress the instability is now being analyzed and is described herein.

INTRODUCTION

In 1975, Dr. R. T. Schneider at theUniversity of Florida conceived the ideaof a flexible flywheel made of rope forenergy storage. The idea was to developa cheap, small scale, energy storage de-vice to make solar or wind-powered gener-ators practical for home or farm use.

Beginning in 1976, a flywheel testfacility was developed and the first ropeflywheels were spun up. It soon becameevident that one of the advantages of theflexible flywheel, its self-balancing fea-ture, had been bought at the price of asubsynchronous dynamic instability causedby internal friction, since the whirl crit-ical speed is well below operating speed.

After it became evident that the maj-or (and apparently the only) technicalproblem of the flexible flywheel is one ofrotor dynamics, the author of this paperjoined work on the project. The projectwas subsequently moved to Texas A&M Uni-versity when the author moved there. Itis less than two months since funding be-gan, at the writing of this paper.

ADVANTAGES OF THE FLEXIBLE FLYWHEEL

Figure 1 shows a photograph of theflexible flywheel. Notice that the supportropes carry only the weight of the fly-wheel. Radial stiffness is provided bygravity only, as in a pendulum. The ques-tion of torque capacity and accelerationhas been investigated and appears not to

be a problem.

Fig. 1 Flexible Flywheel

The advantages claimed for this con-figuration are:

1. High strength fibers in pure ten-sion with no bonding material tocreate mismatch in elasticity orstrength.

2. Self balancing-highly flexiblerotor operates at supercriticalspeeds.

3. Simple construction should becheap to manufacture.

4. Gradual failure mode, with earlywarning by whiplash or hoopgrowth.

5. Less destructive failure mode.These advantages mean that the flexibleflywheel should be safer and less costlythan conventional solid flywheels.

206

FLEXIBLE FLYWHEEL DESIGN EQUATIONS ANDCONSTRAINTS

Analysis has shown that to optimizeenergy density, a flywheel should be con-structed of high strength materials tooperate at high speeds. The maximum stor-age energy in a hoop flywheel is

E = TTRP ft-lb

E = (.3768)(10 )nRP KW-hrs.

(1)

where R = hoop mean radius, ft.P = cumulative strength of all

hoop fibers, lb.

Notice that the material mass densitydoes not appear in the equation. Mater-ials with high mass density (heavy mater-ials) do not optimize energy density,which is contrary to popular intuition.

The mass density does, however, affectthe speed at which the maximum energy isstored.- This speed is given by

rpm (2)

where w = hoop specific weight, lb/ftg = 32.2 ft/sec2

Although there are many advantages tousing super-strong fibers to take advan-tage of equation (1), the resulting highrotational speeds (most high-strength fi-bers are not heavy) pose rotor dynamicsand bearing problems which must be proper-ly appreciated in the preliminary designphases of any modern flywheel. For exam-ple, contemporary electrical motors andgenerators are designed to operate wellbelow the speeds dictated by equation (2)for a Kevlar R wheel.

There are also practical constraintson the dimensions of a flywheel to be usedin the home or on a small farm. Figure 2gives dimension limits for a fiber hoopflywheel.

Fig. 2. Dimension Constraints

Applying the above equations and con-straints to the design of a flexible fly-wheel allows the calculation of hoopsizes, weights and costs for variouschoices of material. Results for Dacron R,steel, and KevlarR are shown in Table 1.

For successful energy storage, themost important parameter is the cost. How-ever, before the low cost advantages ofa flexible flywheel can be realized, thetechnical problem of rotor dynamic sta-bility must be successfully dealt with.

ROTOR DYNAMICS

It was recognized early that the sig-nificant technical problems associatedwith the flexible flywheel would be inthe area of rotor dynamics, specificallythe problem of subsynchronous whirlingdue to internal friction. The advantageof self-balanced operation at high super-critical speeds must be purchased withthe price of suppressing or avoiding aself-excited dynamic instability. Thisis a challenge with a high payoff forsuccess and one which the author believescan be met.

Ever since Jeffcott's analysis ofsynchronous rotor whirl in 1919, rotordynamicists have known that a flexiblerotor displays a "critical speed inver-sion," in which the center of mass comesinside the whirling rotor center atsupercritical speeds.

The flexible flywheel, by virtue ofits low stiffness rotor, always operatesat speeds which are highly supercritical,where Jeffcott showed that synchronouswhirl (due to unbalance) amplitudes areminimized. Experiments to date haveverified that the flexible flywheel pro-duces extremely low levels of synchronousvibration, with no precision balancingrequired.

Not long after Jeffcott's resultsbecame widely known and applied, it waslearned that rotating machinery can be-come dynamically unstable in subsynchron-ous whirl at supercritical speeds, if theratio of internal friction (in rotatingparts) to external damping is high enough.

Early tests of the flexible flywheelshowed subsynchronous whirling which tend-ed to grow with speed and/or time. Figure3 illustrates the mechanism of the inter-nal friction excitation. For subsynchron-

207

ous whirl, the spin speed ft is fasterthan the whirl speed $. As support rope3 moves around to position 1, its rate ofstrain is a maximum at position 2, thusgenerating the friction force F on thehoop which is tangential to the whirlorbit in the forward direction.

Fig. Instantaneous Configuration ofRope Ring

Rotor dynamics theory and analysishas identified several ways of suppressingself-excited subsynchronous whirl. Theyare:

1. Flexible bearing supports.2. Asymmetric bearing support stiff-

ness.3. Bearing support damping.4. Bearing support mass (dynamic

absorber effect).

Before considering how to apply thesemethods to a flexible flywheel., it is use-ful to also look at the other "system"design considerations.

FLEXIBLE FLYWHEEL DESIGN CONSIDERATIONS

1. A shaft seal through the vacuumchamber wall is expensive.

2. A disconnect clutch also increasesthe total system cost.

3. A new motor/generator must bedeveloped to match flywheeltorque-speed characteristics.

4. - Low friction bearings must bedesigned for the application, tooperate in a vacuum environment.

5. The total number of bearingsshould be minimized, for lowestcost.

When these design considerations arecoupled with the design requirements tosuppress the subsynchronous_whirl, a designphilosophy for the flexible flywheel em-erges. The basic elements of this philos-ophy are shown in Table 2.

A DESIGN CONCEPT

Figure 4 shows how the motor/gener-ator can be gimballed on nonintersectlngaxes to provide the low support stiffness,stiffness asymmetry, and bearing supportmass (the motor itself), which are theparameters important to whirl stability.

Since the motor/generator will be putinside the vacuum chamber, a cooling sys-tem is required. The coolant can servedouble duty as the bearing lubricant. Themotor bearings will be designed to supportthe flywheel, thus minimizing the numberof bearings and eliminating the necessityfor a clutch.

Work is now in progress to verifythis concept, both experimentally andanalytically.

A table-top-sized model is being con-structed for preliminary evaluation ofwhirl stability characteristics. Thedistance from the y axis down to the motorcenter of mass is being made adjustable,so as to be able to vary the stiffnessasymmetry.

A mathematical model with four de-grees of freedom is being analyzed to pre-dict the whirl stability characteristics.The characteristic equation (eighth order)will be checked for conditions whichguarantee all real parts of the roots to •be negative.

It is hoped that this two-prongedapproach will yield a stable design with-out resort to a damping mechanism, whichIs often difficult to keep in properadjustment over a long period of time.

Once a stable design has been achiev-ed, the existing larger-scale prototypewill be redesigned correspondingly todemonstrate the potential of home or farmenergy storage using a flexible flywheel.

Fig. 4. Flexible Flywheel onGimballed Support

208

Table 1. Parametric Values for a 10 kw-hr Flexible Flywheel(Safety Factor = 2)

MATERIAL

Dacron(1/4")

Steel(1/2" IWRC)

Kevlar(1500 Den.-"29")

HOOP WEIGHTlbs.

1298

2123

162

MAX. RPM

5,167

4,039

14,606

LENGTH Lft.

1.74

.642

.287

MAT'L COST$

2654

1938

1380

Table 2. Flexible Flywheel Design Philosophy

Design Factors or Contraints

*Shaft seal is expensive

*New motor/generator required

*Clutch increases cost

*New bearings required

*Minimize no. of bearings

*Need low support stiffness

*Need asymmetric support stiffness

*Need support damping and mass

Solution or Conclusion

Put motor inside vacuum chamber

Support flywheel directly frommotor shaft

Gimbal motor/generator onnonintersecting axis

209

PROJECT SUMMARY

Project Title: Conceptual Design of a Flywheel Energy Storage System

Principal Investigator: F. C. Younger

Organization:

Project Goals:

Project Status:

William M. Brobeck & Assoc.1235 Tenth StreetBerkeley, CA 94710(415) 524-8664

The development of a conceptual design for a flywheel energy storagesystem suitable for on-site interfacing with small scale solar energysources. The basic design objective is to provide a generous marginof safety and above average reliability and efficiency at the lowestpractical cost.

A basic concept for a flywheel energy storage system to interfacewith solar energy sources is being studied. This concept utilizesa constant voltage motor/generator directly connected to a flywheelrotor. The field current of the motor is controlled to maintainthe voltage level at the optimum operating voltage of the array ofphotovoltaic cells. When a surplus of electrical energy from thesolar source occurs, the voltage of the array tends to rise abovethe set voltage of the motor/generator, and surplus power drivesthe motor to accelerate the flywheel. When a deficiency in solarenergy occurs, the voltage of the array tends to drop below the setvoltage, and power is drawn from the flywheel via the generator.

A variety of motor types and rotor types have been studied to arriveat a conceptual design to meet the program objectives. A fiber-composite rotor driving a separately-excited dc motor appears tosatisfy the basic requirements. Design details are being examinedto confirm the dynamic stability of the rotor suspension system andto determine the energy losses and system efficiency.

Optimization studies, design layouts, detailed specifications, andcost estimates have not yet been made.

Contract Number: 07-3663

Contract Period: July 1978 - June 1979

Funding Level: $139,053

Funding Source: Sandia Laboratories, Albuquerque

211

CONCEPTUAL DESIGN OF A FLYWHEZL ENERGY STORAGE SYSTEM

Francis C. YoungerWilliam M. Brobeck & Associates

1235 Tenth StreetBerkeley, California 94710

ABSTRACT

A conceptual design of a flywheel energy storage system suitable for on-site inter-facing with small scale solar energy sources is being developed. The basic design ob-jective is to provide a generous margin of safety and above average reliability andefficiency at the lowest practical cost. This paper describes the basic design conceptand the general approach to the conceptual design of the energy storage system. Thebasic concept to interface with solar energy sources utilizes a constant voltage motor/generator directly coupled toa flywheel rotor. The voltage level of the motor ismaintained at the optimum opera+ing voltage of the array of photovoltaic cells. Poweris drawn from the flywheel-driven generator when the voltage output of the array dropsbelow its optimum value and power is delivered to the flywheel via the generator (thenoperating as a motor) when the voltage rises above the optimum value. A variety of motortypes and flywheel rotor types has been studied to arrive at a conceptual design to meetthe program objectives. A fiber-composite rotor driving a separately-excited dc motorappears to satisfy the basic requirements. To obtain acceptable efficiency and low run-down losses, the flywheel operates in a vacuum and has a combination of magnetic thrustsupport and ball bearings to reduce bearing friction in an economical fashion.

INTRODUCTION

The Flywheel Energy Storage System(FESS) is intended to enhance the valueand utility of small solar power generatingplants. The timely development of econo-mical energy storage concepts is crucial inmany solar power applications. The variablenature of sole • energy severely limits itsapplications if some practical storagesystem is not made available. Batteries,pumped-hydro, compressed air, thermal andmechanical energy storage all offer thepotential for suitable systems applica-tions. Economic factors may favor onesystem in a particular application and yetanother in a different application. Thus,it is critical to subject all such systemsto careful examination of performance andcost characteristics. However, at presentsuch an examination is difficult becausean adequate basis for cost and performanceanalyses does not exist for emergingtechnologies. New designs with generousmargins of safety have not yet been demons-trated for mechanical energy storage systems.A concept selection and design programwhich could lead toward a practical demons-tration is presented.

This paper describes an effort forthe development of a conceptual design fora FESS suitable for on-site interfacingwith small scale solar energy sources. Thenew design will have as objectives a gen-erous margin of safety, high reliabilityand efficiency obtained at the lowestpractical cost. The design will be for astationary application, in which no costpremiums are jllowed for minimizing sizeand weight. The basic design objectiveis a system which can be economicallymanufactured in existing industrial facil-ities using conventional production methodsCommercially available components will beused to the maximum practical extent.

Attainment of the highest practicalcharge/discharge efficiency will be a goalof the highe-st level priority. This re-quires very careful attention to a myriadof design details, as well as the selectiorof the most appropriate engineering concepi

BACKGROUND

The flywheel energy storage system foiuse with a small solar energy source mustsatisfy several specific requirements inaddition to being economically attractive.The program objective is the design of

storage units suitable for use with smallto intermediate scale solar sources. Thenominal energy storage capacity is in therange of 5 kWh to 100 kWh; the specificrequirements are for the detailed designof a 10 kWh storage system and for a de-tailed cost estimate for a 50 kWh system.The required power rating of the 10 kWhsystem is 5 kW. These units are to have anelectrical interface for 60 Hz/220 Voltssingle-phase power. The control systemmust maintain the desired frequency andvoltage over the full rotor operating speedrange.

The round-trip efficiency for chargingand discharging must exceed 70%. Thisrequires that the one-way efficiency exceed84%. The rate of loss of stored energy dueto friction and windage must not exceed5% per hour.

The design must have a generous factorof safety and high reliability. It must bedesigned for a life expectancy of at leasttwenty years during which a minimum of 10,000charge/discharge cycles may be encountered.These cycles cover the full speed range ofthe design. Allowance may have to be madefor any overspeed tests which might occasion-ally be performed.

PROJECT DESCRIPTION

The project is divided into four maintasks covering (1) the basic system concept;(2) optimization; (3) design of a 10 kWhunit; and (4) estimate for a 50 kWh unit.

The basic system concept effort andsubsystem options consist of an evaluationof workable engineering concepts and thenarrowing down of choices to the best sub-system options. The product of this workwill be a documentation of the comparativeanalyses and the selection process. Thisis being done in the context of a provi-sional layout of the total system. Sub-system options include flywheel configura-tion, materials, flywheel housing andcontainment, seals, type of motor/generator,suspension system, mounting, controls,auxiliaries, and instrumentation. Problemsof safety, efficiency, losses and parasiticloads, manufacturability, maintenance, andeconomics are being identified and solved.

Various flywheel configurations andmaterial combinations have been consideredfor energy storage systems. It has beenshown that high energy density is potentiallyachievable with materials possessing a high

strength-to-we^ght ratio. In many engineer-ing applications light weight is veryimportant and a premium may be paid for amaterial with a high strength-to-weightratio. However, for a utility application,additional expense merely to save weightwould be unreasonable. The appropriatedesign criterion is maximum energy atminimum cost. Fiber-composite using E-glassappears to satisfy this criterion.

CONCEPTUAL DESIGN

The candidate flywheel energy storagesystem is shown in Fig. 1. It consistsof a filament-wound fiber-composite fly-wheel directly connected to a drive motorwhich acts as a generator during dischargeof the system. The flywheel is containedin an evacuated enclosure to keep the aero-dynamic drag low. The enclosure also providesa means of support and containment forfracture debris in case of failure.

The flywheel (Ref. 1) shown representsa fiber composite concept which has been greatlyde-rated to achieve a high factor of safety.The energy density and resulting stressesare but a fraction of that which m.ght beobtained at some future time when fiber-composite flywheels are more fully developed.

The vacuum requirements are being setby trade-off studies where the benefitsassociated with low aerodynamic drag areevaluated relative to the added cost,complexities and power consumption of thevacuum pump. These studies show that a hardvacuum is not required. A vacuum of 10"'*Torr is quite adequate.

The drive motor acts to accelerate theflywheel during charging and acts as agenerator to decelerate the flywheel duringdischarge. A variety of options forcontrolling the power flow to and from amotor are available. These must take intoaccount the fluctuating nature of the load,the output characteristics of the solargenerator, and the speed variations of theflywheel.

The selection of a motor type isdependent upon the power control method toa large extent but is also dependent uponother factors such as the operating speedrange, voltage range, and operating envi-ronment.

For the conceptual design shown inFig. 1, a range of sizes and speeds isrequired to satisfy the range of energy

213

storage desired. Table 1 shows the sizesand ratings for the various components of acandidate design for a 10 kWh demonstrationmodel and for a 50 kWh unit.

MATERIAL PERFORMANCE, COST ANALYSIS ANDOPTIMIZATION

For a given configuration, the energydensity in Wh/kg is proportional to thestrength-to-weight ratio of the rotormaterial; thus, the energy storage perdollar of material is directly proportionalto the strength-to-weight ratio divided bythe cost per pound. The cost of the vacuumenclosure is dependent upon the size of therotor which, fora given energy storage andconfiguration, is dependent upon the absolutestrength of the rotor material. The sizeof the rotor rather than its weight is aprimary factor influencing the fabricationcost. The heavier material, in additionto permitting the use of a small flywheelfor a given energy, may also allow a lowerrotational speed which, in turn, may yieldlower friction and windage losses. Theoverall effect of rotor material propertiesand cost upon tne complete system cost arebeing examined for a complete system designin which the cost of the enclosure, bearings,and motor are included. These costs dependto a considerable extent upon the propertieso'- the rotor, principally rotor shape andmaterial.

Of prime importance in the selectionof flywheel materials is high strengthbecause the energy which can be stored perunit of volume is directly proportional tothe allowable stress. A survey of flywheelprojects in progress shows that there havebeon no complete life cycle tests and eval-uation of high-strength high-energy densityflywheels to date.

Several types of flywheels using differ-ent types of materials are being examinedincluding concepts based on isotropicmaterials (metals) and orthotropic materials(fiber-composites). Analysis of perform-ance and total costs and projections ofreliability and safety to personnel arebeing considered. At this time, it appearsthat a fiber-composite flywheel is the bestchoice.

Table 2 shows the properties of severalcandidate fiber composite materials, ofthese, the E-glass is the least expensive.Although it has the lowest strength, theratio of strength to cost makes it the mostattractive candidate. Fatigue test on

E-glass shows a significant loss in strengthdue to static fatigue. Much of this lossis believed to result from the chemicalaction of moisture in the air which maybe prevented by exclusion of moisture orby operating in a vacuum. Long-term spintesting in a vacuum is one way to determinethe static fatigue limit for E-glass. Untilsuch testing is completed, the only safeapproach is to allow an adequate reductionin the design stress.

The Kevlar 49 compusite materials havea higher specific strength than the fiber-glass materials and do show less sensiti-vity to static fatigue; however, thesecomposite materials are much more expensive.In spite of the higher cost of Kevlar 49a limited use of such a material may pro-vide significant design advantages. Thehigh modulus of elasticity and low densityof Kevlar 49 makes it an ideal candidatematerial for use as an overwrap to reducethe radial tensile stress in a thick-walledring by creating a radial compression atthe interface.

A biannualte rim, shown in Fig. 2,consisting of an inner ring of E-glass/epoxy with an overwrapped outer ring ofKevlar 49 can permit a considerable in-crease in radial thickness for the completerim without giving excessive radial tensilestress. This large radial thickness alsogives good space utilization. The ratioof Kevlar to fiberglass in the biannulaterim will be determined by design studies.Increasing the amount of Kevlar may increasethe cost faster than it increases the energystorage. It appears from preliminary cal-culations that only a small amount ofKevlar is required to give an optimumdesign.

Increasing the amount of Kevlar in therim at the expense of the fiberglass hascertain disadvantages. In the limit of anall-Kevlar rim, the maximum radial thick-ness must be quite small to prevent excessiveradial tension. The utilization of spacewould then be poor. For a given energy,the flywheel would be much larger and wouldhave a higher rim speed. Consequently,the aerodynamic drag would be higher andthe vacuum enclosure larger. The Kevlaris a more expensive fiber and the largerenclosure for a Kevlar flywheel would alsobe more expensive.

214-

FLYWHEEL HOUSING AND CONTAINMENT

This component serves a number of im-portant functions:

• An evacuated enclosure to reducewindage losses.

• An encasement to prevent corrosionand mechanical damage to theflywheel.

• A structural member to transmitloads.

A structure for mountingbearings and seals.

the

• A means of containment in theevent of failure of the wheelor bearings.*

For a demonstration unit, the housingcould be a steel or aluminum alloy weld-ment. The housing for a production version"baseline" flywheel module is expected tobe a casting. A steel casting will be aprime candidate. Another housing candidate,which will be considered in view of itspotential for low costs, is a fiberglasscomposite material.

The highest loads which must be trans-mitted through the housing are thoseassociated with a flywheel failure of sometype. High torques and high internalradial loads are expected if failure occurs.

The flywheel energy storage systemhas a large energy and momentum storage.A sudden and uncontrolled release of thisenergy and momentum would have seriousconsequences, unless, by careful design,this energy can be released as heat. Thereduction in angular momentum will producea torque proportional to the rate of changein angular momentum.

For a flywheel with adequate inertia tostore 10 kWh of energy, the torque requiredto remove all of its angular momentum in onesecond would be in excess of 68,000 lb-ft.This is an enormous amount of torque andquite likely is greater than the flywheelmounting structure could support. Itseems clear that efforts to limit the rateof change of momentum are required.

*Placing the FESS in an underground con-crete-lined pit may provide the most cost-effective containment means.

The design of the system should be suchthat in the event of flywheel failure onlya small fraction of the system momentumand energy must be rapidly dissipated.The larger remaining fraction can bedissipated very gradually.

Two controlled braking methods arebeing considered. One will provide acontrolled deceleration of fracture debrisrrom the failure of the flywheel rim andthe other will simultaneously provide avery much slower deceleration of the fly-wheel hub. The first of these brakingsystems consists of a containment ringsurrounding the flywheel rim. This ringis permitted to rotate within the vacuumchamber with substantial frictional re-sistance to its rotation being used toprovide braking. Fracture debris from arim failure would impinge upon this con-tainment ring which would be strong enoughto hold the failed fragments and keep themrotating around on its inner surface. Thefracture debris and the containment ringmay both be rotating within the vacuumchamber until the friction eventuallybrings them to e. stop.

The second mode for stopping the hubprovides aerodynamic braking by permittingthe pressure in the vacuum chamber to rise.This braking is very gradual and will notresult in excessive torque or excessivetemperature rise.

The containment ring must be strongenough to resist the initial impact of thefracture debris and the continued centrifugalforce of the rotating debris as its speedgradually reduces. Analysis shows thatthe required strength is related to thekinetic energy of the fracture debris.The relationship for the ratio of energycontainment to weight of containment ringis the same as that for a rotating ring.That is:

£ _5-W " 2w

where E = contained energyW = weight of containment ringa = allowable stress in the ringw = density of containment ring

The importance of the strength-to-weight ratio of the containment ring isshown in this equation. It seems clearthat a high strength-to-weight ratioof the material may be just as importanthere as it is in the flywheel itself. A

215

high strength-to-weight ratio fiber-compo-site could be used for the containmentring. A composite of E-glass may be moreeconomical than high strength steel forthis application.

ENERGY LOSSES

Table 3 lists the categories of lossesand parasitic loads of an electrically-coupled flywheel energy storage system.

Table 3.Loads.

System Losses and Parasitic

Flywheel Unit Losses:

Flywheel Bearing LossesFlywheel WindageVacuum Seal FrictionVacuum Pump PowerMotor WindageMotor Bearing LossMotor Brush Friction*Motor ExcitationMotor Iron LossMotor Armature Copper LossLubrication Pumping Power

Power Conversion Losses:

Semi-Conductor LossesOther Component LossesControl PowerAuxiliary Conversion Losses

*If brushes are used.

Flywheel Unit Mechanical Losses. Thefriction of the flywheel ball bearings isbeing estimated from the bearing manufac-turer's test data and from tests of fly-wheels of comparable design.

For the flywheel candidate design,the ball bearing loss will be 80 to 120watts for the 10 kWh unit at 10,000 rpm.

Windage loss of the flywheel properdepends, for a given geometry and speed,on the pressure and composition of the gasin which the flywheel runs. The dependenceof windage loss upon pressure and flow regimehas been calculated and used to estimatethe aerodynamic drag on the flywheel.

In the pressure range in which theflywheel operates, the mean free path ofthe molecules of gas is long compared tothe wheel clearance in its housing, andthe windage loss is proportional to pressure.

It is desirable to keep the pressure below10"° atmosphere where loss is around 30watts. If the pressure is allowed to riseto 10-4 atm. the loss become excessive.To maintain a pressure below 10"° atm.will require a mechanical pump.

Vacuum seal friction is due to therubbing contact between the fixed androtating faces of the seal. This loss willvary from 50 watts to 100 watts over thespeed range. The bearing lubricating oilwill also lubricate and cool the seal. Theseal will be designed to balance most of theatmospheric load so as to limit the rubbingforce to a low value.

MOTOR SELECTION STUDY

A successful motor application dependsprimarily on selecting a motor that satisfies,as nearly as possible, the kinetic require-ments of the driven machine without exceedingthe temperature or torque limitations ofthe motor. The first step in motor selectionis to determine the load characteristics-power, torque, speed, and duty cycle.

In this application the motor drivesthe flywheel shaft to accelerate duringthe charging time and acts as a generatorduring discharging time. The motor mustprovide good efficiency in both modes ofoperation. The most widely used motorsystem which provides good efficiency asa motor and as a generator is a dc motor.However, ac motors can also be used as acgenerators but the electrical controlequipment is costly and will complicatethe system. An economic study is beingconsidered to test the feasibility ofboth systems.

The schematic diagram (Fig. 3) illus-trates the different modes of operation.

The functional schematic diagramshown represents the block diagram of theFESS. The different blocks shown in thediagram are as follows:

i) Flywheel; which is mechanicallyconnected to a motor system.

ii) Solar input; which has anoutput of 250 Vdc ±50V.

iii) Power flow controller; whichcontrols the power flow inand out of the motor systemand to the load.

216

iv) Power conditioner; whichinverts the power from dcor ac with any frequency toexactly 60 Hz/220V singlephase ac.

v) Load; which can take power at60 H2/220V single phase.

There are four modes of operation:

1. Only the solar source isfeeding the load.

2. The solar source and theflywheel both are feedingthe load.

3. The solar source is feedingthe load and charging theflywheel.

4. Only the flywheel is feeding theload.

Note that in Modes 2 and 4, where theflywheel is discharging and feeding theload, the motor is acting as a generator.

The motor has a dual function, i.e.,it acts as a motor when it is chargingthe flywheel and it acts as a generatorwhen feeding the load during discharging.Separately-excited dc motors are easilyswitched from the motor configuration tothe generator configuration and can handlethe speed range for this application. Anac type motor can also be used. However,since the electrical frequency will varyas the speed of the flywheel changes, therewill be a need for electrical power andfrequency control devices to convert thehigh and variable frequency of flywheelsource to the low frequency of the load.

A brush!ess dc motor is very attrac-tive because lower losses and higherreliability are expected.

There are several ways to reduce powerlosses in the motor. One is to reduce lossesin the core, either by adding more materialto the magnetic core structure or by usinga steel with improved core-loss propertiesand thinner lamination. Another method isto increase the cross-sectional area ofconductors to reduce resistance. Anotheralternative is to shorten the air gapto reduce the magnetizing current required.

Losses calculated for motors have beendivided into load-dependent and load-independent losses. The latter determinethe standby power required by the flywheeland is considered separately from the loaddependent losses. The eddy current andhysteresis losses are dependent upon fieldexcitation of wound-field machines. Wehave assumed that the field current willbe turned-off during standby to reducerun-down losses, similar to the design ofRefs. 2 and 3.

CONTROL OF POWER FLOW

Generally, if the solar power genera-tion exceeds the load power requirement,the excess power will be diverted to theflywheel; and if the solar power generationis less than the load requirement, thedeficiency will be made up from the fly-wheel .

Generation of solar power is due tothe photovoltaic effort (PVE) which occursin semiconductors. Charge separationin the photovoltaic cell will cause anelectrostatic potential difference acrossits P-N junction, and by placing cellsin parallel or series, we can obtain therequired voltage or current for a specificload. Four parameters are of interest inanalyzing the performance of a photovol-taic cell. They are the short circuitcurrent (I,-); the open circuit voltage(V ); the current for matched load, i.e.,current under maximum power transfer(operating point) conditions (ImD) andthe corresponding voltage (V _ ) .

Figure 4 shows typical current-voltage characteristics of a P-N junctionsolar cell array under a given amount ofsunlight. Since the maximum power outputoccurs when the current and voltage is at^ D and nro' °Perating at this point makesavailable pthe maximum .power. The controlsystem should, to the extent possible, keepthe solar output voltage at V . If theload demand exceeds the solar poutputeither because there is inadequate sunlightor because there may be a high load condi-tion, then the voltage would start to dropbelow Vmp. To prevent this voltage dropadditional power must be supplied by theflywheel energy storage unit. On theother hand, if the load demand is lowerthan the solar output, then the voltagewould rise above V . To prevent thisvoltage rise the pexcess power must beused to recharge the flywheel. As examples,consider the case where the sunlight is such

217

that the maximum output is 3kW and V is200 Vdc. If the load demand is 8 kW^thecontrol system would bring the flywheeloutput to 5 kW to keep the solar outputvoltage at V . If the load demand is 1 kW,the control psystem would divert 2 kW torecharge the flywheel to keep the solaroutput voltage at V .

During normal operation the sunlightfluctuates going to zero at night, alsothe load fluctuates making abrupt changesas electrical equipment is turned on andoff. As these fluctuations in output anddemand occur, the control system sensesthe changes in voltage and regulates themotor output or input to discharge orrecharge the flywheel unit and maintainsthe voltage. In the design example witha separately-excited dc motor, the fieldexcitation would be raised or lowered bythe control system in order to keep thedesired voltage. The usual dynamic prob-lems of feedback control systems will besolved in the conventional way.

For other types of motors, the controlproblems are similar. The control systemwill regulate the motor output or inputpower of the flywheel unit to keep thevoltage output of the solar array at V .

If the flywheel is completely chargedor if it is discharged, then the controlsystem will not be able to regulate thevoltage. In this case, the voltage mustbe allowed to rise or fall. If the FESSis too small, excess power from the solarunit will be lost after the FESS is fullycharged.

The state of charge/or discharge ofthe flywheel is simply determined by itsspeed. When the flywheel is being chargedfrom the excess power of the solar unit, anoverspeed sensor can be used to shut offthe flow of charging power. Similarly,when the flywheel is being discharged, theunder speed sensor (or inadequate voltage)can be used to shut off the output power ofthe motor. However, the normal operationshould be such that the flywheel is dis-charged only to its minimum rated speed,where the motor can no longer generatethe required output voltage at maximumfield excitation.

SUSPENSION SYSTEM

The suspension systems being consideredprovide sufficient flexibility to permitthe flywheel to rotate about its own masscenter. The total deflection needed tosatisfy this requirement can be quitesmall and is dependent upon the unbalanceof the flywheel. The mode of operationwhich permits this deflection of the masscenter involves rotational speeds greatlyin excess of the first critical speed.Operation at such speeds would be unstableif special precautions to insure stabilityare not taken. Means of preventing whirlinstability are known and will be incor-porated into the suspension system (Ref. 4).

A variety of suspension systems arepossible. It is difficult to prove thatone system has an overwhelming advantage overall others. Our experience (Refs, 5 & 6) with apendulous mounting shows that this systemis acceptable when precision ball bearingsare used. A stradle mounted system may bepreferable where magnetic bearings are used.

The analysis of run down losses indi-cates that some magnetic support to reducebearing loads is required. Thus, asuspension system using a combination ofmagnetic and ball bearings will be designed.Such a system will have magnetic supportabove the flywheel which will be rotatedabout a vertical axis. The magneticsupport will be passive. The electricalpower requirements for such a bearingappear, from preliminary calculations, tobe very small. Of course, the power mustbe smaller than the frictional loss ofconventional bearings in order to justifythe use of the magnetic support.

REFERENCES

1. Younger, F. C., "Tension-BalancedSpokes for Fiber-Composite FlywheelRims," 1977 Flywheel TechnologySymposium, CONF-771053, March 1978.

2. Brobeck, W. M., "Design Study for aFlywheel-Electric Car," IEEE 28thVehicular Technology Conference,Denver, Colorado, March 22-24, 1978.

3. "The Application of Flywheel EnergyStorage to Electric PassengerAutomobiles," William M. Brobeck &Associates, Report No. 8600-28-R1,February 1977.

218

4. Thomson, W. T., Younger, F. C , andGordon, H. S., "Whirl Stability ofthe Pendulously Supported FlywheelSystem," Journal of Applied Mechanics,Vol. 99, No.s, June 1977.

5. Brobeck, W. M., "Flywheel EnergyStorage for Utility Applications,"IEEE Power Engineering Society SummerMeeting, Paper No. A 77 652-1.

6. Brobeck, W. M., "Flywheel Developmentfor the Electric Power ResearchInstitute," 1977 Flywheel TechnologySymposium, CONF-771O53, March 1978.

219

MOTOR-GENERATOR

Figure 1 . Candidate Flywheel Energy Storage Unit

220

Figure 2. Candidate Fiber-Composite Flywheel Rotor

221

SUNLIGHT

FLYWHEEL

/

to

SOLARINPUT

POWERFLOW

CONTROLLER

MOTOR/GENERATOR

POWERCONDITIONER si

O <OgCMCM

Figure 3. Functional Schematic Diagram

Figure 4. V-I Characteristic of Photovoltaic Device

223

Table 1. Characteristics of a Candidate Flywheel EnergyStorage System.

Energy, kWha

Power, kW

Speed, rpm

Flywheel Diameter, in

Flywheel Height, in

Flywheel Weight, lbs

Motor Type

Converter

Efficiency - round-trip

Run Down Losses

Output

10 kWhDemonstration

Unit

10

5

10,000

50

13

613

dc separatelyexcited

dc to acinverter

71%

4«/hr

60 Hz/220 V

50 kWhUnit

50

10

5,850

8505.

22.2

3,065

dc separatelyexcited

dc to acinverter

72%

4%/hr

60 Hz/220 V

Nominal Energy Rating

ac to dc rectifier may also be needed if utility power isused for recharge

224

Table 2. Fiber Composite Properties

rotoui

Fiber:Vol% Fiber:

Density:

Mechanical Properties

Elastic Constants:Longitudinal Young's Modulus, E-|iTransverse Young's Modulus, E22Shear Modulus, G]2Major Poisson's Ratio, u-j2Minor Poisson's Ratio, U21

U l t i mates:Longitudinal StrengthLongitudinal Ultimate StrainTransverse StrengthTransverse Ultimate Strain

Longitudinal StrengthLongitudinal Ultimate StrainTransverse StrengthTransverse Ultimate StrainShear Stress at 0.2% offsetShear Strain at 0.2% offset

E-Glass (Owen-Corning)Nominally 65 vol%

2.1

52.1514.036.30.2070.056

gm/cm3

t 0.89 GPat 0;61 GPat 0.5 GPat 0.016i 0.011

Tension1108

(2.167.50.054

t 25 MPat 0.11%)t 1.1 MPat 0.009%

Compression530

(1.11 :(78

(0.68 :22.4 :0.546 :

t l lO MPat 0.27%)t 4 MPa)t 0.10%)t 1.7 MPat 0.045%

S2-G1ass (Owens-Corning)Nominally 60 vol%

56.215.77.410.2820.079

2.08

± 2 . 7 GPa± 1.0 GPa± 0.56 GPa± 0.031± 0.014

Tension1615

~ 241.00.292

± 127

± 6.3± 0.064

Compression460

0.92111.8

2.9330.40.620

± 60 MPa± 0.06%± 2.2 MPa± 0.32%± 0.98± 0.041

Kevlar 49 1420 denierNominally 60 vol%

1

81.85.101.820.3100.0193

.38

± 1.5 GPa± 0.10 GPa± 0.09 GPa± 0.035± 0.0014

Tension1850

2.237.90.161

±50MPa± 0.06%± 1.1 MPa± 0.023%

Compression(235

(0.48(53(1.4124.41.55

± 3 MPa± 0.3%)± 3 MPa)± 0.12%)±-2.4 MPa± 0.16%

Matr ix: 100 parts Dow Chemical DER 332 (bisphenol-A epoxy res in ) , 45 parts Jefferson Chemical Jeffamine T403(polyether triamine)

Cure: 16 h @ 60°C (E-Glass & S2-Glass)Cure: One day at room temperature and 16 h at 85°C (Kevlar)

Reference: L. L. Clements and R. L. Moore, Lawrence Livennore Laboratory, Report UCRL-79262 (1977.

PROJECT SUMMARY

Project Title: Residential Flywheel with Turbine Supply

Principal Investigator: T; W. Place

Organization: Garrett AiResearch2525 W. 190th StreetTorrance, CA 90509

Project Goals: Develop the conceptual design of a cost effective flywheel energystorage system that interfaces with a wind turbine energy source.The system is to be sized for residential use. Design emphasisis to be on the use of current rather than future technology.Accordingly, the manufacturing methods are to involve conventionalprocessing to the maximum practical extent.

Project Status: A literature search was conducted for evaluating promising systemconcepts. This has culminated in the selection of a baselinesystem concept for analysis. This concept consists of five majorsubsystems. These are a gearbox, generator, variable speedtransmission, flywheel rotor, and controls. The input to thesystem is mechanical shaft power from a vertical axis windturbine. The system output will be based on the load demands ofan all-electric single family residence with a floor area of139 m 2 {1500 ft 2). Time varying input power and electrical loaddata have been furnished in card-deck form by Sandia and will beused as guidance in evaluating the conceptual system.

Trade-off criteria were established for the comparative ratingof subsystem options, and several variations have been selectedfor each subsystem for cost and performance evaluation.

A computer code is being written for analyzing the baselinesystem with various combinations of subsystem options. Theseanalyses are scheduled for completion at the end of November.

Contract Number: 07-9093

Contract Period: July 1978 - July 1979

Funding Level: $135,899

Funding Source: Sandia Laboratories, Albuquerque

227

RESIDENTIAL FLYWHEEL WITH TURBINE SUPPLY

Theodore W. PlaceAiResearch Manufacturing Company of California

A Division of The Garrett Corporation2525 W. 190th Street, Torrance, California 90509

ABSTRACT

This paper examines a flywheel system that stores energy from a wind turbine sourceand converts the energy to 60-Hz, 220-v output for residential use. The typical residencehas a 1500-sq ft floor area with a maximum power level of 5 kw.

The purpose of this study is to determine the cost/benefits of storing wind energyin a flywheel and using it on a demand basis. The study examines the systems and theflywheel rotor materials that offer the greatest promise for reducing the initial cost.

The paper presents the progress to date on th i s program and 'Jescr i bes the workplanned to complete the study.

INTRODUCTION

The vertical wind or Darrieus turbineconverts wind energy to mechanical energy.The mechanical energy is stored in a fly-wheel for use as electrical power in a1500-sq ft house. Calculations have shownthat the typical residential demand in anarea such as Albuquerque, New Mexico, is2.4 kw, and that the demand slowly undu-lates as a function of time. The firststep in applying the stored energy is tochange it to electrical power and use iton a demand basis in the home.

Tests conducted on the wind turbinehave shown that the wind energy is inter-mittent and that the available energy canbe much greater than demanded by residen-tial electric use. One solution is tostore this pulse of energy in a flywheelthat has been sized to accept the avail-able wind energy, and to use it to supple-ment the utility power. These sizingstudies will be conducted to determinethe system effectiveness.

PROJECT GOALS

The project goal is to identify acost-effective meciianical storage device.In addition, the device must be safe andreliable for 20 years. The study con-straints include a 5-kw output generatorand a 10-kwhr flywheel system. Themechanical storage system is to usecurrent technology rather than future

technology. The method of manufacturing isto be conventional processing. The studyis expected to stimulate industry to pro-duce and market a mechanical energy storagesystem that will produce a significantreduction in the use of utility power. Theenergy storage system will also have anacquistion cost that produces a net savingsin life-cycle costs.

BACKGROUND

In 1925, G.J. Darrieus,1'2 of Paris,France, applied for a U.S. patent for anew type of windmill to generate power.The vertical axis turbine shown in Fig. 1has several blades of symmetric airfoil incross section, and is curved in the shapethat a flexible cable of uniform densityand cross section would assume if spunabout a vertical axis. The blade shapehas been designated troposkien. The advan-tages of the Darrieus turbine over a con-ventional propeller type wind turbine areits ability to accept wind from any direc-tion, its lower and simpler constructioncosts (since its qenerator is located onthe around), and the resulting lowermaintenance.

Several tests have been run in which themeteorological data were collected at vari-ous test sites and used to calculate bothsupply (wind velocity) and demand (solarinsulation, air temperature, and windvelocity). These data are illustrated inFig. 2 for the Blue Hill test site inMassachusetts.

228-3CZT

AIRFOIL SECTION

Fig. 1. Vertical axis wind turbineshowing modified troposkien configuration.

INPUT

TUES WED THRU

Fig. 2. Wind energy (input) versusresidential demand (output).

The present study will use data for a 365-day year for both a minimum wind site suchas Albuquerque, New Mexico, and a maximumwind site such as Blue Hill, Maine.

Flywheels have been constructed andtested that have been made either of iso-tropic material such as steel or of com-posite material such as fiberglass or

kevlar. These devices have been construc-ted for applications such as subway railvehicles or passenger vehicles. These datawill be used in the study.

PROGRAM PLAN

The program is divided into four func-tional tasks plus a documentation task.These are shown in Table 1.

The first task is to select the basicsystem and to establish the component tradeoptions. The second task will evaluateperformance, cost, and reliability. Thiswill be done by evaluating selected com-ponents in the baseline system, using acomputer study. The test data collected atBh.e Hi I I and Albuquerque w!! I be input.Task III will provide a specification andestimate of a 10-kwhr system. In Task IV,a 50-kwhr system will be extrapolated andest i mated.

PROGRAM SCHEDULEWORK DESCRIPTION

IK I1.0 Of V I I I SASIC SVSTEMi.o SELECT IUSSYSTCM

OPTION!:I* . FLYWHEEL ENEROV

STORAflE UNITU.I I ICTI I ICAL MOTOR/

GENERATOR2c. THANSMII DON2*. CONTROL!TAIK II

.0 UTUP COftT-MRFORMANCEMOO1L

!.Q ESTIMATE PERFORMANCE3.0 ESTIMATE SAFETY AND

RELIABILITYIASK III

ULYOUT AftD SPECSnoKWiSELECT KANOrACTURERBCOST ESTIMATEANNUALIZE COST

TASK IVEXTRAPOLATION TO ID KWESTIMATE COiTANNUALtZCD COST

At REPORTSTATUS i

Table 1. Schedule for Sandia flywheelenergy storage system study.

PROGRAM METHODOLOGY

The method that has been chosen forthe study is as follows. First, a litera-ture search has been accomplished in orderto evaluate both the baseline system andthe components. Second, a baseline systemwill be selected and preliminary sorting of

components will be estimated. The assump-tions will be defined along with thecharging and discharging cycle.

A performance and cost computer modelwill be established for the baseline system.Selected components will be Inserted intothe program, and the performance and costdata wilI be evaluated for input wind and•electrical demand data. The baseline sys-tem with selected components will be traded

229

off to optimize performance in terms ofcost. The model will then be evaluated forsafety and reliability.

The 10-kwhr system preliminary designwill be finalized and defined by cross sec-tion drawings and specifications. Finally,a 50-kwhr system will be extrapolated. Theresults will be documented in a final report.

SYSTEM DESCRIPTION

The mechanical energy storage systemfor a residential application consists offive major components. These are a gear-box, a generator system, a variable speedtransmission, a flywheel storage system,and controls. A typical system is illus-trated in Fig. 3. In the system analysis,candidates for each of the above componentswill be substituted into the performanceand cost analysis to select the system.

DARKMINDTUHB

EUS

NE

60 HZ220 V

CLUTCH

|/\J

JOTOR/,

tVOLT.REG

BOX

FLYWHEELCONTROLLER

SPEEDTRANS-MISSION _

Fig. 3. System description with ac generator.

The input to the system is mechanicalshaft power supplied at either a constant12t>- or 75-rpm speed, depending on the testlocation of the wind turbine. The windpower has been supplied for a worst caseand best case in the form of test data!supplied by Sandia Laboratories.

The output of the system (electricaldemand) has been calculated assuming an all-electric house with a floor area of 1500sq ft. These data have also been suppliedon a card deck which describes the yearlydemand of the residence for 60-Hz, 200-vpower.

COMPONENT DESCRIPTION

The selection of components wilI beaccomplished in two steps. The first stepis to select candidate components anddetermine their characteristics.

FLYWHEEL

The flywheel will have the followingvariables:

(a) Shape

(b) Material

(c) Vacuum

(d) Bearings

(e) Seal

Flywheel Shape. The variation of shapes'for the steel or isotropic material fly-wheel will be limited to a truncated coni"-cal disc, solid disc, and pierced disc.The composite material will use a concen-tric ring approach. Fig. 4 illustrates theshape factor K s that is used in calculatingenergy density. The second variable forconsideration is the length-to-diameterratio. The selected minimum diameter is2.0 ft and the selected maximum diameteris 6.0 ft.

1

2

3

TRUNCATEDrnmrAi ^~-~DISC " - 5 ^ ;TAPER RATING = .3

msr. 1f—

CONCENTRIC CYLINDERS J(Al RESEARCH) ~|

-A I—i.c

1 1o. o.

1

I

0.8

0.6

0.3

0.3

Fig. 4. Flywheel rotor candidate shapes.

Flywheel Material. The available materialfor constructing flywheels is listed inTable 2. The table shows the specificenergy per pound of flywheel material.Considering present-day technology anddemonstrated success, several candidateswere selected for preliminary cost analy-sis. Those candidates marked with anasterisk will be examined in the systemperformance and cost model. Table 3 showsthe flywheel rotor cost comparisons.

230

Table 2. Flywheel material candidates.

Mater i a 1

*High-strength steel 4340

•Maragtng steel 18 Ni + 300

*High-strength loadalloy 100 AR

Aluminum 7075

Tivanillin 6A1 4V

Depleted uranium

*E-glass-epoxy andaluminum hub

S-glass-epoxy andaluminum hub

Kevlar 49 and epoxyand aluminum hub

Plywood and birch

DensityIb/cu in.

0.283

0.289

0.283

0.101

0.160

0.683

0.080

0.079

0.052

0.022

Allowablestress m

140,000

150,000

too,ooo

26,000

70,000

25,000

120,000

130,000

170,000

8,000

Energy density versus ;

0.8 0.6

12.4

13.0

8.8

6.5

11.0

0.9

9.3

9.8

6.6

4.9

8.2

0.7

6.8

;hape factor

0.3

4.7

4.9

3.4

2.4

4.1

0.3

14.0

15.5

30.7

3.4

^Selected for system evaluation!

Table 3. Flywheel rotor comparison (preliminary)

FlywheelMaterial

Hiqh-strenqthsteel 4140

Maraginq steel18 NI -300

High-strengthlow al loy 100 AR

E-g1ass-epoxyand aluminum hub

E-g1ass-epoxyand steel hub

Kevlar 49 andepoxy andaluminum hub

E w-hr

12.4

13.0

6.6

14.0

12.6

30.7

Labor

0.25

0.25

0.36

0.84

0.84

0.84

Costs

Too 1i ng

0.10

0.10

0.05

0.05

0.05

0.05

dollars/1b

Unuseable

0.07

0.10

0.07

0.01

0.01

0.01

Mater i a 1

1.32

1.92

0.32

0.87

0.72

6.75

Cost per kwhr,dollars

140

182

121

126

128

249

231

Vacuum for Flywheel. The fIywheeI systemwindage losses will be insignificant if thecavity pressure is maintained between 1 and10 microns. Fig. 5 shows that the loss isinsensitive to the lei.^th-to-diameter ratio.The variables for the study will then be

the cost impact of a hermetic sealed,magnetic coupled flywheel versus a sealedshaft, and the related pump down equipment.

between the 1200-rpm gearbox and the fly-wheel, and since the design horsepoweris approximated 26 hp, then it appearsthat either the traction or the variable-pulley rubber belt would satisfy thedesign. The relative costs chart listedin Fig. 7 shows the variable-pulley rubberbe 11 to be most attract i ve on a f i rst-costbasis. These two selections will becarried on for final analysis.

The second part'of the transmissiontradeoff is the cost of a transmission.The transmission will vary from zero tofu 11 speed versus a speed change of 3 to1, plus the special motor that must bedesigned to tolerate the heat and largeslip angle.

Fig. 8 shows the candidates for zero-speed versus the straight 3-to-1 belt system.

100,000

10,000

Fig. 5. Effect of length/diameter ratio versus a.windage losses.

Bearings for the Flywheel System. The ucandidate bearings'* for the performance and |cost analysis will be ball bearings, roller °bearings, and hydrostatic bearings. The §evaluated performance and cost will be the £criteria for selection. The magnetic bear- 3ings have been eliminated since the loss is j|low for the lower cost bearings.

Seals for Flywheel System. The seals4 forthe flywheel system that have been selectedas candidates are the carbon seal, and thehermetic sealed/magnetic coupled unit.The ferro fluid seals have been eliminatedbecause of hiqh losses for the large-diameter seal. The life-cycle cost will beevaluated as part of the vacuum systemanalysis.

Structure for Flywheel System. The fIy-wheel containment housing4 must be sealedfor vacuum. The candidates for cost evalu-ation are a hermetic sealed metal container,a metal and concrete container, and a metalcontainer with earth embankment or pit.The vertical axis and the earth orientedaxis will be examined for the flywheel.TRANSMISSION

The purpose of the transmission is toprovide a variable speed input to the fly-wheel system so that mechanical energy maybe stored and extracted as a resuIt of aspeed change of the flywheel. The trans-mission type was first selected between thecandidates listed in Fig. 6. Since thetransmission is placed in the sysTem

1,000 10,000 100,000

RATED OUTPUT SPEED, RPM H m 7

Fig. 6. Design rationale for selectionof transmission options.

100,000

~ 10,000

1,000

100

FLUID

METAL-CHAIN ANDWOOD-BLOCK BELT

1 10 100 1,000 JO, 000

RATED POWER CAPACITY, HP M I 7 M

Fig. 7. Cost rationale for selectionof transmission options.

232

MOTOR/GENERATOR

The main function of the motor/generator (M/G) is to provide a.nominal220 rms (110-v, line-to-neutral, single-phase) at a nominal 60 Hz. In the motormode it can be used to start the wind tur-bine, using the utility as a power source.

The M/G could also be applied as avelocity governor for the wind turbine ifit is referenced to the utility to maintainan accurate 60-Hz frequency. If it is notreferenced to utility frequency, the fly-wheel and CVT must be controlled to providethe frequency required. The frequencywould then be accurate enough for allhousehold loads except synchronous clocks.

The three M/G types that are commer-cial I y available as well as technicallyacceptable are:

(a) Salient pole synchronous alternator

(b) Squirrel cage induction alternator

(c) Separately excited dc generator

The salient pole synchronous alter-nator is the most easily controlled of thethree candidates. With slip rings to trans-fer dc field current to the rotor, voltageregulation can be readily accomplished. Itcannot control suddenly applied overloadssince it will lose synchronism if pullouttorque is exceeded. Amortisseur windingsare required if it is to be used as astarting motor.

TRANSMISSION TRADEOFFS

! i 1

TOROIDALVARIABLEBELT AND

SPEEDINCREASER

VARIABLE BELTANP SPEEDINCREASERWITH FLUIDCOUPLING

VARIABLEBELT AND

DIFFERENTIALGEAR DRIVE

alternator, it is less stable under vari-able loads and it operates at a lower powerfactor.

The separately excited dc generator isreadily voltage-regulated, reasonably effi-cient, capable of absorbing overloads, andcapable of wide speed variation underloaded conditions. It must be augmentedwith a solid-state inverter to provide thehousehold 60-Hz power. The beneficialoperating characteristics are offset bythe high cost of the required inverter.

FUTURE WORK

A base Iine computer model has beenestablished. In this way, the candidatecomponents may be substituted into the sys-tem and input/output performance examinedfor rotational losses. In addition, a life-cycle cost analysis will be accomplished toselect the lowest cost system.

After the system is selected, a layoutdrawiny and specifications will be accom-plished for the lu-kwhr storage system. Acost estimate will be prepared for both a10-kwhr and a i>G-kwhr extrapolated system.

REFERENCES

1. J. F. Banas, W. N. Sullivan:Engineering of Wind Energy Systems(Sandia Laboratories, SAND75-0530,January 1976).

2. B. F. Blackwell: The Vertical AxisWind Turbine, How It Works (SandiaLaboratories, SAND74-0160, December1974).

3. D. A. Towgood: An Advanced VehicularFlywheel System for ERDA ElectricPowered Passenger Vehicle (U.S.Department of Energy, CONF 77J053,October 1977).

4. Economic and Technical FeasibilityStudy for Energy Storage Flywheels(U.S. Department of Energy, HCP/M1066-01 , May 1978).

Fig. 8. Transmission tradeoffs.

The squirrel cage induction alternatoris capable of continuous near-synchronousoperation despite short term overloads.However, its efficiency is not as high asthe synchronous alternator unless it iscontrolled to a low slip operation. Ingeneral, when compared to the synchronous

233

SESSION V: SUPERCONDUCTING MAGNETIC ENERGY STORAGE

235

PROJECT SUMMARY

Project Title: Superconductive Energy Storage

Principal Investigator: R. W. Boom

Organization: University of Wisconsin531 E.R.B.Madison, WI 53706608/263-5026

Project Goals: To develop hardware components, produce engineering systemdesigns, procure manufacturing and assembly equipment designs,revise cost estimates, and assess operational efficienciesfor large underground superconductive storage systems.

Project Status: Component development in the areas of cryogenics, conductors,and structures is under way and to be completed in FY-81. Rockmechanics design and experimentation is under way and to be com-pleted in FY-81. Parallel efforts supported by the WisconsinUtilities are under way on the same schedule in five other areas:Magnetics, environmental studies, system design, electrical, andsafety.

Contract Number: EY-76-C-02-2844-000 Department of Energy

Contract Period: FY 1970 - FY 1981

Funding Level: $600,000 FY-78 D.O.E. *

Funding Source: Department of Energy, Division of Energy Storage Systems

Total Funds: $3,174,000 FY-70 to FY-78

ERDA, DOE $1,600,000NSF 715,000IKW. 554,000Wisconsin Utilities 250,000General Electric 55,000

'Included in this project are the following:

S. W. Van Sciver, "Recent Component Development Studies for Super-conductive Magnetic Energy Storage," page 247.

237

SUPERCONDUCTIVE DIURNAL ENERGY STORAGE STUDIES

R. W. BoomEngineering Experiment Station

University of WisconsinMadison, Wisconsin 53706

ABSTRACT

A general description of a large central energy storage unit employing super-conducting magnet technology is presented. Work on this concept has been under wayat the University of Wisconsin since 1970. Economic and engineering optimizationhas shown that a superconducting energy storage magnet system would be competitivewith other storage schemes for units larger than 1000 MWh. The device consists ofa large superconducting magnet of approximately 100 m radius interfaced with thepower system with a three-phase Graetz bridge. Economic analysis dictates that themagnet be a single layer segmented solenoid buried in bedrock and operated in super-fluid helium. An overall discussion of progress on the Wisconsin project is pre-sented including recent cost estimates and time schedule for commercialization ofthe technology.

INTRODUCTION

Energy storage studies have beenunder way at the University of Wisconsinsince 1970. The originating Wisconsinidea was that a three-phase Graetzbridge could be used to convert dc cur-rent in a superconducting storage magnetinto ac1current in a three-phase powersystem.' The Wisconsin work has con-centrated on large central storage unitsfor diurnal use while a subsequent energystorage effort at Los Alamos (see com-panion papers by J. Rogers et al. in thissession) has evolved to include a concen-tration on small storage units for utilitystabilization purposes. The efforts atWisconsin and at Los Alamos representthe major worldwide activity in super-conductive storage although there areindications that Japan and Russia arebecoming interested.

The concept is that large butsimple superconducting solenoids storeenergy in the form of dc currents in aninductance equal to V2LI2, where L isthe inductance and I the dc current. Thesolenoid turns are superconducting toeliminate I'R losses, where R is zeroresistance for superconductors. The useof superconductors necessitates the useof cryogenic systems and a liquid heliumcoolant. The solenoid would be mountedin bedrock which is the least expensivemechanical support structure available.

According to the virial theorem,M > § E, the structural mass M per kWh ofstored energy E must be greater than264 lbs/kWh for stainless steel with0 = 50,000 psi and p = 7.86 g/cm3. Theweight per pound can be reduced only byusing lighter material at higher stresslevels. The practical result is thatsuch structural requirements absolutelyforbid any purchased structure, only in-expensive bedrock is available. Fly-wheels also suffer from the same absoluteprohibition if $/kWh is significant.

The three-phase ac/dc Graetz bridgeand the dc energy storage magnet coilform an inductor-converter (I-C) unit.Typical uses for large I-C units are2-12 hour discharges at rates of 100 to2000 MW for nighttime charging and day-time discharging.

Large central storage I-C units areunique in that 90-95% efficiency is pos-sible. Another unique property is thespeed of response; within 50 milli-seconds an I-C unit can change its powerlevel from full charge to full dis-charge. We are not aware of any otherstorage system with these two advantages.

A total of $3,174,000 has beenexpended at the University of Wisconsinfrom 1970 through FY-1978 from thevarious sources listed in Table 1. The

University of Wisconsin and GeneralElectric Corporation provided initialfunding followed by National ScienceFoundation support which led to thepresent ERDA-DOE grants. Throughout theeight years the seven electric utilitycompanies in Wisconsin have providedworking staff in addition to the fundslisted which were transferred to theUniversity from the Wisconsin ElectricUtility Research Foundation.

Table 1. Aggregate funding for theUniversity of Wisconsin Energy StorageProject from 1970 to October 1, 1978.

SourceUniversity of WisconsinGeneral Electric Corp.Wisconsin UtilitiesN.S.F.E.R.D.A.D.O.E.

*Also provided in house

staff.

Amount$ 554,000

55,000250,000*715,000300,000

1,300,000$3,174,000

studies and

During this period faculty and stafffrom five engineering departments haveworked on the storage project. Therehave been 13 faculty, 12 research asso-ciates and assistant scientists (post-doctoral s), 5 permanent staff, and 32research assistants (graduate students).The project was fortunate to have avail-able specialists in all necessary disci-plines: superconductivity, cryogenics,rock mechanics and geology, metallurgyand materials, power engineering,mechanical design, and stress analysis.

This paper is one of two describingthe Wisconsin effort. In the secondpaper by S.W. Van Sciver the new andrecent research and development is pre-sented. In this paper is presented themore general aspects of the Wisconsinproject development with emphasis onsystem specifications, system design,electrical power system design and rockmechanics. Tentative cost projectionsand project schedules are discussed.

NEED FOR SUPERCONDUCTIVE STORAGE

REQUIREMENTSThe desirable amount of power from

storage in an electric utility system is

generally estimated to be between 5% and10% of the peak available power fromgenerators. The duration of the powerdelivered from storage would vary from2 to 10 hours in different utilitiesacross the country with a trend towardsneeding 12 hours from storage. The peakpower period starts in the morning after8 A.M. and persists through the lateafternooon on weekdays. Occasionallypeak needs for storage even arise duringa weekend. The peak power in Wisconsintwo years ago was on a Sunday that wasexceptionally hot and humid.

As an example let us take the stateof Wisconsin whose power requirements areabout average for the 50 states. Thepeak power is about 8000 MW which impliesthat 800 MW would be desirable fromstorage for about 10 hours, as has beendetermined by our utility collaboratingengineers. Thus 8,000 MWh might beneeded for the average state. We predicteconomic competitiveness for I-C unitslarger than 1000 MWh and thereforerecommend for Wisconsin, as an example,two I-C units of 4000 MWh each. Largerunits are less expensive per unit ofstorage but lack the reliability ofredundant smaller units. Thus compromisebetween size, cost and redundancy wouldbe made after operating experience isobtained.

The magnitude of U.S. storage needs isiaken as 50 states x 4000 MWh * 200,000 MWh,which assumes that the average state wantsstorage at half the Wisconsin rate.Wisconsin is fortunate in having 30% of itspower capacity from inexpensive new baseload generators which are available 90% ofthe time, a notably reliable performance.As expected, storage couples well withefficient generation. In addition, storagecouples well with intermittent generation,as would be available from future photo-voltaic cells, for example.

DISPERSED STORAGEThe magnitude of the utility storage

needs seems to preclude widely distributedsmall storage units. Most utility advisorswould not want 80 I-C units of 100 MWh eachand would prefer only two or three units ofequivalent total energy. Small units aregenerally inefficient and costly to buildper unit power. I-C units costs scale asEd/i, where E is the total stored energy.In Table 2 these cost trends are illus-trated. We can predict 200,000 MWh of

239

storage countrywide in 50 large unitswould cost 27% of the cost for 2500smaller units. In addition, maintenance,siting and environmental controls mightscale even less favorably for many smallunits.

Table 2. Relative capital costs ofI-C storage units.

Size10,000 MWh8,000 MWh5,000 MWh2,000 MWh1,000 MWh500 MWh100 MWh

POLLUTION AND SITING

Capital Cost/MWh

1.001.081.261.702.162.724.63

One of the very few pollution prob-lems associated with I-C storage is theelectromagnetic interference arisingfrom the three phase Graetz bridges.This problem is common to most forms ofelectrical storage, especially storagebatteries, which rely on ac/dc conver-sion. The bridges must be shielded toprevent interference on telephone systems.It is much easier to shield 50 interfer-ence sources than 2500 sources, which againmitigates against widely dispersed storage.

There is no need to locate storageunits near generators, it is onlyrequired that adequate transmissionlines exist between the storage unit andthe power system.

ALTERNATIVES

The two main competitors for loadleveling today are pumped hydro storageand load management through time of daymetering. Pumped hydro storage iseconomic and would be attractive whereverthe terrain allows for upper/lower bodiesof water and environmental standards canbe met. There are very few sites avail-able in the central U.S. and environ-mental disadvantages are extensive.

Time of day metering is a costlymetering snd billing process which addsnothing productive to a system. WisconsinPower and Light, which is the 60th largestutility in the country, is a leader in

time of day metering. It will require$100 M to completely install metars 1nthe WPL system. We estimate that $100 Mwould buy a 1500 MWh I-C unit. Suchstorage is 10% of WPL peak power for10 hours and would probably eliminate theneed for time of day metering with itsimplied disruption in life style.

ADDITIONAL CREDITS

Supplementary uses for I-C unitsin power systems, such as AGC (automaticgeneration control), transient stabilityregarding major disturbances and voltageregulation have been discussed in ourreports. The major use, of course, isthe diurnal storage and release of energy.What makes I-C storage high quality is itsspeed of response. No other storagesystem can reverse power direction within50 milliseconds. Such speed might preventsystem blackouts following losses of loador generators. Load following second bysecond can be provided by an I-C unitsimultaneously with its major charge-discharge function. Such load follow-ing is otherwise unavailable and shouldgreatly reduce the wear and tear on"old" generators which normally provideload following functions.

In summary, load leveling in electricutility systems by large central storageunits is needed. Superconductive I-Cstorage may prove to be the best option topumped hydro storage and time of daymetering.

SUPERCONDUCTIVE ENERGY STORAGE SYSTEMDESIGN STUDIES

EARLY RESULTSThe early work between 1970 and 1976

was primarily a feasibility study whichindicates that superconductive storage istechnically and economically feasible. Theresults of these early studies have beenpublished in Vol. I and Vol. II, Supercon-ductive Energy Storage Reports, Universityof Wisconsin and in the first 19 paperslisted in the Wisconsin bibliography.The major results are:

1. Bedrock structure is needed.

2. Cryogenic stability for the con-ductor is required.

3. Pool cooling with superfluidhelium is preferred.

240

4. An aliiminum. stabilized NbTi com-posite is planned.

5. The conductor, dewar, dnd asso-ciated structure are to be rippled atapproximately 1 meter radius of curvature.

6. A one layer, thin wall, highcurrent, solenoid is probably best.

7. A multi-tunnel, sectored sole-noid results in less cold structure andgreater safety.

Since 1976 the project has enteredthe component research and developmentphase, as reported in papers 19 to 55 ofthe bibliography and in the 1976 AnnualReport published in May, 1977. In thecompanion Wisconsin paper by S.W. Van Sciverthe study of conductor, structure andcryogenic systems is given. Here we in-clude the electrical system and rockmechanics studies.

ELECTRICAL

The external electrical circuitryconsists of three-phase Graetz bridges,such as those used on dc transmissionlines. By arranging the converters inseries and parallel large dc currents andreal and reactive power control can beachieved.1 Superconducting short circuitswitches across the storage magnet are notplanned because 10 hour charge and 10 hourdischarges use up the time available. Theenergy loss in the leads and bridges is aminor loss and can be tolerated for 24 hourperiods. If a short circuit is needed thena thyristor bypass switch is probably thebest choice.

The internal magnet electrical con-siderations are unique to superconductingmagnets. Magnet dc voltages up to 10 kVand currents up to 330,000 A are plannedfor storage magnets with fields up to2.5 T. Because of the large currents itis necessary to limit the number ofexternal leads, preferably to only two.Voltage breakdown in helium liquid andvapor, and through and across insulators,limits internal design voltages to a fewhundred volts. The lowest voltage break-downs occur in helium vapor. Our experi-ments show that 1 kV is the maximumvoltage across reasonable separations of afew mm.5Z Me plan for voltage drops to betaken across solid insulators so that thehelium will not be subjected to largepotential differences.

Magnet safety will be achieved bysubdividing the unit into severalseparate vacuum and helium sections sothat any local problems are isolated toa fraction of the total unit. Energycan be removed quickly from a section bymutual coupling to other magnet sectionsin conjunction with some external energydischarge through the Graetz bridge. Inthis way magnetic fields and magneticfluxes are maintained approximatelyconstant, thus avoiding induced voltagesand ac losses.

ROCK MECHANICS

The preferred design is a multipletunnel unit shaped for a favorable fieldand force distribution. One examplestudied is the "C" shaped three tunnel1000 MWh unit sketched in Fig. I.36

This three tunnel unit minimizes shearloading on the walls by arranging theturns in a "C" shape. The magnetic loadis transferred to the outer wall of thecentral tunnel and the inner horizontalwalls of the outer tunnels as shown bythe arrows in Fig. 1.

Fig. 1. Three dimensional view of threetunnel magnet system. Arrows denote direc-tions in which the rock is loaded. Thecentral tunnel is 15 m high; its inner wallradius is 65 m (not drawn to scale).

Site studies and associated labora-tory work are under way in order todetermine criteria for I-C storage units.Site investigations include geologicmapping, hydrologic studies, core holestudies, laboratory and field mechanicalstudies, in situ rock stress measurements,and finite element analysis. Four coreholes have been drilled; in granite,

241

guartzite, rhyolite and carbonates(dolomite) to depths of between 200 mand 300 m. The data obtained is used toconduct the finite element analysis ofthe rock structure around the threetunnel unit. One result is shown inFig. 2, where surface tensile stresses orsurface shear failures are reduced withappropriate rock bolting.36

• ft ffli •

fl CC fl C a• a

Fracture Geometry and Hydrology. Fracturegeometries combine with stress distribu-tion to cause unstable zones in thetunnels; of particular importance arezones where joint systems parallel theinner walls of the central tunnel andthe area between the tunnels in the 90°profile. Rock masses must have lowpermeabilities controlled by fractureswhich can be successfully grouted.Bolting may be necessary in some partsof the inner walls of the centraltunnel, the roof of the upper tunnel,and the floor of the lower tunnel toavoid joint deformation which mightcause joint aperture increases and waterflow problems.

O'-W0«45*

Fig. 2. (a) Potential shear failure zonesbefore reinforcement, 0° profile, (b)potential shear failure zones before rein-forcement, 90° profile, (c) rock boltpattern; note that the bolt patternchange's with profile, (d) shear failurezones after reinforcement, 0" profile,(e) shear failure zones after reinforce-ment, 90° profile.

Some general conclusions from therock mechanics work are that three majorfactors affect the stability of storagetunnels:

Rock Strength. The rock mass must bestiff enough to contain the magneticloads without deforming so much as todamage the conductor. Because of jointsand discontinuities, we have found somereinforcement is necessary even in hardrock masses.

In Situ Stress. Stress distributions inthe tunnel vary strongly with profileabout the tunnels. The least stablezone appears to be the 90° profile whichIs parallel to the greatest horizontalin situ stress. The 0° profile isperpendicular to the greatest horizontalin situ stress. Depths of at least 300-400 in are necessary to avoid verticaland tangential tensile stresses in partsof the tunnels.

COST ESTIMATES

Over the years cost estimates andcost reduction engineering research hasbeen emphasized. In 1976 a particularlycareful cost optimization and design wasundertaken. The following tables out-line engineering progress and develop-ment over the years. The cost basis inall cases is the 1976 dollar and changesin costs result only from engineeringimprovements. The cost in mills/kWhstored is based on delivering 90% of thestored energy in a 10 hour day. Theyearly cost of the unit is taken as 16%of the original capital cost for interest,taxes, dividends, maintenance, etc. and20% for interest during construction.These were typical rates for 1975-76.

In Table 3 the initial cost esti-mate for storage is 101 mills/kWhdelivered. The copper in the compositeconductor and the stainless steelstructure are too expensive. The storagecost in mills/kWh delivered is a bettermeasure than $/KW because differentdischarge times drastically effect thekilowatt rate for a given storage magnet.The reader i s referred to Vol. II for acomplete set of cost presentations.

In Table 4 the design status in1974 shows substantial improvementresulting from replacing copper withless weight aluminum and steel structurewith bedrock structure. The storagecost estimate of 27.1 mills/kWh istolerable when one notes that in 1976a cost of 60 mills/kWh was attached togas turbine peaking power.

242

TlMt >. lallltl Kaae «>l|» l> 1171.

Mriacttr

Cwtactar

Strvctafa

N w cajripaiat

t r j w 1 1 ' CMUIwr

•afrlMratar

K<al< Hall*

Moata tlualia

can staiaitii suel

CMVtrttr I r i t t n

SUtHlfU Stn) '

lf«iM Hailaa

Jtatlltala

hillatu

JMIlaala

Stantara.ftallaala

ftnlitklt

•ntlaalt

ttillikit

1.1 Mlli/llk

n.«u.oJ. I

t . t

O.I

0.1

T5 allliAla,

by better and less expensive reinforcedpolyester struts.

T«tn: 10 ailli/Ua

OMClnlais Ceaatr tl Slractiira toa b t n t t o

tO.OOO attialt lain

CtnlactarCanfcctor

Stnctarw

•War Cqutpmt

Cr>a|aiilc CaatalMr

atfrllarator

tl«aM Haltaa

•Stmtt

aall

M »

HackCafara

Canarur I r te j t i

•laalnaa Alia;

SwarfliiM Hallaa

AvaiUMi

ESSES-SHclaiiittEKavatiaa

[ItctHul a>MWigt Statlai

<>allaila.

S B S U -uHrj""^Mtllakta

Oa^waM

1.1 anil/Ma)

t . t

t . l

t . t

0.0

0.1

0.4

t . l

17.1 a l l l l / u *

•all: Contuctar, strati. Hack Hacktnici ant HtfrliaraKr »Mf

CaMlinlan: Alwlnui anf Rack tavar. aaia fieoa Prvttactf ncca»ta.li

In Table 5 the benefits of engi-neering optimized component design isevident. Costs are now reduced to18.7 milis/kWh due to the properselection of magnetic field, strutlength and conductor.

inn «. Ortiatirt fciin. m i .

CMactar

CaiakclaF

Smctan

IkMtr Cwjiaatat

•tfrliaratar

l l talt Ikilaa

Strata

M l

alialma

Hack

IrlHaa

« Allay

Si«trf1a<. Htllui

n u , cirrmitu Rnwractarv

Haaa-i atallcatltaOtyaiopacnt

l.talllt /kW

t . l

•vallaala

Urt t . A.lllrtla

1.1

t . l

t . l

t . l

t.t

11.7 allls/UD

Halt: Cantoctar, stnt . Itfrlgtratar ana Hacl Ikchmln Unaar Omlaaaaiit

lawlulaa: Coiavcur Oaaccaataki.

In Table 6 the current reviseddesign costs slightly more at 19.2milis/kWh. The main advance 1s thatthe conductor is now deemed to bemanufacturabie. Heat transfer data ismeasured to be better than previouslypredicted and epoxy struts are replaced

Tiaia t . anlna anl|a, 1170.

OaaKUr

CwaKUr

Stractara

• n t r Iwlfaaat

Covtanlc CaAUlHar

RafrliaraMr

UiaM Malta*

Stmtt

M l

•1

kKt

•rlajn

•1 alley

JwartlaH

—

Nljattrr Plbar«1a»

ItntrntM

Rw&aB1"*

Milatte

tSSSXiItHtauat

lantrMtaai

lau bpaatlat.batrlaaMtl

MMIWW.

t . t

l.t

I . I

«.«

I . I

l . t

t.r

it.l oiiiinaa

•ata: k t t n i w r M i u l Oxaitaaau t f aaaaa Caa««ur. Slnrt aaj~~*" Hallda Kaat frantfar an. Iraai»art

Canclailani Cawoiwt t t l l l " an. Ttit Wtl IMaita». » » t Ma> I rDavaiepaant af rakricatian E l t

Both engineering and cost progressare seen in the above tables. The costbasis is given in Chapter II, Vol. IIand is purposely conservative.

PROJECT SCHEDULE

COMPONENT DEVELOPMENT

For the next two years, FY-79 andFY-80, component development and testwill be completed for the conductor,strut, and cryogenic system. Rockstudies will continue with emphasis on asite near Madison. In FY-81 a substan-tial portion of the funds will be usedto procure the Industrial design ofcomponent manufacturing and assemblyequipment (conductor, struts, cryo-genics). Specialized rock excavationequipment may need to be designed.The system construction equipmentwill be designed in FY-81.

Following the development of thecomponents and the design of theassembly equipment it will be possibleto accurately assess the cost estimatesand recommend the future constructionof the first model unit.

MODEL UNIT

At this time we anticipate that themodel should be in bedrock and that allideally developed fabrication and con-struction equipment should be used. Themodel will not be built to show that itoperates. The magnet is so conservativethat operation is absolutely guaranteed.For example, the composite conductoruses the well developed Fermi National

243

Laboratory conductor which carriescurrent without all of our aluminum.The additional aluminum insuresreliability and trouble free operation.

The model is to be built to demon-strate that the proper constructionequipment has been used and to verifythat costs are known.

Scaling is easy. Since only fullscale conductor is used at full currentat full field at full radius of curva-ture there is no extrapolation needed tolarger sizes. The first unit could bein the 10-50 MWh range and the demon-stration unit to follow in the 100-1000 MWh range.

CONCLUSIONS

Superconductive energy storage istechnically and economically attractive,according to the Wisconsin developmentand design studies. Storage efficienciesup to 95% seem probable. The furtherdevelopment of components and the designof manufacturing and assembly equipmentwill provide a more advanced confirmationof this favorable assessment. The firstmodel to be built in bedrock with theproper equipment will provide the finalconfirmation of the use of superconductivestorage for load leveling in utilitysystems.

Wisconsin Bibliography (1972.78)

Reports

1. Wisconsin Superconductive Energy Storage Project Volume . 1 , Universityof Wisconsin. July 1 , 1974.

I I , Wisconsin Superconductive Energy Storage Project Volume 11, Univer-sity of Wisconsin. Januiry 1 , 1976.

111. Wisconsin Superconductive Energy Storage Project Annul! Report,Un1vors>ty of Wisconsin, Hay I , 197/.

Publications

1 . -Superconductive Energy Storage for Power Systems," Sow, R. V. andH. A. Peterson. IEEE Transactions on Higneties. Vol. HAG-B. Ho. 3,September, 1972, pp. 701-701.

2. "Superconducting Energy Storage," Boom, H. If., G. E. ffclntosh, H. A.Peterson and W. C» Young. Advances in Cryogenic Engineering. Vol. 19,plenu* Press 1974, pp. 117-126.

3. "Inductive Shielding for Pulsed Energy Storage Magnet's," Hoses, R. U.,Jr.and J . K. Ballon, IEEE Transactions on Magnetics. Vol. HAG-11, No. 2,Harch 1975, pp. 495-*96.

4. "External Field Reduction of Superconducting Energy Storage Solenoids,"Ballou, J . K. «nd R. H. Moses, Jr . . IEEE Transactions on Hagnetict.Vol. HAG-11. No. 2. ««rch 1975, pp. 497-499.

5. "Cellular Concrete - A Potential Load-Bearing Insulatton far CryogenicApplications." R1ch»ro, T. G., J . A. Dobogai, T. 0. Gerhardt andW. C. vouns. IEEE Transactions on Hagnetics. Vol. HAS-11, Ho. 2,Harctt 1975, pp. 500-502.

C. "Flu* Olffuslon Losses 1n StabilUed Conductors," Hl la l , M. A. andft. H. Soon. IEEE Transactions on HagnetJcs. Vol. KAG-11, Ho. 2 .Harch 1975, pp. 444-54/.

1. "Superconductive Energy Storage for targe systems," B O M , R. If.,8. C. Hiinson, C. E. Mclntosn. H. A. Peterson and W. C. loung, U KTransactions on Hagnc.ics. Vol. MAS-11, No. 2. Harch 1975, pp. 4TC4B1.

B. "Superconductive Inductor-Converter Units for Pjlsed Power Loads,"Peterson, H. A., N. Hohan. W. C. Young ind R. w. 8otm. Energy Stc—•Compression, and Switching. Plenum Press, 1976, pp.

9. "Superconductive Energy storage Inductor-Converter Units for PowerSystems," Peterson, H, A., K. Hohan and R. W. Boon, HEE Transactionson Power Apparatus and Systems. Vol. PAS-94, Ho. 4, July/August I t / * ,pp. 1337-13*6.

1U. "A Look at Superconductive Storage," Peterson, H. A., It. U. lorn,N. C. Storck and W. C. Young, Electrical World. Harch 1 , Wi,

11. "Wisconsin Superconductive Energy Storage Project," Peterson, tl. A.,R. W. Boon and U. C. Young, Proceedings of the American PowerConference, Vol. 37, 1975, pp. 1046-1057.

12. "Magnet Design for Superconductive Energy Storage for Power 5ystems,"Bom, R. w., H. A. Hi la i , ft. W. Hoses, G. E. Kclntosh, H. A. Peterson,R. L. Willig and W. C. Young, Fifth International Conference on MagnetTechnology (HT-5) Proceedings Roma. (EUltjV Italy, April 21-25, 1975.Laboraton National) del CNLN, July 1975, pp. 477-483.

13. "Configurattonal Design of Superconductive Energy Storage Magnets,"Hoses, R. U., Jr . , Advances in Cryogenic Engineering. Vol. 21, PlenunPress, 1976, pp. MET*?:

14. "Optionalten of Mechanical Supports for Large Superconductive Magnets,"Hi ia l , M. A. and R. U. Boom, Advances In Cryogenic Engineering. Vol. 22,Plenum Press, 1976. pp. 224-23~2l

15. "Cryogenic Design Elements for Large Superconductive Energy StorageMagnets," Hllal , H. A. and G. E. Hclntosh, Advances 1n CrvocenicEngineering. Vol. 21, Plenum Press, )976, pp. 69-7?.

16. "Electrical and Mechanical Properties of Dilute Alumiiium-Gold Alloysat 300, 77 and 4.2 K," Hartwig. K. T . , F. J . Worzala and H. E.Jacltson, Advances in Cryogenic Engineering. Vol. 22, Plenun Press.1976, pp. 472-476.

17. "Low Temperature Resistance Studies on Cyclically Strained Aluntnw,"Segal, H. R. and T. G. Richard, Advances 1n Cryogenic Engineering.Vol. 22, Plenum Press. 1976, pp. 486-469.

18. "Superconductive Energy Storage for Tokarok fusion Reactors," L«»J. W., H. A. Peterson and R. W. Boom, Proceedings of the SixthSymposium on Engineering Problems of Fusion Research. 1975. pp. 291-95,IEEE Pub. Ho. 75CHI09J-5-NPS.

19. •Superconductive Energy Storage Inductor-Converter Units for PowerSystems," Peterson, H. A., EFC Session on SuperfiywhceU and Super-conductive Storage, Asilonar, California, February 8-13, 1976.

20. "Magnet Design for Superconductive Energy Storage for ElectricUtility Systems," Boon, R. u . , B. C. Haimson, H. A. Hl la l , R. W.Hoses, d, E. Hdntosh, H. A. Peterson, R. L. Hlllig and U. C. Young,EFC Session on Superflywheels and Superconductive Storage, Asilomar,California, February 8-13, 1976.

21. "Dilute Aluminum alloys: Their Potential in Superconducting Devices,"Hartwig, K. T. and F. J. Woriala, 105th A1KE Annual Meeting, Las Vegas.Nevada, February 22-26, 1976.

22. "Determination of Magnetic fields at the Conductor for Solenoids «ndToroids," Moses, R. W. and R. L. WilHg, Proceedings of the CQHPUHAfiConf. on the Computation of Haonetic Fields. Oxford. England.March 31-April 2, 1976.

23. "Oilute Aluminum Alloys: Their Potential in Superconducting Devices,"Hartwig, K. T. and F. J. Worzala, Proceedings of the Sixth Interna-tional Cryogenic Engineering Conference. Grenoble. France. Hay 1976.pp. 406-410.

24. "Free Convection Heat Transfer to Supercritical Helium," Hl la l , M. A.,R. w. Boom and M. H. El-Hakil, Proceedings of the Sixth InternationalCryogenic Engineering Conference_and Exhibition. Grenoble. France.Hay 1976, pp. 327-329.

25. "Transition and Recovery of Cryogenically Stable Conductors," HJU1,M. A. and R. M. Boom, Proceedings of the 9th Symposium on FusionTechnology. Pergamon Press, Oxford and New York, 1976, pp. 67-93.

26. "Optimization of Current Leads for Superconducting Systems," Hllal ,M.-A., IEEE Transactions on Hagnetics. Vol. HAG-13, Ho. 1 , January1977, pp. 690-693.

27. "Reinforced A'uminum as a Superconducting Magnet StaMIIier," Segal,H. R., [EFE Transactions on Hagnetics. Vol. HAG-13, No. 1 , January1977, pp. 109-1. \

28. "Problems Associated with the Use of High Purity Aluminum In theDesign of Composite Conductors 1n the Elasto-Plastic Region,"Ladkany, s. G. and W. C. Young, IEEE Transactions on Magnetics.Vol. HAG-13,. Ho. 1 , January 1977, pp. 105-108. ~

"Properties of RF Sputtered Nb,Ga Superconducting Films," Burt, ft. J .and F. J . Worzala, IEEE Transactions on Magnetics. Vol. HAG-13, Ho. 1 ,January 1977, pp. 323-326.

"A Method for Preventing Pressure Oscillations In Tubes ConnectingLiquid Helium Reservoirs to Room Temperature," H l la l , H. A. andG. E. Kclntosh, Cryogenics. February 1976, p. 122.

"Superconductive Inductor Storage and Converters for Pulsed PowerLoads," Hohan, N. and H. A. Peterson, IEEE International PulsedPower Conference, Lubbock, Texas, November 9-11, 1976,

"Site Characterization for Tunnels Housing Energy Storage Magnets,"Haimson. 6. C , T. W. Doe, S. R. Erbstoesser and G-F. Full. Proceedingsof the 17th U.S. Symposium on Rocfc Mechanics. Snowbird, lftan\August, 1976, pp. 34-1-9.

"Th1n-Walled Solenoid End effects," WllUg-, R, L. and R. W. Hoses, Or.,presented at the 15th Annual International Magnetics Conference,Intermsg 77, Los Angeles. Calif . , June 1977.

"Inductor-Converter Superconductive Energy Storage Systems for ElectricUtil ity use," Peterson, H. A., R. W. Boom and W. C. Young, presentedat the World Electrotechnical Congress, Moscow. USSR, June 21*25, ¥977.

"Structural Design for Large Superconducting Magnets," Young. If. C ,R. W. Boom and S. G. Ladkany, presented at the World ElectrotechntcalCongress, Moscow. USSR, June 21-25, 1977.

29.

30.

31.

244

K , "Design of Underground Tunnels for Superconductive Energy Storage,"Fun. G. F., T. One and I . C. Haimsor, 18th U.S. Sy*po>1u> on lockMechanics, Keystone, Colorado, June 22-24, 1977.

37. "Constant Tension and Constant Field Solenoids," M. * . El-Derlni,R. W. tool and N. A. Hi la l , Advances 1n Cryogenic Engineering.Vol. 23, Plenum Publishing Corp., Hew Yorto, 1970. pp. B8-l».

31. "Hew Crltorla for Refrigerator and L'quefler Cycles Design,"H. A: Hi la l , presented at CEC-ICHC. Boulder, Col., Aug. 2-5. 1977.

99. "Heat Transfer to Subcooled Liquid Helium," Ibrahim. E. A. .R. y . Boom and G. E. HclntDsh, Advances In Cryogenic Engineering.Vol. 23. Plenum Press, 1978, pp. 333-339.

40. "Dielectric Strength of Helium Vapor and Liquid at TemperaturesBetween 1.4 K and 4.2 K," Hwang, K. F.. Advances In CryogenicEngineering. Vol. 23, Plenum Press, 1978, pp. 110-117.

41. "Resistance to Strain Degradation 1n Preliminary IIWWC TF CollConductors for Fusion Reactors," Kong, S. 0 . , P. F. Hichaeison,I . H. Sviatoslavsky, and U. C. Young. Advances In CryagenicEngineering. Vol. 23, Plenum Press, I9)&\

42. "ThermodynMlc 0pt1m1ial1on Study of the Helium Hulti-Englne ClaudeRefrigeration Cycle," Khaiil, A. and G. E. Hclntosh, Advances InCryogenic Engineering. Vol. 23, Plenum Press, I97B,

43. "High Current Al-T1Nb Composite Conductors for Large Energy StorageHignets." Ladkany, S. G.. Advances 1n Cryogenic Engineering. Vol. 24,Plenum Press, I97B,

44. "bnpressive Strength of Glass Fiber Reinforced Composites at RoomTemperature and 77 K,' Stone, E. L. and w. c. Young, advances 1nCryogenic Engineering, vol. 24, Plenum Press, 1978,

45. "Kapitia Conductance of Aluminum and Heat Transport from I FlatSurface Through a Large Diameter Tube to Saturated He I I , " Van Sciver,S. U-, Advances In cryogenic Engineering. Vol. 23, Plenur. press, 197B,pp. 341-348.

46. "The Effect of Thermal Treatments on the Critical Current Oensity ofa Commercial NbTI Filamentary Superconductor," Larbalesticr, 0. C ,R. Flach and D. G. Hawksworth. Proceedings of the Sixth InternationalConference on Magnet Technology. Aug. 29-Sept. 2 . 1977.

47. "Formation of Superconducting A-15 V-Ge Compound by • Compositt-Olffusion Process," Tachlkawa, K., K J . Burt and K. T. Kartirlg.Journal of Applied Physics, vol. 48. No. 8, August 1977, pp. 3623-25.

48. "Geotechnice! Investigation and Design of Annular Tunnels for EnergyStorage," Haimson, B. C.. T. Doe and G. F. Fun, presented at theInternational Symposium on Storage 1n Excavated Rock CtntmtSeptember 1977.

49. "Underground Caverns for Energy Storage Using Superconductive Magnets,"Hilmson,.B. C , K. T. Hartwig and T. w. Doe, Underground Space.Vol. 2 , Pergamon Press, 1978, pp. 137-142.

'Cryogenlcatty Stable' Holloa Conleliutt." H l la l , H. A. and R. u. ISymposium on Engineering Problem:

t Conductors Cooled sy Supercritical. . W. Boon, Proceedings of the Seventh

Symposium on Engineering Problems of Fusion Research, Knoxvtlie,Tent... October 25-28, 1977, Vol. 1 . pp. 695-699. IEEE Pub. Ho.7JCH12SM-WS.

SI . "Design of Energy storage Solenoids for Tokamak Reactors," Cl-Derinl,«. « . , K. u. loon and H. A. Peterson, Proceedings of the Seventh5yanos1un on Engineering Problems of Fusion Research, Knoxvilic,Tern., October 25-28, 1977. Vol. I I , pp. 1367-1370, IEEE Pub. No.77011267-4-HPS.

£2. 'Dielectric Breakdown of Liquid and Vapor Hellun In Bulk and AcrossEpoicy Insulation," Hwang, K. F. and S. 0. Hong, Proceedings of theSeventh Symposium on Engineering Problems of Fusion Research, Knoxville,Tenn., October 25-28, 1977, Vol. I I , pp. 1531-1534, IEEE Pub. r».T7CH1267-4-NPS.

53. "F;t1gue Tests with Small Colls of Filamentary ffl>-Sn,r Larbalestler,D. C. and S. 0. Hong, Proceedings of the Seventh Symposium on Engineer-ing Problems of Fusion Research, KnoKiHlle, Term., October 25-28. 1977.Vol. I I . pp. 1260-1262. IEEE Pub. No. 77CH1267-4-HPS.

54. "Compressive Strength of Glass Fiber Reinforced Composites at RoomTemperature, 77 K and 4.2 K," Stone, E. L. and V.. C. Young, Proceedingsof the Seventh Symposium on Engineering Problems of Fusion Research,(nonville. Term.. October 25-28, 1977, Vol. I I . pp. 1510-1512. IEEEPub. Ho. 77CH12S7-4-KPS.

55. "CryosteblHution of Large Superconducting Magnets using Pool BoiledHelium I I , " Van Sclver, 5. » . . Proceedings of the Seventh Symposiumon Engineering Problems of Fusion Research. Knoxville, Tenn.,October 25-28, 1977, Vol. I , pp. 690-694, IEEE Pub. Ho. 77CHI267-4-W5.

245

RECENT COMPONENT DEVELOPMENT STUDIESFOR SUPERCONDUCTIVE MAGNETIC ENERGY STORAGE

S. U. Van SciverEngineering Experiment Station

University of WisconsinMadison, Wisconsin 53706

ABSTRACT

For the past two years the Wisconsin Superconductive Energy Storage project hasbeen in the component development phase. Work during this period has been principallydirected toward development of the superconducting composite conductor and the fiberreinforced strut needed to transfer the loads from the magnet to the bedrock structure.Conductor work has involved design as well as research into manufacturing techniquessuitable for mass production. In a related activity, heat transfer studies in super-fluid helium has shown potential for substantially better conductor stability thanoccurs in normal helium. Compressive testing of fiber reinforced composite materialshas been employed to evaluate the commercially available products. A current surveyof progress in the component development studies at Wisconsin provides the main contentof the present paper. The direction of future activities is also discussed.

INTRODUCTION

In 1976, The Wisconsin Superconduc-tive Energy Storage Project entered thecomponent development phase. Thisactivity, which has followed the earlysystem optimization and conceptualdesign studies, has required initiationof development programs in several keyareas. Based on the conceptual designstudies, development has centered aroundthe superconductive composite conductorand the fiber reinforced compositestructural supports of the storage magnet.The principal goal of this phase is toidentify the best materials and manu-facturing techniques so that equipmentcan be designed to mass produce thecomponents for a full scale energystorage magnet.

It is the purpose of the presentpaper to provide an up-to-date survey ofthe progress that the Wisconsin group hasmade during the component developmentphase of the project. The discussionwill include the directions for futureactivities.

COMPONENT DESIGN

The Wisconsin approach to the designof superconductive energy storage magnetshas been discussed in detail in theadjoining article' as well as innumerous papers in the literature. "° Abrief review of the overall concept is

presented here to provide a basis fordiscussion of the components.

A superconductive energy storagemagnet for electric utility load levelingand peak shaving has been estimated to becommercially competitive with alternatestorage schemes for units larger than1000 MWh.6 In order to make the devicecost competitive, it must be buried inbedrock, which provides the structuralsupport at the rather low cost of exca-vation. A schematic of a superconductiveenergy storage unit is shown in Fig. 1.The dimensions of a 1000 MWh magnet areapproximately 70 m radius and 30 mheight, for a magnetic field at thecenter of the solenoid of 5 T. Thesolenoid is segmented into threeseparate magnets to minimize the coldstructure and to reduce the shearloading on the rock wall.

Fig. 1. Conceptual design of a super-conductive energy storage facility.

247

A detailed view of the magnet and sup-port structure are shown in Fig. 2. -Themagnet consists of a single layer solenoidwound from aluminum stabilized-NbTi super-conducting composite conductors cooled bypool boiling superfluid helium at 1.8 K.The conductors have a round cross sectionand are rippled in a radius of curvatureof approximately 1 m. The ripple isnecessary to reduce the hoop stress in theconductor. The radial load is transferredto the bedrock through fiber reinforcedcomposite struts. If the magnet were notrippled in the above fashion, it wouldrequire a more complicated and costly sup-port structure.

Fig. 2. Detailed view of magnet supportstructure.

The superconducting composite conduc-tor requires an active research and develop-ment program. The end product of thisprogram must be a high current conductorwhich can be manufactured in continuouslengths and at moderate cost. The Wisconsinanswer to the above requirements is shownschematically in Fig. 3. The diameter ofthe full scale conductor is 8.0 cm whichhas a design current of 286,000 A.9 Theoverall current density of the conductoris 5700 A/cmz.

The conductor has three principalcomponents. The superconductor, locatednear the outer surface, consists of abraid of NbTi copper composite verysimilar to that used by the NationalAccelerator Laboratory in the EnergyDoubler Project. At this time, nodevelopment programs are required for

the superconductor since it is currentlyavailable in industry.

HIGH STRENGTH ALUMINUM SHELL

HIGH STRENGTHALUMINUM WEB

SUPERCONDUCTINGFILAMENTS

HIGH PURITYALUMINUM STABILIZER

Composite NbTi-aluminum conductorFig. 3.design.

Stability of a magnet is a basicrequirement so that it can operatewithout unplanned quenching. Full cryo-genic stability must be the approach forlarge superconducting magnets since thetotal energy stored is very large. Byfull cryogenic stability we mean thatthe conductor contains sufficient highpurity normal metal to carry the currentshould the superconductors go normal forany reason. We determine the requirednormal metal by using the cryogenicstability criteria10

\c p/A = qS O)where the joule heat generated per unitlength, I2 p/A, must be dissipated inthe cryogenic coolant through the con-ductor surface, S. The allowable heattransfer, q, is determined by theproperties of the helium.

An important design decision earlyin the project was to propose to coolthe magnet with liquid helium at 1.8 Krather than at its normal boiling pointof 4.2 K. This decision was motivatedby two considerations. From the stand-point of cost, the superconductorrequired for 1.8 K operation is roughlyhalf that needed for 4.2 K.11 Thisreduction in NbTi more than offsets theadded refrigeration for the lowertemperature operation. The other aspectto the decision was that at lower tem-peratures, T < 2.2 K, helium becomessuperfluid having substantially betterheat transfer characteristics. Thiscooling concept requires further research

248

to evaluate the heat transfer charac-teristics of superfluid helium.

High purity aluminum has been chosenas the stabilizing material in the energystorage magnet conductor. This choice isprincipally because aluminum can be easilyand cheaply manufactured in the very purestate. In addition, aluminum has a lowermagneto-resistance than copper. The dis-advantage of using high purity aluminumis that it has a very low yield strength.Under the loading which is present in theenergy storage magnet, the high purityaluminum yields completely and must becontained. The conductor is designedwith a skin of intermediate strengthaluminum to contain the stabilizer.

The structural cruciform, the thirdcomponent of the composite conductor, isnecessary to carry the hoop tensionbetween the support struts. Also, theconductor must be able to transfer loadsto the struts requiring bearing stresscapabilities. The star cross section hasbeen chosen to insure against collapseof the conductor at the point where ittransfers loads to the strut.

Finally, the conductor has beendesigned with a circular cross section topermit better manufacturability. Eachcomponent can be extruded separately andassembled with a combination of drawingand swaging processes. These procedurescan be carried out nearly continuouslyallowing very long lengths to bemanufactured.

The design of the fiber reinforcedcomposite strut is shown in Fig. 4.Since the strut separates the conductorwindings at liquid helium temperaturesfrom the ambient temperature rock wall,it must have both a high strength and alow thermal conductivity.

The design scheme calls for coolingthe strut at intermediate points, 11 Kand 70 K, to minimize the heat load atliquid helium temperatures.IZ Refriger-ation power required to recover thelosses through the strut amounts to themajor loss mechanism in the energystorage magnet. In addition, thechoice of material must allow easy lowcost manufacture since many struts arerequired for an energy storage unit.

Apparently the best material forthe strut is a composite of glass fibers

in a polymer matrix such as polyester orepoxy. The ultimate strength's of thesematerials ure typically greater than100,000 psi (690 MPa) while thermalconductivities are in the range 0.1 to1 W/mK.

SECTION *-*

Fig. 4. Conductor, dewar and strutassembly.

COMPONENT DEVELOPMENTCONDUCTOR

At Wisconsin, we have demonstratedmanufacturability of the conductor conceptdiscussed above. In a small productionfacility, we have been able to producequarter scale test samples. A photographof a completed conductor sample is shownin Fig. 5. The cruciform is extruded inindustry from 6061 aluminum. Wedges ofhigh purity aluminum are then insertedwith the superconductor, also purchasedfrom industry. The entire assembly issubsequently inserted into a tube of in-termediate strength aluminum, such as6063, and swaged together to form atight monolytic unit. Consideration isbeing given to alternative aluminumalloys particularly for the skin wherestrength and thermal conductivity mustboth be high.

Future programs in conductordevelopment are directed toward studyingmechanical, electrical and thermal pro-perties of the conductor samples whichare being produced. A 1,000,000 lbtesting facility and a 3 T large bore

249

Fig. 5. Quarter scale conductor samplemanufactured at Wisconsin.

superconducting magnet are available toinvestigate the performance of thisconductor.

CRYOGENICS

The volume of high purity aluminumrequired to stabilize the magnet conduc-tor is largely determined by heat transferto the helium bath. The maximum designsurface heat flux (often the recoveryheat flux in a boiling experiment) mustnot be exceeded in steady state. AtWisconsin we have been evaluating theheat transfer behavior of superfluidhelium to compare its performance withnormal pool boiling at 4.2 K. Plottedin Fig. 6 are heat transfer and boilingcurves for superfluid and normal fluidhelium. For normal helium at 4.2 K, therecovery heal flux is around 0.2 W/cm2.13

Substantially better heat transfer isobserved in superfluid helium where therecovery heat flux is about 0.7 W/cm2 at1.9 K and 0.02 atm (saturated vaporpressure). Further enhancement in therecovery heat flux is achieved by pres-surizing to around 1 atm, whererecovery is observed to occur at1.9 W/cmz.14'lb This improved heattransfer means that less normal metal isrequired to stabilize the conductor thanwould be needed in normal helium, furtherreducing the overall conductor cost.

STRUCTURE

The best materials for the structuralmember in an energy storage magnet appearto be glass fiber reinforced composites.

AT(K)

Fig. 6. Heat transfer curves for super-fluid and normal fluid helium.

This choice is principally due to theirhigh strength to thermal conductivityratio (a/k), which is a figure of meritfor cryogenic structural materials. AtWisconsin, we have been investigatingcommercially available glass fiber rein-forced composites to determine which wouldbe best for a strut. This survey has in-volved compressive testing at room as wellas cryogenic temperatures. Both ultimatestrengths and cyclic fatigue data have beencollected on five different compositematerials.

Ultimate compressive strength dataat room temperature, 77 K and 4.2 K areshown in Fig. 7. In general, th'.- uni-directional composites are stronger thanthe cloth reinforced. Strengths of theunidirectional composites are typically100,000 psi at room temperature in-creasing to nearly 200,000 psi at 77 Kand below. Cloth reinforced materialsare roughly half as strong ranging from50,000 to 100,000 psi ultimate strengthat reduced temperatures.16

An important observation of thisdata is that the strengths of materialsare relatively independent of whetherthe matrix is epoxy, polyester or vinylester. This effect is important sincethe cost of polyester is substantiallylower than epoxy. In addition, thepolyester is more durable under cooldownbeing less inclined to crack. Thus,based on the ultimate strength measure-ments it appears that glass reinforcedpolyester is the best choice for thestrut.

250

ox

2 0 0

150

100

0

!« *

-a <=>I D~ • ,

" OEP0XY- CLOTH

A POLYESTER A,ZopOLYESTER A,- V POLYESTER B- O VINYL ESTER

-

V m

§ :• • "

ROD 1ROD 5

i i

=115

- 10

- 5

O"5

0 100 200 300TEMPERATURE (K)

Fig. 7. Average ultimate compressivestrengths versus temperature for fiberreinforced composites.

Since the energy storage magnetvaries its loading over the diurnalcycle, it is necessary that the strut bedesigned to withstand the 10b cyclesthat occur during the lifetime of aunit. Therefore, cyclic fatiguemeasurements have been carried outon the glass-fiber reinforced polyester.Results of these measurements are shownin Fig. 8 which is a plot of the peakstress versus median fatigue life of thesample. The allowable peak stress underlO1* cycles is approximately 70% of theultimate strength for both the roomtemperature and cryogenic temperaturedata.17

10' 10* 10°MEOIAN FATIGUE LIFE (CYCLES)

Fig. 8. Peak compressive stress versusmedia fatigue life for glass/polyestercomposite.

Future activities in the structuralsupport research and development programare to test model struts under compressiveand shear loads. Initially the workwill be carried out at room temperature,with later modifications allowing forcooling the strut to low temperatures.Half scale struts will be tested up to1,000,000 lbs total load.

CONCLUSION

The principal achievements during thefirst two years of the component develop-ment phase of the Wisconsin Superconduc-tive Energy Storage project have beensummarized above. Activities in thedevelopment of the superconductor com-posites conductor have shown manufactur-ability of a fully stable aluminum-NbTicomposite to be cooled by 1.8 K superfluidhelium. Structural support developmenthas tentatively selected glass rein-forced polyester as the strut material.Testing of the conductor and strut areplanned in the near future. The ulti-mate goal of these activities is todevelop manufacturing equipment inindustry for producing the components ofan energy storage magnet.

REFERENCES

1. R.W. Boom, paper V-l, this con-ference.

2. Wisconsin Superconductive EnergyStorage Project, Vol. I, EngineeringExperiment Station, University ofWisconsin, July 1974.

3. Wisconsin Superconductive EnergyStorage Project, Vol. II, Engi-neering Experiment Station, Uni-versity of Wisconsin, January 1976.

4. Wisconsin Superconductive EnergyStorage Project, Annual Report,Engineering Exper:ment StationReport No. 47, University of Wis-consin, May 1977.

5. S.W. Van Sciver and R.W. Boom inProceedings 21st Midwest Symposiumon Circuits and Systems, pp. 615-619 (1978).

6. H.A. Peterson, R.W. Boom and W.C.Young in Proceedings of WorldElectrotechnical Congress, June21-25, 1977, Moscow.

251

7. R.W. Boom, G.E. Mclntosh, H.A.Peterson and W.C. Young, Advancesin Cryogenic Engineering 19, 117(1976).

8. R.W. Boom, B.C. Haimson, G.E.Mclntosh, H.A. Peterson and W.C.Young, IEEE Trans, on Magnetics,HAG 11 , 475 (1975).

9. S.G. Ladkany, Advances in Cryo-genic Engineering, 24 ( to be pub-l ished) .

10. Z .J .J . Stekly, Journal of AppliedPhysics 37, 324 (1966).

11. P.E. Hanley and M.N. B i l t c l i f f ei n Proceedings Fourth Interna-t ional Cryogenic EngineeringConference, p. 224-226 (1972).

12. M.A. Hi lal and R.W. Boom,Advances in Cryogenic Engineer-ing 22, 224 (1977).

13. D.N. Lyon, Advances in CryogenicEngineering J£, 371 (1965).

14. S.W. Van Sciver in Proceedings7th Symposium on EngineeringProblems in Fusion Research,pp. 690-694 (1977).

15. S.W. Van Sciver, Cryogenics JjS,415 (1978).

16. E.L. Stone and W.C. Young, re f . 14,pp. 1510-1514.

17. E.L. Stone, L.O. El-Marazki, andW.C. Young in Nonmetallic Materialsand Composites at Low temperatures(to be published).

252

Project Title:

PROJECT SUMMARY

Power System Stabilization Using Magnetic Energy StorageDynamic Characteristics of the BPA System

Principal Investigator: R. L. Cresap

Organization: Bonneville Power AdministrationP. 0. Box 3621Portland, OR 97208503/234-3361 Ext. 4419

Project Goals: To determine the dynamic characteristics of the BonnevillePower Adnini strati on (BPA) electric system in order toestablish the feasibility of using a superconductingmagnetic energy storage (SMES) unit to damp poweroscillations in the western U. S. power system.

Project Status: Field tests on the BPA system with the dynamic brakewere conducted to determine the dynamic characteristicsof the western U. S. power system. The analysis of thetest data and the operating experience with v*e modulationof the DC intertie show that a small magnetic energystorage unit will provide damping of power oscillationsfor the pacific AC intertie. The AC intertie powervariation together with the system transfer functionallowed the sizing of a SMES unit. It was found that a30-MJ/10-MW unit is adequate for damping 99% of the ACintertie power oscillations. The project has beencompleted.

Contract Number: None

Contract Period: None

Funding Level: Not Applicable

Funding Source: BPA (in-house)

253

POWER SYSTEM STABILITY USING SUPERCONDUCTING MAGNETIC ENERGY STORAGEDYNAMIC CHARACTERISTICS OF THE BPA SYSTEM

R. L. Cresap and J. F. Hauer

Bonneville Power Administration

P.O. Box 3621

Portland, Oregon 97208

ABSTRACT .

Modulation of the Pacific HVDC Intertie has shown that a small amount of control

can provide damping for the western system, which has a history of negatively damped

oscillations. Because an extended outage of the DC Intertie could reduce the capa-

bility of the AC Intertie an alternate source of damping is desirable. This could be

done with a small special-purpose SMES unit. In order to establish the size of the

unit, field tests were conducted to determine the dynamic characteristics of the

western system. Analysis of the data indicates that, under normal operating conditions,

a 10-MW device would provide damping equivalent to dc modulation.

INTRODUCTION

The western power system has a long

history of negatively damped syn-

chronizing oscillations. This tendency

to oscillate is a factor which can

significantly reduce the maximum trans-

fer capability of the Pacific AC Inter-

tie. Because surplus hydro-generation

in the Northwest is often used to dis-

place oil-fired steam generation in the

Southwest it is important that this

Intertie have the maximum possible

rating.

Modulation of the Pacific HVDC Intertie

has shown that large amounts of damping

can be achieved by control of an ac-dc

converter . The availability of this

additional damping was a key factor in

permitting a 400 MW uprating of the

Pacific AC Intertie. The possibility

exists that loss of the DC Intertie, such

as occurred in 1971 when an earthquake

partially destroyed the Sylmar Converter

station, could cause a reduction of the

AC Intertie transfer capability.

Due to the use of an ac-dc converter,

SMES has excellent dynamic response

characteristics. Experience with dc

modulation has shown the amount of

control necessary during normal opera-

tion of the system to be quite small

254 -.

(5 MW peak to peak). As a result it

appears possible to provide an alternate

source of system damping using a small

special purpose SMES unit.

DESCRIPTION OF THE

WESTERN POWER SYSTEM

The western power systen, as comprised

by the Western Systems Coordinating

Council (WSCC), includes all or part of

the 14 western states, and British

Columbia. These interconnected systems

operate about 75,000 miles of trans-

mission lines, 115 kV and higher, and

have a total installed generating capa-

city of about 92,500 MW.

The WSCC system can be characterized by

its diversity. The Pacific Northwest is

a winter peaking region with predomi-

natingly hydro generation, while the

Pacific Southwest is a summer peaking

system which has primarily oil-fired

steam generation. In order to take

advantage of this seasonal load diver-

sity and the availability of surplus

Northwest hydro generation, the northern

and southern portions of the WSCC

system are connected by a system of

interties. The 500-kV Pacific AC Inter-

tie is the primary alternating current

interconnection between the Northwest

and Southwest. This Intertie is com-

posed of two lines with a total rating

of 2,500 MW. A system of 230-kV and

345-kV ties extending around the eastern

side of the system also connects the two

regions. In addition to the ac inter-

ties, the + 400 kV bipolar, direct

current Pacific HVDC Intertie provides a

1,400-MW interconnection between the

Pacific Northwest and the Los Angeles

area. Figure 1 shows the Pacific AC and

Pacific HVDC Interties.

Portion" - , j

Pacific

Ocean

\\ \OBEOON

^ S S \ \ NEVADA

Fig. 1 Pacific Northwest-Pacific

Southwest intertie system.

During the spring and early summer high

stream flows in the Northwest dictate a

heavy export of surplus hydro generation

to the Southwest. Because this dis-

places high-cost oil-fired steam genera-

tion in the Southwest, large economic

and energy conservation benefits are

realized. At other times, this system

of interties provides a capability to

exchange off-peak energy.

In order to enhance the transient sta-

bility of this interconnection, BPA

installed a 1,400-MW braking resistor in2

its system . During a large transient

swing the brake is applied to the system

for 30 cycles, thereby decelerating

Northwest generation. Stability studies

have shown that this improves the trans-

ient stability of the north-south inter-

connection by about 900 MW.

255

CAUSES OF POOR SYSTEM DAMPING

Traditionally, the stability of power

systems has been assessed in terms of

maintaining synchronism between the

various parts of the system, both in

steady-state (steady-state stability)

and following a severe disturbance such

as a 3-phase fault (transient stabil-

ity). Usually, if the system survives

the first swing, the transient subsides

because of the natural system damping

resulting from machine windings, tur-

bines, loads, etc.

Control systems on generators can lead

to another type of instability. Norm-

ally, in tightly-connected systems, the

frequency of electromechanical swing

modes range from about 1 Hz to 2 Hz and

are usually well damped. However, when

large systems are connected by long,

relatively weak interties, lower fre-

quency modes result. The response of

generator controls to the synchronizing

swings associated with these low-

frequency modes can produce sufficient

negative damping to cancel the natural

damping of the system. When this hap-

pens, oscillations of increasing ampli-

tude will occur (dynamic instability).

Although a system could operate beyond

its transient stability limit provided

no . severe disturbances such as 3-phase

faults occur, it could not operate

beyond its dynamic stability limit. The

constantly occurring small changes in

operating conditions (loads, voltage

levels, etc.) would give rise to

oscillations of increasing amplitude.

The primary source of negative damping

of the one-third Hz swing mode associ-

ated with the Pacific AC Intertie is the

response of generator excitation sys-

tems. The negative damping occurs when

excitation systems correct changes in

machine terminal voltages caused by

swings in power angle between the inter-

connected systems . These fluctuations

in excitation cause sizeable negative

damping torques to be exerted on the

rotating masses. Other less important

sources which contribute to the negative

damping of this swing mode include the

high-frequency response of hydro gover-

nors and certain types of loads.

Figure 2 shows a typical example of a

negatively damped oscillation.

1820 •

2 1780 -

0 10 20 30 40 50 60 70 80 90 100 110 120

TIME IN SECONDS

Fig. 2 Negatively damped Pacific AC

Intertie oscillation

SIZING THE SMES UNIT

Because the rating of the SMES unit

depends on the size of the synchronizing

swings to be damped, the types of dis-

turbances which occur in a power system

need to be considered. Disturbances

range from random load changes and

256

switching of lines and shunt compensa-

tion occurring during normal operation,

to large disturbances such as faults and

loss of major generators. Experience

with dc modulation has shown that,

during normal operation, strong damping

of the swing mode associated with the

Pacific AC Intertie can be accomplished

by small changes in dc power. However,

transient stability studies indicate

that to provide the same degree of

damping for swings caused by serious

faults would require very large changes

in dc power.

Fortunately, large disturbances occur

infrequently, and system nonlinearities

cause large swings to be better damped

than small ones. In any case, it would

be prohibitively expensive to provide

sufficient capacity in a dedicated SMES

unit to strongly damp a large swing. On

the other hand, random load changes

occur constantly and the resulting

swings must be damped.

As a result, a criterion was used to

size the SMES unit that would insure

damping during normal operation. Be-

cause of the stochastic nature of

Intertie swings, caused by random load

variation, it was decided that the

rating of the SMES converter should be 3

standard deviations, or be inside its

limits 99.7-percent of the time.

While experience with dc modulation

provides some insight, it cannot be used

directly to establish the rating of a

SMES unit. The response of a system

depends on the location of the input.

In general, the response of the system

to modulation of a SMES unit would be

different than its response to dc modu-

lation. As a result, the rating of the

SMES unit required determination of a

system transfer function obtained from a

test, design of a suitable controller,

am. measurements of the statistical

characteristics of load variations.

LOCATION OF THE SMES UNIT

AND DETERMINATION OF A

TRANSFER FUNCTION

The efficient control of a swing mode

requires that the associated torques

exerted on machine rotors be maximized.

As a result, the best location for

control of the AC Intertie swing mode is

near the center of northwest generation.

Transient stability studies to determine

a site for the dynamic brake showed that

a good location was Chief Joseph Sub-

station in North Central Washington.

Favorable operating experience with the

brake, together with its availability to

perform the testing needed to establish

a system transfer function, also made

Chief Joseph an attractive location for

the SMES unit.

A 0.5-second brake application produced

the AC Intertie response shown in

Figure 3, and the Bode plot shown in

Figure 4 was obtained from a Fourier

analysis of this data. A transfer

function was fitted to the frequency

domain data using a criterian which

included both gain and phase . Fitting

was done in the frequency domain rather

257

TUM (N HCONOS

Fig. 3 Time response of AC Intertie

phase current to Chief Joseph brake

application.

Fig. 4 Frequency response of AC Inter-

tie to Chief Joseph brake application.

than the time domain because accurate

gain and phase replication was needed

over a wide frequency range in order to

design the compensator. Complex values

for the poles and zeros are listed in

Table 1.

The poles describe the damping and

frequency of the modes of the system,

while the zeros provide the proper model

composition. The first two modes are

believed to be due to the dynamics of

hydro-plant water passages, and the

composite response of speed governors.

Table 1. Poles and zeros for AC Inter-

tie response to Chief Joseph brake

application.

Poles Zeros

-0.295

-0.422 + jO.374

-0.254 +j2.214

-0.445 + J3.560

-0.519+ j4.989

-1.123+ j6.650

-1.570 +j7.338

-0.788 + J9.089

0.084

-0.336 + jO.782

-0.386+ j3.240

-0.365 + J4.051

-1.135 + J5.700

-25.484 +j7.294

-0.134 +j7.670

-0.456 + J9.168

Gain = 0.01065 AC Intertie Phase Amps/

Brake MW

The third pole is the AC Intertie swing

mode. Remaining poles are needed to

describe the multitude of high frequency

swing modes associated with tightly

coupled groups of machines. This trans-

fer function has the time domain re-

sponse shown in Figure 3. Because

"Hanning" was used to obtain smooth

frequency response data, the damping of

the transfer function response is some-

what greater than the measured response.

COMPENSATOR DESIGN

To get maximum benefit from the SMT'f;

unit it is necessary that control sig-

nals have the proper gain and phase

relationship to AC Intertie swings.

Because the objective is to damp a

specific swing mode, it is desirable

that components of the control signal

due to other swing modes and noise be

minimized. These objectives were suc-

258

cessfully achieved with dc modulation.

As a result, it was anticipated that a

similar control system would be used for

the SMES unit.

Figure 5 shows the compensator used for

dc modulation . For the SMES system the

input to the compensator would be phase

current magnitude derived from fast

responding current transducers located

at the northern terminal of the AC

Intertie. This signal would be differ-

entiated and telemetered to Chief

Joseph. The low-pass filter is used to

control bandwidth, and the lead compen-

sator provides additional signal shap-

ing. High-frequency swing modes (1 Hz

to 2 Hz) are removed by the notch fil-

ter. The output is applied to the

regulator of the SMES converter.

•2 • V • -2m

' " " L D 1 "LP- \tt-

"LDl

"IB

:o.o

20.0

o.o

1.5

50.0

50.0

I.a

10.0

LOW-PASS DIFFERENTIATOR CAINFILTER

rldluu/tacond

ndlus/sacond

radlws/stcond

rsdlans/second

^rsdUns/sacond)*

( r i d l . n i 2

seconds-1

Fig. 5 DC modulation compensator.

The SMES unit was assumed to have an

instantaneous response over the fre-

quency range of interest. The system

transfer function, listed in Table 1,

was pessimistically modified by placing

the AC Intertie pole pair on the ima-

ginary axis of the s-plane (undamped).

Also the zeros closest to these poles

were moved to the imaginary axis. The

compensator parameters employed for dc

modulation were used as the starting

point for determining values for the

SMES system. Figure 6 shows the cor-

responding root-locus where the sub-

scripts c and s have been used to dis-

tinguish compensator poles and zeros

from those of the system. The "best

gain" was choosen to maximize the damp-

ing of the AC Intertie mode. An exten-

sive parametric study showed that, with

a suitable gain change, dc modulation

compensator values were also best for

the SMES system. This is due to similar

system transfer functions aad a compen-

sator design which is insensitive to

parameter changes.

IVFRSF OECAV TIME CONSTANT IN SEC

Fig. 6 Root-locus of dc modulation

compensator contolling SMES unit.

C - Compensator dynamics

S - System dynamics

0 - Best gain (0.46)

259

NOiSE CHARACTERISTICS OF THE

AC INTERTIE

Fourier analysis of AC Intertie power

variations, measured under a wide range

of operating conditions, provided spec-

tral characteristics of the response of

the AC Intertie to random load switch-

ing. Figure 7 shows a typical spectrum

with HVDC modulation off, fitted with a

5th-order model (Table 2). At low

frequencies the effect of random load

switching is evident, producing the

initially linear decline with log

frequency like the spectrum for integrated

white noise. The peak at 0.35 hertz is

amplification of load switching noise

by the AC Intertie swing mode. The

activity at 0.17 hertz is of undetermined

origin. However, sustained oscillations

have recently been observed at that

frequency.

Table 2. Poles and zeros for spectrum

of AC Interie power variations.

a1 is-

'°1

Poles Zeros

0 -0.167 + j0.996

-0.136+ jl.028 -0.137 + jl.854

-0.290 + J2.266

Gain = 5.8

RATING OF THE SMES UNIT

The AC Intertie noise spectral informa-

tion, together with the system transfer

function and compensator, permits estima-

tion of the converter rating of the SMES

unit. For analysis purposes noise

spectra can be modeled as the output of

a transfer function T (s) with white

noise as the input. Referring to

Figure 8, modeling the feedback effect

of the SMES unit requires conversion of

the noise spectra to be an input at

Chief Joseph. Tg(s) is the system

transfer function listed in Table 1 with

a gain change to convert A I._ to

APA(;. Tg(s) is Tg(s), possibly modified

to represent changes in system dynamics

such as making the Intertie mode nega-

tively damped.

Fig. 7. Spectrum of AC Intertie power Fig. 8 Model for sizing SMES unit,variations

260

This technique was tested by using noise

spectra obtained with dc modulation out

of service to predict the spectra which

would have been observed with

modulation in service. These predictions

were compared to later measurements made

with modulation in service, with excel-

lent results. Consequently, it appears

to be a highly accurate method for

determining the converter rating of the

SMES unit.

Figure 9 shows the effect of the SMES

unit on AC Intertie noise. Also shown

is the spectrum for the output of the

SMES unit. By Parseval's Thecrm the

area under the power spectrum of the

output of the SMES unit is its variance,

or standard deviation squared. Standard

deviations were calculated for a number

of noise spectra, obtained unaer a wide

range of operating conditions, and with

postulated changes in system dynamics.

Values ranged from 2.5 MW under normal

operating conditions, to 3.9 MW for

high system noise and assuming a nega-

tively damped Intertie mode. As a

result a 10 MW converter would provide

3-standard deviations for most system

conditions.

CONCLUSION

System measurements and operating

experience with dc modulation show that

a small special purpose SMES unit coma

provide danping for the western system.

Because the damping capability of dc

modulation would be lost in the event of

an outage of the DC Intertie, an alter-

nate source of damping is desirable.

20

noise at compensator output,with loop closed

FREQUENCY IN HERTZ1.0

Fig. 9. Noise Spectra

Also such an application would provide

valuable information about the relia-

bility and maintainability of super-

conducting equipment in a power system

environment.

While the primary purpose of SMES is

load-leveling, it is well suited to

control applications. Because stability

problems often impose serious operating

constraints, this capability can be

valuable. Early demonstration of the

control capabilities of SMES will add an

important dimension to the technology,

and may lead to the development of a

variety of special-purpose control

devices.

261

KEFERENCES

R.L. Cresap, D.N. Scott, W.A. Mittel-

stadt, and C.W. Taylor, IEEE Trans.

Power Apparatus and Systems. PAS-97,

1053 (1978).

2M.L. Shelton, P.F. Winkelman, W.A.

Mittelstadt, and W.J. Bellerby, IEEE

Trans. Power Apparatus and Systems.

PAS-94, 602 (1975).

3J.F. Hauer, BPA Report. (19?7).

4F.P. Demello and C. Concordia, IEEE

Trans. Power Apparatus and Systems.

PAS-88, 316 (1969).

262

PROJECT SUMMARY

Project Title: "Energy Storage Systems and Identification of PowerSystems Equivalents"

Principal Investigators: D. P. Carroll and D. M. Triezenberp

Organization: School of Electrical EngineeringPurdue UniversityWest Lafayette, IN 47907317-493-3813

Project Goals:

Project Status:

One major objective of this project is to investigatetechniques for identifying power system configura-tions, load magnitudes, and generation levels, fromtime series measurements taken at a limited number ofpoints in the system. The other major objective of theproject is to develop simplified system models and toexplore techniques for controlling energy storagedevices to improve the transient performance of powersystems. A significant part of this latter objectiveis the study of low power superconducting magneticstorage systems for damping controls.

In the research area of power system identification,algorithms are under development for multiple outputARMA process parameter estimation. Some success hasoccured in estimation of simplified model parametersusing data taken from detailed system simulations.

In the area of energy storage applications, simpli-fied analytical models have been developed and veri-fied for line-commutated and force-commutated powerconverters su-itable for interfacing batteries andsuperconducting magnets with power systems. Simplifiedand detailed computer simulations are being used tostudy the dynamics and control of candidate systemsinvolving energy storage. A preliminary design hasbeen completed for a damping control using a low powersuperconducting inductor-converter unit.

Contract Number: EC-77-S-02-4206.A000

Contract Period: Jan. 1977 - Dec. 1979

Funding Level: $517,818

Funding Source: Department of Energy

263

HYBRID COMPUTER STUDY OF A SMES UNIT FORDAMPING POWER SYSTEM OSCILLATIONS

P.C. Krause and O.M. TriezenbergEnergy Systems Simulation Laboratory

School of Electrical EngineeringPurdue University

ABSTRACT

The work conducted at Purdue University regarding the use of SMES to damp oscilla-tions in the BPA system is reported. This investigation was conducted in parallel withwork at LASL and BPA and serves as a verification of the work at BPA. In particular,the dynamic characteristics of the BPA system were simulated on the hybrid computerusing the response obtained from the Chief Joseph Dynamic Brake test. The SMES unit wassimulated using information supplied by LASL. The compensator was simulated using thedesign developed by BPA. The hybrid computer study illustrates the dynamic response ofthe BPA system with the SMES unit in service for sinusoidal variations in AC Intertiepower and for random variations of AC Intertie loading.

INTRODUCTION

In mid 1977 an existing DOE contractwith Purdue University was extended to in-clude an investigation of a possible ap-plication of SMES. In particular, thestudy considered damping of power oscil-lations on the West Coast North-South ACIntertie with an SMES unit and an AC to DCconverter. If this application proved tobe feasible, this inductor converter (IC)unit could provide backup for the PacificHVDC Intertie which is presently beingmodulated to damp the AC Intertie powerosci1lations.

Personnel from LASL, BPA and PurdueUniversity, with H.A. Peterson acting asa consultant, were involved in this inves-tigation. Important, to this study wasthe background of BPA personnel who haddesigned a modulation control for thePacific HVDC Intertie. Purdue's principalrole was to provide verification of tt ework being conducted at BPA [1]. This wasto be accomplished from a hybrid computerstudy of a simplified representation ofthe western system. The results of thisstudy are recorded in this paper.

SYSTEM STUDIED

The system studied on the hybrid com-puter at Purdue University is shown inblock diagram form in Fig. 1. The IC unitis represented from information received

from LASL, and the transfer functions G(s),H(s), and C(s) from information providedby BPA.

The IC unit is connected to the 230kV Chief Joseph Substation through a 12.75MVA transformer bank with 1 kV line-to-line voltage on both secondary transfor-mers (12 pulse converter). The inductance,L, Is 2.4 H and the maximum inductor cur-rent, l(c, Is 5000 A. The commutatingreactance, Xf, is 0.00784 ohms which is10% of base impedance. The harmonics areneglected in the representation of theconverter. That is, the output voltage ofthe converter, V,Q, and the inductor oroutput current, I\Q, are the average val-ues of the actual variables. The IC unitis equiped with an open-loop power controlwhich senses I|Q a pd calculates the nec-essary voltage, V r ef, to satisfy the de-sired or reference power, Pref. In thiscase the reference power is zero where-upon the signal from the compensator, Pc,determines the desired power. The refer-ence voltage and the output current, areused to calculate a control signal, ec,which establishs the firing angle of theconverter and thus V|r,. The power outputof the IC unit is denoted Pjr,. That isthe power consumed or supplied by the ICunit at Chief Joseph Substation.

264

From information obtained as a resultof an 0.5 sec. application of the dynamicbrake at Chief Joseph, 6PA established thetransfer function between Chief JosephSubstation injection and AC Intertie cur-rent magnitude [1]. This transfer func-tion, which is given in [I], Is G(s) withone exception; AC Intertie power P|f, Isused rather than AC Intertte current.Although not explained herein, this Isaccomplished by multiplying current Inamperes by 1.07 to obtain AC Intertiepower in HW.

whereR(x) = E{x(t)x(t-T)}

This random signal is generated by samplesof a random number generator held for 0.1sec. each, with the variance of the randomnumber generator corresponding to (2.5

Fig. 1 Block diagram of BPA system-transfer function approach

The transfer function H(s) providesa means of modeling the semi-random poweroscillations recorded on the AC IntertieMonitor. H(s) consists of three blocks inseries. The output, P^, is added to Pi-!-,the output of G(s), to yield the AC In-tertie power, Pjy. The input to the trans-fer function H(s) is a bandwidth limitedwhite noise with an average random powerof 0.25 (MW)2-second, uniform for fre-quencies in the range of 0 to 6 rad/sec.Average random power for the signal x(t)is here defined as

MW) 2. Three sets of data for H(s) werefurnished by BPA corresponding to thrrcdifferent inertie loads; 1680, - 1'sC and920 HW. The 1680 MW load (export) Is thecase considered In this study. For thiscondition H(s) is

Function

15.6s

265

1.138

s 2 + .3108s + 1.A31

s 2 + .9428s -i- 5.565s 2 + .AO53s + 5.395

The compensator design was performedby BPA [1]. The output signal of the com-pensator is added to the power referenceto establish the desired power output ofthe IC unit.

COMPUTER STUDY

A frequency response analysis was per-formed with a linearized representation ofthe converter, making the transfer ratioof P|T(S) to PN(S) equal to H(s)/(l-C(s)G(s)). From this analysis It was estab-lished that for a noise frequency of 2.29rad/sec. (about 1/3 Hz) the ratio Pp (s)/Pj-](s) has an amplitude of 2.1k. This fre-quency response calculation is verified infig. 2 wherein traces of Pn(t), PfJ(O,

P|C(t), PjT(t). P.T(t), P (t), V. (t), andl)c(t) are given for PN(t) a 1/3 Hz sinu-soid. In steady state the amplitude ofPN(t) is approximately 12 MW while thesteady state amplitude of P|T(t) is ap-proximately 5 MW, a reduction by a factorof 2.4.

Figures 3 and k show the system withPn(t) represented as a "bandwidth limitedwhite noise" with S(w) = 0.25 MW2-sec,uniform for w from 0 to about 6 rad/sec.The same mode of operation is depicted inboth figures. The recording scale is ex-panded in Fig. k for the purpose of portra-ing the noise input. In all the studiesperformed, the output of the compensatorPc did not exceed 10 MW and the convertervoltage, V|C, did not exceed 2.5 kV.

CONCLUSIONS

The material presented in this paperprovides a verification for analysis andrecommendations of BPA. Hence, the con-clusions of [1]. are essentially the con-slusions of this paper. Clearly, if thetransfer function between power at theChief Joseph Substation and the AC Intertieand if the variations in the AC Intertiepower are as used in this study then, asrecommended by BPA, a 30 MJ 10 MW IC unit(SMES) placed at Chief Joseph bus shouldbe sufficient to damp AC Intertie oscil-lations.

REFERENCES

[1] R.L. Cresap and J.F. Hauer, "PowerSystems Stability Using Super-conduc-ting Magnetic Energy Storage DynamicCharacteristics of the BonnevilieSystem," presented at the First AnnualMechanical and Magnetic Energy StorageContractors1 Information-ExchangeConference, Luray, Virginia, Oct. 2k-26, 1978.

ACKNOWLEDGEMENTS

The authors greatly acknowledge theassistance of H. Boenig, R.L. Cresap, andH.A. Peterson.

JtCCUCHART niliu'mtnl eytttmt DMllon

- . .

' • v i • • ! • • ! -i j I --); i I i . i J .i . t I

PICMW

j ! ; t ' ' ' I i ' ' '

..4 : .

t - '• - ' - +

r MftnMA/iAAMArtAMJ VvUVVv Vv v wvi/UVJvVv

•

• T • - "

25.0

P.T 0

-25.0 E-12.5

-12.5

.2.5

0

FI9. 2 Response of system sham In FI9. 1to 5fnuso|d«l noise Input,

267

25.0 p"IT 0 L

-25.0 L-

.2 .5

-...sL-^/lijuA

12.5 ,

?K>J\f,' "•

Fig. 3 Response of system shown in Fig. Ito bandwidor Hroftetf White noise input,

268

ACCUCHART OouM me.. InMnmwnt SyMMii

-12.51

25.0PIT oKW

-25.0

12.5f.T 0

-12.5

E-

12.5,Pc 0

-12.5 1

4-.. H-LHu

'-!•+•

t-ri^-:..i"r"

t:rL.L.i:tt"T.ri:rr-: :

•jf :ofIP....

• • • • * - * -+

Ftg. 4 Same as Fig. 3 with expanded time scale.

PROJECT SUMMARY

Project Title: Superconducting Magnetic Energy Storage (SMES)

Principal Investigator: John D. Rogers

Organization: Los Alamos Scientific LaboratoryUniversity of CaliforniaLos Alamos, NM 87545505/667-5427

Project Goals: The goals of the SMES program are two-fold, the first isto design, fabricate, and place into operation a 30-MJ,10-MW SMES unit for electric utility transmission linestabilization on the Bonneville Power Administration (BPA)system by 1982-83. The second goal is to design and haveconstructed a 10-to-50-MWh SMESunit. This unit willbe for a completely detailed engineering prototype demon-stration of an electric utility diurnal load levelingsystem.

Project Status: The conceptual engineering desiqn of the BPA SMES systemis complete. An engineering specification design for the30-MJ superconducting coil will be complete by 10-1-78,and an RFQ for fabrication design and actual coil fabri-cation will be issued within 60 days. The superconductingwire for the entire 30-MJ coil and 50- and 100-m lengthsof prototype 5-kA superconducting cable are on order. TheSCRs and sinks for the converter are on order. An RFQfor the converter fabrication will be issued soon.

An entirely new reference design for a 1-GWh diurnal loadleveling SMES unit is underway. Emphasis is being givento a better understanding of the dewar, the dewar supportstructure, rock mechanics, excavation details and costs,stabilization of the excavation walls, and high purityaluminum superconductor matrix costs.

Contract Number: W-7405-ENG-36

Contract Period: Continuing

Funding Level: $1,050,000 FY78

Funding Source: Department of Energy, Division of Energy Storage Systemsand Division of Electrical Energy Systems.

271

SUPERCONDUCTING MAGNETIC ENERGY STORAGE*

J. D. Rogers and H. J. BoenigLos Alamc Scientific Laboratoryof the University of California

Los Alamos," NM 87545

ABSTRACT

Superconducting inductors provide a compact and efficient means of storing elec-trical energy without an intermediate conversion process. Energy storage inductorsare under development for diurnal load leveling and transmission line stabilizationin electric utility systems and for driving magnetic confinement and plasma heatingcoils in fusion energy systems. Fluctuating electric power demands force the elec-tric utility industry to have more installed generating capacity than the averageload requires. Energy storage can increase the utilization of base-load fossil andnuclear power plants for electric utilities. Superconducting magnetic energy storage(SMES) systems, which will store and deliver electrical energy for load leveling,peak shaving, and the stabilization of electric utility networks are being devel-oped. In the fusion area, inductive energy transfer and storage is also being devel-oped by LASL. Both 1-ms fast-discharge theta-pinch and l-to-2-s slow tokamak energytransfer systems have been demonstrated. The major components and the method of op-eration of a SMES unit are described, and potential applications of different sizeSMES systems in electric power grids are presented. Results are given for a 1-GWhreference design load-leveling unit, for a 30-MJ coil proposed stabilization unit,and for tests with a small-scale, 100-kJ magnetic energy storage system. The resultsof the fusion energy storage and transfer tests are also presented. The common tech-nology base for the systems is discussed.

INTRODUCTION

The nondissipative dc-current carryingcapability and the low ac-current gener-ated losses of a superconductor place thedesign of high current inductors withinreach of useful application. Only in thelast 15 years has the superconductingtechnology been sufficiently advanced tomake the applications covered in this pa-per become practical. Only more recentlyhas energy transfer within a few secondsor less, into and out of a superconductingenergy storage coil been demonstrated.

Electric utilities experience periodicload variations on a seasonal, weekly, anddaily basis. The daily maximum and mini-mum loads of a power company typicallydiffer by about a factor of two. The poorload factor is an economic burden to theutilities because their installed cr acitymust be capable of meeting the peak demandand much of the generating capacity isidle during periods of low demand. Inex-pensive but inefficient units, such aspeaking gas turbines, are used to meet thepeak loads.

*Work done under the auspices of the DOE.

Energy storage units can be used tomeet the peak-power requirements and toabsorb excess available during periods oflow-power demand. To date, only pumped-hydro storage, with units up to 15,000Mwh, has been used very effectively.1Other energy storage technologies includechemical storage in the form of batteriesand hydrogen, thermal storage, compressedair storage, and magnetic storage.'-4Economic considerations eliminate iner-tial storage in flywheels for utilityapplications.

Superconducting magnetic energystorage has several advantages. SMESunits will have fewer site restrictions.Large SMES units can be constructed forstructural support in the rock formationsnear most large load centers, and exten-sive new transmission systems will not berequired. SMES units will have a re-sponse characteristic of less than a cy-cle to power system demand, which canimprove power system stability. With ahigh efficiency of 90% the energy isstored electromagnetically without anintermediate mechanical or chemical ener-gy state. The cost for a large SMES unit(10 GWh) is estimated to be about $30 to$35/kWh.

272 -

The LASL and the University of Wiscon-sin (UW) are developing SMES systems forelectric utility applications.5>6 Thesuperconducting coils for these systemsrange in size from small units a few me-ters in diameter and height, which willstore as litle as 30 MJ (8.3 kWh), up tolarge installations several hundred metersin diameter and height, which will storeas much as 10 GWh.

A technology development program forpulsed superconducting energy storage sys-tems for fusion applications has been un-derway at LASL since 1968. Both high-theta-pinch? and low-3 tokamak ohmic-heating°~l" systems will need nondissi-pative energy storage to achieve over-allpower balance. Liners, Z-pinches, lasers,and pulsed electron beam machines are ex-amples of fusion devices which requirelarge, fast energy delivery systems. Thetoroidal Reference Theta-Pinch Reactor(RTPR) would require about 60 GJ deliveredin 30 ms, the linear theta-pinch fusion-fission hybrid reactor needs about 25 GJin 2 ms,H and a liner reactor may re-quire about 10 GJ in 1 ms. The ohmic-heating coils in present US designs oftokamak experimental power reactors haveabout 1-2 GJ of stored energy, and thestorage currents must be reversed in 0.5to 2 s to induce plasma current.12-16

Feasibility experiments for MagneticEnergy Transfer and Storage (METS) systemswith 1-ms discharge from 300-kJ to 540-kJsuperconducting coils have been success-fully demonstrated for delivery of energyto an adiabatic theta-pinch plasma com-pression coil for fusion.1? Pulsed en-ergy simulation of both the tokamak plasmaohmic-heating and burn cycles has alsobeen demonstrated with a superconductingenergy storage coil and a dc-commutatedmechanical capacitor.18

SUPERCONDUCTING MAGNETIC ENERGYSTORAGE SYSTEM DESCRIPTION

The SMES coil is immersed in a liquidhelium bath in a dewar, which keeps itsuperconducting at a temperature below 4.5K. A closed-cycle refrigeration systemcools and liquifies the boiloff helium gasand returns it to the liquid bath. Foreconomic reasons, the inductor is a shortsolenoid, a coil with a ratio of height todiameter of about 1/3. A transformer anda converter connect it to a 3-phase utili-ty bus and regulate the power flow. Dur-ing the charge phase of the energy storage

cycle, the converter rectifies ac powerto dc for charging the coil. Stored en-ergy is returned to the utility bus forpeak-load demands by operating the con-verter as an inverter. Commerciallyavailable thyristors are used as theswitching elements in converters. Fig. 1shows the SMES system components in blockdiagram.

A full-wave Graetz bridge, as shownin Fig. 2, is the fundamental buildingblock of a line-commutated converter.Phase-angle control of the thyristors inthe converter determines the dc-outputvoltage. Such a converter requires reac-tive power from the ac bus during bothmodes of converter operation. A reactivepower compensation network, such as acapacitor bank, a synchronous condenser,or a static, reactive-power controllingdevice is needed to provide power factorcorrection.

SMES APPLICATIONS IN ELECTRICUTILITY SYSTEMS

The types of energy storage systemsfor the utilities may be separated ac-cording to the duration of the load vari-ation. On a seasonal basis the utilitiestypically use some form of fuel storageto meet the winter or summer peak load.The daily and weekly load variations aremet by pumped-hydro storage, gas tur-bines, and old, inefficient fossil-firedpower plants. At present, the short-termload variations are met by adjusting thepower output of one or more power plantson the system. Each of these short-termload variations is discussed below, and aSMES unit which might meet their powerand energy requirements is described.

LOAD LEVELING

Power demands are generally met by acombination of three or more types ofpower generation, including the base loadgeneration consisting of the more effi-cient fossil fueled or nuclear powerplants; the intermediate load generation(midr^ige peaking), consisting of older,smaller, less efficient fossil-fueledplants anu energy storage units; and thepeak-load generation,'consisting mainlyof gas turbines and energy storageunits. A representative weekly loadcurve, which is for the Michigan ElectricCoordinated Systems is shown in Fig.

273

Threepferse

bus

Reactivepower

M F ipC nSQi IQfl

Traraformer Smarter Siper conduct inocoil

SCRfiringcircuit

Controller

1Coil

protection

LHe

Wigeratw

I Control sicnoHo/fromelectrical power system

Fig. 1. Components of a superconducting magnetic energy storage system.

3-PHASEAC SYSTEM COIL

TRANSFORMER GRAETZ- BRIDGE

Fig. 2. Schematic of a full-wave, 6-pulse Graetz bridge.

14,000

12.000

§10,000

~ 8,000

O 6,000

4,000

I2M I2M I2M I2M I2M I2M I2M I2MI2N I2N I2N I2N I2N I2N I2N

SUN MON SUN WED THU FRI SAT

Fig. 3. Typical weekly load distribution for an electric utility.

A SMES unit with the same capacity asthe pumped-storage unit in Ludington, MI,which has a storage capacity of 15 GWh anda power capacity of 2076 MW, would be asolenoid about 340-m diam and 114 m high.Whereas Ludington cost $351 x 106 in1973 (or $503 x 106 in 1978 based on 7%per year inflation), the estimated cost ofan equivalent SMES unit is about $480 x106.

The superconducting coil could be con-structed several hundred meters under-ground in solid rock, which acts as astructural material to contain Lhe magnet-ic forces on the coil. For a 10-GWh unita shield coil with a radius of about fourtimes that of the main coil would be con-structed just below ground level to reduce

the magnetic fields in the vicinity of aSMES unit. It is possible to store largeamounts of energy in a relatively smallvolume because superconductors allow highmagnetic fields. For example, theLudington plant occupies about 10(km)2. The equivalent SMES unit wouldrequire only 1.5 (km)29 including allthe land area within a shield coil.

SYSTEM STABILITY AMD SHORT-TERM LOADVARIATIONS

The output of power generationplants must be adjusted to balance randomand periodic load variations, such asthose caused by steel rolling mills, arcfurnaces, etc. An energy storage unit

274

capable of leveling these short-term powervariations would be of great value to autility. The SMES units for diurnal loadleveling might have a converter capacityof 1000 to 2000 MW. As the response timeof the converter to a power demand is lessthan a cycle, it will be possible to meetthese short-term power demands by varyingthe power in the converter by a few per-cent. This particular function could alsobe satisfied by a small SMES unit thatstored only 100 to 500 MJ which had a con-verter capable of delivering only 20 to 50MW.

Occasionally, load variations and thesubsequent generation response cause anelectrical power system to become unsta-ble. System instabilities can be avoidedby limiting the load variation, by chang-ing the electrical characteristics oftransmission lines, by reducing the timeresponse of the generation plants, and/orby providing system damping. One specificlocation where an energy storage devicemight improve the stability of a powersystem is on the intertie between thePacific Northwest and southern Califor-nia. Two ac lines and one dc line trans-mit power along this corridor. Undercertain conditions an instability ariseson the ac line.20 This instability hasbeen overcome by installing a feedbacksystem that controls the converter powerat the northern terminal of the dc line,thereby damping the power oscillation onthe ac line. This solution is not com-pletely satisfactory as the power flow onthe ac line depends on the dc line workingproperly. If the dc line fails, the powerflow on the ac line should increase totake up the load, rather than decreasebecause of reduced stability.

A small SMES unit, storing 30 MJ andhaving a 10-MW converter, could damp theoscillations which occur at a frequency ofabout 0.35 Hz. The fast response of aSMES unit should improve system stabilityand provide for spinning reserve^! (dis-cussed below). Table I shows the majordesign parameters for the 30-MJ stabiliz-ing unit. Analysis of the Bonnevi^ePower Administration system by P. C.Krause at Purdue University has recentlyconfirmed the earlier work of Cresap andcoworkers at BPA.setting the energy stor-age, power rating, and control behavior ofsuch a system. Much of the technologybase for the 30-MJ coil has been estab-lished as a part of the fusion programpulsed inductive energy storage work.

100.359.10304.92.82.24.5

1502.5

1.51.2

0.42

MWHzMJMJkATkVK

WHmmm

Table 1. Design Parameters of a 30-MJSMES System Stabilizing Unit

Maximum power capcityOperating frequencyEnergy exchangeMaximum stored energyCoil current at full chargeMaximum field at full chargeMaximum coil terminal voltageCoil operating temperatureCoil lifetimeHeat load at 4.5 KInductanceMean coil radiusCoil heightWinding thickness

Spinning Reserve. The electric powerutilities are required to have a minimumspinning reserve capacity which amountsto about 10% of the load or 1.1 times thelargest generation unit on line. Addi-tional converter capcity on a large SMESunit may substitute for the spinning re-serve. During the periods of low-powerdemand, spinning reserve on the system isachieved through the ability of the con-verter to change from charge to partialcharge or discharge in less than onecycle. During the times when the unit isneither charging nor discharging, theSMES unit will be a substitute for spin-ning reserve. During the longer periodsof high-power demand the capacity of theunit will normally only be used at afraction of the maximum converter capa-city. The excess discharge capacity ofthe system may then take the place of thespinning reserve for the utility system.

A 1-GWh SMES SYSTEM DESIGN

The LASLdesign for a 1load leveling,of a referencestarting pointdesigns. Someunit are given

is developing a reference-GWh SMES unit for diurnal

One of the major purposesdesign is to provide afor detailed engineeringof the parameters of thisin Table 2.

ENERGY STORAGE COIL AND SUPPORT STRUCTURE

The coil is a thin-walled, 132-m-diam, 44-m-high solenoid as shown in Fig.4. The size and shape are the result ofa cost optimization and the dimensionsare determined by the maximum field.

275

Table 2. SMES 1-GWh Reference Design En-ergy Storage Coil Specifications

Energy stored at full charge3.96 x 10 1 2 J (1.10 GWh)

Energy stored at end of discharge3.36 x 10 1 1 (0.1 GWh)

Current at full charge 50 kACurrent at end of discharge 15 kAMaximum power output or input 250 MWTerminal voltage "to provide Pmax at end

of discharge 16.7 kVInductance 3.17 kHMaximum field at conductor at full charge

4.5 TOperating temperature 1.85 KMean coil radius 66 mCoil height 44 mCoil radial thickness 0.30 m

allowable stress in the conductor, about70 MPa (10,000 psi), then they are trans-mitted through struts to the rock, thecoil is placed below the surface of theearth where the compressive stresses inthe rock can be used to offset the mag-netic loads of the coil.

CONDUCTOR

Superconductor for a SMES coil mustbe reliable (this includes but is notlimited to stability considerations),must cost as little as possible, must becapable of being fabricated with existingtechniques or extensions of those tech-niques, and must be flexible enough to bewound into a coil in a 3-m-wide tunnel.

Fig. 4. Artist's concept of large SMES unit constructed underground.

The magnetic forces must be containedby rock to reduce the cost of the system.If stainless steel bands were used in aSMES coil, their cost alone would far ex-ceed the cost of other types of storagesystems. A set of struts and rods is re-quired to transmit the forces from thecoil at 1.8 K to the rock at about 300 K.The stresses and deflections associatedwith the thermally induced contraction ofthe coil during cooldown and the magnetic(Lorentz) forces on the conductor aretaken up by a short length curvature inthe dewar and conductor repeated on aboutl-to-2-m centers. Axial loads are al-lowed to accumulate until they reach the

Operation at 1.8 K rather than at 4to 6 K and the use of NbTi rather thanNb3Sn keep the total system cost low.The use of high purity aluminum insteadof copper as the current stabilizer isless costly and reduces the size of theconductor. A 1-GWh SMES unit is estima-ted to cost about $80 million or $80/kWh.

To fabricate the conductor with ex-isting techniques, a design has been cho-sen in which the NbTi is extruded in cop-per and the aluminum is added in subse-quent fabrication steps. To meet theflexibility criterion, a conductor designis considered in which several insulated

276

subconductors are in parallel electricallyand are cabled to reduce hysteretic losses.

CONVERTER

Line-commutated, solid-state convert-ers are used extensively in high-voltage,dc-power transmission. In comparison,converters for large SMES systems willhave medium to high voltage and currentratings. A 1-GWh energy storage capacityand a 4-h charging or discharging periodwill require a power of 250 MW. Becauseof the purely inductive load and the re-quirement that the maximum power be avail-able at all operating currents, the maxi-mum voltage and maximum current do notoccur at the same time. Thus, the con-verter has to be designed for a powergreater than the maximum power flow everexpected through the converter. For the1-GWh unit the voltage rating is 16.7 kVand the current rating is 50 kA.

Phase-controlled converters generateharmonics and absorb reactive power.

Figure 5 shows one possible circuitconfiguration of a converter. To reducethe reactive power requirement, the con-verter is designed as a series connectionof four 12-pulse modules each with itsown power transformer. Each module isdesigned for 4.5 kV and consists of two6-pulse bridges connected in parallel byan interphase reactor which balances thecurrent flow in the two bridges. Twomodules have a current rating of 50 kA,and the two remaining modules have a cur-rent rating of 30 kA and 20 kA. At maxi-mum coil current, each 6-pulse bridgewill provide 25 kA dc. Each 12-pulsemodule can be bypassed by a mechanicalswitch when the module voltage is zero,and then the module can be disconnectedfrom the 3-phase bus. This improves theoverall converter efficiency by removingthe forward voltage drop of at least fourseries-connected thyristors.

The installed converter power ratingand the converter cost can be decreasedby designing those modules which are

3-PHASE BUS

AC SWITCH

TRANSFORMER

6-PULSEBRIDGE

INTERPHASEREACTOR

DC SWITCH

SUPERCONDUCTINGMAGNET

1

Fig. 5. Four series connected 12-pulse converter modules forming a converter.

Advanced converter circuits are used tominimize these unwanted effects. The har-monic content of the ac-line current isreduced by using 12-pulse or 24-pulse mod-ules. Tuned filter networks can removethe remaining harmonics. The reactivepower requirement can be reduced by sub-dividing the converter into several seriesconnected modules. During operation, thephase-delay angles of all but one moduleare kept at 0°. That one module has aphase-delay angle that depends on thevoltage requirement. All those convertermodules at 0° require only a small amountof reactive power caused by commutation.

switched out of the circuit first (i.e.at a low current) for the current atwhich they are switched off, rather thanfor maximum current. Theoretically, ifthere were an infinite number of con-verter modules, the converter could bedesigned for a ratingPmaxtl + 1" ( W W

SMALL-SCALE SMES SYSTEM TEST

A small scale system, which includedall the components shown in Fig. 1 exceptthe refrigerator, was tested to evaluatethe electrical characteristics of a SMESunit.22 The system contained a 70-Hsuperconducting coil built by stacking

277

eight 3000-turn coils in series. Thequench current above which the 70-H coillost its superconducting property was45 A. A 12-pulse, solid-state converterand a power transformer with a 6-phasesecondary winding interfaced the coil tothe 3-phase laboratory bus. The maximumconverter output voltage used in the ex-periment was 150 V. The control systemfor the model SMES unit was designed withall the features necessary for the auto-matic operation of a large SMES unit onthe utility bus.

The total system was tested with dif-ferent power demands. The transition timefor rectifier-inverter and inverter-rectifier switching was measured to be 5to 6 ms. Figure 6 shows the coil current,coil voltage, and coil power for randompower demand. Because of the fast timeresponse of the converter, the coil powerfollowed the power demand closely. Theresponse to sinusoidal power demands wasmeasured at frequencies up to 30 Hz.

Reactor (SFTR),23 for adiabatic com-pression of a fusion plasma. A designoptimization study2** for the METS sys-tem'5 led to modular energy storagecoils of approximately 400-kJ size.These were to be charged in series anddischarged in parallel for nearly 10055efficient energy transfer.

The theta-pinch, METS system ischaracterized by the resonant circuit ofFig. 7. Coil charging is accomplishedwith a shunt switch external to thedewar. Discharge is initiated by openingthe shunt followed by opening of the HVDCinterrupter, B. The interrupter is coun-terpulsed to extinguish the arc by a cur-rent from the transfer capacitor, Ct,which has been back charged. Currentthen transfers to the compression coilwith a peak voltage across the circuitdeveloping at one-half the transferperiod. The energy is then trapped inthe compression coil by closing the igni-tron crowbar, IGcg, for the fusion

10 20 50 60 7030 40TIME(s)

Fig. 6. Coil voltage, current, and power responseto a random power demand signal (L=33H).

Mo control system instabilities were ob-served during the experiments with a su-perconducting coil.

INDUCTIVE ENERGY STORAGE FOR FUSION

THETA-PINCH MAGNETIC ENERGY TRANSFER ANDSTORAGE (METST

The METS inductive energy storage sys-tem was developed to deliver 488 MJ in 0.7ms to a 40-m radius toroidal theta-pinchsystem, called the Scyllac Fusion Test

burn cycle to be completed during the250-ms L/R decay constant of the loop.

The parameters for the METS coils made byLASL and Westinghouse are listed in Table3. The LASL coil tested successfully to12.5 kA and 386 kj of energy stored. Itwas pulse tested at 10 kA and 35 kV andwith transfer times as short as 1 ms inan L-C-L circuit as shown in Fig. 7.

The Westinghouse coil26 w as

operated with pulsed energy transfer at

278

300 M SUPERCONDUCTINGSTORAGE COIl

SHUNTSWITCH m i

I G C P

1 AUAIATIC PtASMA' COMPRESSION COIl

COUNTER-PULSE J

•OWBISUPPIY

Table 3. Parameters of 300-kJ EnergyStorage Coils

LAST WH~~Inductance, mH 4.87 6.05Resistance at 20 C, fi 0.0896 0.165Stored energy at 10 KA, kJ 244 302Length, cm 73.0 79.1Mean radius, cm 28.7 25.5Winding thickness, cm 0.508 4.74Number of turns 122.5 159.5Number of layers 1 4Central field at 10 kA, T 1.82 2.23Conductor support method self groovesMatrix ratio, Cu:NbT1 6:1 2.5:1Wire diameter, mm - 0.813Filament diameter, m 32.3 18Number of filaments/wire 2640 529Wire twist pitch, crrr1 0.13 1.42Number of active wires

in cable 1 72Type of transposition none RoebelCable width, cm 1.016 1.69Cable thickness, cm 0.508 0.84

the 300-kJ level for 50 cycles. This wasat the nominal design level of 10 kA and2.23-T central field on the winding. Theenergy transfer period was 2.4 ms. Thecoil was then charged to 13.4 kA and 2.99 Tfor a stored energy of 0.54 MJ. This energywas pulsed from the coil with transfer volt-ages near 40 kV. A fast transfer at reducedenergy with a 1-ms time safely reached 58 kV.The coil has been run in a subsequent testto 14.1 kA and a field of 3.1 T with 0.6-MJstored energy.

Fig. 7. Prototype SFTR-METS circuit.

TOKAMAK OHMIC-HEATING AND BURN CYCLESIMULATION

The Westinghouse-METS coil was alsoused to demonstrate the use of supercon-ducting magnetic energy storage in a sim-ulated tokamak ohmic-heating and burncycle.18 This performance, althoughmuch more demanding tha.i that requiredfor the 30-MJ SMES unit intended for theBPA system, established the feasibilityfor the extension of the METS technologyfor transmission line stabilization units.

Figure 8 gives the circuit used forbipolar operation of the coil in conjunc-tion with a commutated dc generator act-ing as a mechanical capacitor. Figure 9shows the oscilloscope trace of the ex-perimental current. The storage coil wasfirst charged to -12 kA with a continuousduty dc homopolar and then oscillatedthrough zero current to near +5 kA byconnecting it in parallel with the dcgenerator. The damped half sinusoidalenergy transfer corresponds to the energychange expected to occur in plasma heat-ing in a tokamak. After this bipolaroperation, the rectifier bank (power sup-ply) was connected across the coil tocharge it to +12 kA. This corresponds tothe burn phase of the tokamak cycle. Theentire cycle can then be repeated; how-ever, the test of Fig. 9 was concluded inthe damped oscillation caused by a dis-sipative resistance of the coil, dc-generator parallel loop.

279

/-LOSS MEASUREMENT/ APPARATUS eI »9

RECTIFIERBANK

Fig. 8. Circuit for operation of the Westinghouse-METScoil in a tokamak ohmic-heating cycle.

Fig. 9. Oscilloscope trace of current versus timeof the experimental tokamak ohmic-heating cycle.

CONCLUSIONS

Superconducting magnetic energy stor-age units should prove to be effectivecomponents of electric power systems.These devices can be used for load level-ing and peak shaving, can satisfy spinningreserve requirements, and can improve sys-tem stability. The fast time response ofthe control system will allow a fairlysmall SMES unit to damp oscillations onpower systems.

ACKNOWLEDGEMENT

The author wishes to thank W. V.Hassenzahl, W. S. Ranken, R. I. Schermer,W. D. Smith, R. D. Turner, P. Thullen, 0.D. G. Lindsay, and D. M. Weidon for theirwork contributing to this paper. The con-tent of this paper follows quite closely

that of papers submitted to the Instituteof Gas Technology and presented to theInstrument Society of America (Oct.16-18, 1978).

REFERENCES

1. IEEE Committee Report, May/June, 1976,"Survey of Pumped Storage Projects inthe United States and Canada to 1975,"IEEE-PAS, Vol. 95, No. 3, pp. 851-858.

2. Ramakumar, R., 1976, "Survey of EnergyStorage Techniques," Energy, IEEE Re-gion Six Conference Record, Tucson,Arizona, pp. 105-110.

3. Yao, N. P., Birk, J. R., Aug. 1975,"Battery Energy Storage for UtilityLoad Leveling and Electric Vehicles: AReview of Advanced Secondary Batter-ies," Record of the Tenth IECEC,Newark, Delaware, pp. 1107-1119.

280

4. Mattick, W., Haddenhorst, H. G.,Weber, 0., Stys, Z. S., 1975, "Huntorfthe World's First 290 MW Gas Turbine 15.Air Storage Peaking Plant," Proc. ofthe American Power Conference, Vol.37, pp. 322-330.

5. Los Alamos Scientific Laboratory Su-perconducting Magnetic Energy StorageProgram Progress Reports, Hassenzahl,W. V., editor, LAPR Reports Nos. 5258, 16.5415, 5472, 5588, 5786, 5935, 6004,6117, 6225, 6434, and 7132.

6. Boom, R. W., Peterson, H. A., et al.,1974 and 1976, "Wisconsin Supercon-ductive Energy Storage Project," Vol.I and II, Engineering Experiment Sta-tion, College of Engineering, Univer-sity of Wisconsin, Madison, Wisconsin. 17.

7. Ribe, R. L., Krakowski, R. A.,Thomassen, K. I., Coultas, T. A.,1974, "Engineering Design Study of aReference Theta-Pinch Reactor (RTPR)," 18.Special Supplement on "Fusion ReactorDesign Problems: to Nuclear Fusion;"International Atomic Energy Agency,Vienna.

8. Kulcinski, G. L., Project Director,1973, "UWMAK-I-A Wisconsin Toroidal 19.Fusion Reactor Design," UWFDM-68.

9. Kulcinski, G. L., Project Director,1974, "Major Design Features of theConceptual D-T-Tokamak Power Reactor,UWMAK II," IAEA-CN-33/G1-2.

10. Arendt, F., Komerek, P., Herppich, G., 20.Knobelach, A., Wermer, F., 1974,"Energetic and Economic Constraints onthe Poloidal Windings in ConceptualTokamak Fusion Reactors," Proc. 8thSymp. on Fus. Techn., Noordwijkerhout, 21.Netherlands, June 17-21, 1974,Instituut Voor Plasmafysica, Jutphass,Netherlands, EUR5182, p. 563.

11. Krakowski, R. A., Dudziak, D. J.,Oliphant, T. A., Thomassen, K. I., 22.Bosler, G. E., Ribe, F. L., Dec. 3-4,1974, "Prospects for Converting232Th to 233u in a Linear Theta-Pinch Hybrid Reactor (LTPHR)," DCTRFusion-Fission Energy Systems ReviewMeeting, ERDA-4; Germantown, MD.

12. Roberts, M., Bettis, E. S., eds.,November 1975, "Oak Ridge Tokamak Ex- 23.perimental Power Reactor Study-Reference Design," Oak Ridge NationalLaboratory Report ORNL/TM-5042.

13. Baker, C. C , Project Manager, July1975, "Experimental Power Reactor Con- 24.ceptual Design Study," General AtomicsCorp. Eng. Staff, GA-A13534.

14. Stacey, W. M., Project Manager, June1975, "Tokamak Experimental Power

Reactor Studies," Argonne NationalLaboratory Report ANL/CTR-75-2.Ballou, J. K., Brown, R. L., Easter,R. B., Lawson, C. G., Stoddart, W.C. T., Yeh, H. T., February 1977,"Oak Ridge Tokamak ExperimentalPower Reactor Study - 1976, Part 3Magnet Systems," Oak Ridge NationalLaboratory Report ORNL/TM-5574.Baker, C. C , Project Manager,December 1976, "Experimental FusionPower Reactor Conceptual DesignStudy, Volumes I, II, and III,"General Atomic Co., San Diego, CA,Report GA-A14000, July 1976; andElectric Power Research Institute,Palo Alto, CA, Report EPRI ER-289.Rogers, J. D., et al., "0.54-MJ Su-perconducting Magnetic Energy Trans-fer and Storage," Adv. Cryog. Eng.23, 48 (1978).Thullen, P., Lindsay, J. D. G.,Vogel, H. F., Weldon, D. M., "Super-conducting Ohmic-Heating Coil Simu-lation," presented at the 1978 App.Superconductivity Conf., Pittsburgh,Pennsylvania, September 25-28, 1978.Forgey, H. L., 1974, "Feasibilityand System Planning Symposium on theLudington Pumped-Storage Hydroelec-tric Generating Station," Proc. ofthe American Power Conference, Vol.36, pp. 797-806.Cresap, R. L., Mittelstadt, W. A.,March/April, 1976, "Small SignalModulation of the Pacific HVDCIntertie," IEEE-PAS Vol. 95, No. 2,pp. 536-541.Peterson, H. A., Mohan, N., Boom,R. W., July/August, 1975, "Supercon-ducting Energy Storage InductorUnits for Power Systems," IEEE-PAS,Vol. 94, No. 4, pp. 1337-1348.Boenig, H. J., Ranken, W. S., 1977,"Design and Tests of a Control Sys-tem for Thyristorized Power Suppliesfor Superconducting Coils," Proc.7th Symp. Eng. Problems of FusionResearch, Knoxville, TN, Oct. 25-28,1977, IEEE, Inc., Piscataway, NJ,IEEE No. 77CH1267-4-NPS, p. 484.Thomassen, K. I., Editor, January1976, "Conceptual Design Study of aScyllac Fusion Test Reactor," LosAlamos Scientific Laboratory ReportLA-6024; Los Alamos, New Mexico.Rogers, J. D., Baker, B. L., Weldon,D. M., 1974, "Parameter Study ofTheta-Pinch Plasma Physics ReactorExperiment," Proc. of the 5th Symp.on Engineering Problems of FusionResearch, Princeton, New Jersey,

281

Nov. 5-9, 1973; IEEE Inc., Piscataway,New Jersey, IEEE 73CH0843-3-NPS, p.432.

25. Rogers, J. D., Williamson, K. D.,April 1975, "Proposed METS-FTS CoupledSuperconducting Prototype System," LosAlamos Scientific Laboratory ReportLA-5918-P; Los Alamos, New Mexico.

26. Mole, J. D., Eckels, P. W., Haller, H.E., Janocko, M. A., Karpathy, S. A.,Litz, D. C , Mull an, E., Reichner, P.,Sanjana, Z. N., KA Superconducting0.54-MJ Pulsed Energy Storage Coil,"Adv. Cryog. Eng. 23, 57 (1978).

282

Project Title:

PROJECT SUMMARY

Superconducting Magnetic Energy Storage for PowerSystem Stability Applications

Principle Investigator: C. R. Chowaniec and P. H. Stiller

Organization: Westinghouse Electric Corporation700 Braddock AvenueEast Pittsburgh, PA 15112

Project Goals: The objective of this investigation was to seek possibleapplications for small superconducting magnetic energystorage (SMES) devices as aids to maintaining powersystem stability.

Project Status: The suitability of small SMES devices as an aid to main-taining power system stability is discussed. It wasconfirmed that small SMES units are not effective forkeeping an electrical system in synchronism after a tran-sient disturbance because of the limited power and storagerating; however, they do increase the dynamic stabilitylimit of interconnected systems and may be used for sub-synchronous resonance damping, provided power conditioningequipment different from a line-commutated converter isemployed.

A questionnaire was prepared and sent to about 25utilities to obtain information about possible SMESunit application sites in utility systems. Theremaining work within the project is the evaluationof the questionnaires.

Contract Number: LP8-9415Ca

Contract Period: July 1978 - Mar. 1979

Funding Level: $24,060

Funding Source: Los Alamos Scientific Laboratory

283

SUPERCONDUCTING MAGNETIC ENERGYSTORAGE FOR POWER SYSTEM

STABILITY APPLICATIONS

C. R. Chowaniec and P. H. StillerWestinghouse Electric Corporation700 Braddock AvenueEast Pittsburgh, PA 15112

ABSTRACT

Superconducting magnetic energy storage (SMES) devices are characterized in termsof reasonable power and total energy ratings as well as other pertinent specifications.The objective of this investigation was to seek possible applications for small SMESdevices in power systems, particularly as stability aids. Small SMES devices are thosewhich are not large enough to be practical as bulk energy storage units. The suit-ability of the SMES device as an aid to maintaining powar system stability is discussedin light of the anticipated capabilities of the device. Possible application of theSMES device as a guard against subsynchronous resonance is also considered.

INTRODUCTION

Huge superconducting magnets havebeen proposed in recent years as anenergy storage device for diurnal loadlevelling on electric power systems. Themuch smaller 30-megajoule magnet; now inthe conceptual design phase at Los AlamosLaboratories, represents a logical stepin the development of superconductingtechnology for power system applicat ions.The 30-megajoule superconducting magnetenergy storage (SMES) system is intendedfor use in improving power system sta-bility. Its characteristics are sum-marized in Table 1.

Prior to the past ten years, sta-bility concerns for power systems wereprimarily related to maintaining thegeneration of a system in synchronism.Acceptable stability performance wasgenerally achieved without much diffi-culty. As power systems expanded, theextensive use of interconnection betweensystems together with an increasingdependence on firm power flow over theselines has renewed the concern of powersystem stability. The consequences ofinstability were most dramatically ex-hibited by the 1965 Northeast power fail-ure.

The generators on a power system,which are connected by the transmission

network, must operate in synchronism.Loss of synchronism occurs when theangular difference between the rotors oftwo generators or groups of generatorsexceeds a certain value (nominally 90°).Although the generating unit controls doinfluence the stability of the system, itis the transmission network which producesthe angular differences between machines.For this reason, the main emphasis in thestudy of stability has historically beenon the transmission network.

System instability can generally beclassified in one of three categories -steady-state, transient, and dynamic. Ofthese three categories, stead/-state in-stability is the least probably to beexperienced. Thus, only transient anddynamic instability were considered inattempting to find applications for the30-MJ SMES unit. Another phenomenon whichcan occur on certain systems is undampedsubharmonic osci1lation. Although it isnot a stability problem in the strictestsense, it can damage equipment nonetheless.Therefore, the potential application ofSMES units to eliminate subsynchronousresonances was also considered.

TABLE 1. CHARACTERISTICS OF A TYPICALSMALL SMES UNIT

Maximum Power 10 MWOperating Frequency 0.35 HzEnergy Interchange 9.1 MJ (2.53 KWH)

(based on 1/2 cycle)Maximum Stored Energy 30 MJ (8.4 KWH)Life 107 CyclesMaximum \l^cMaximum lc|cInductanceDiameter of Coi1Height of Coil

LossesConductorStructure

2.18 kV5 KA2.k H2.7 M0.86 M

58.7 W50. W

Cost of Prototype $3-5 Million

SUBSYNCHRONOUS RESONANCE

When the synchronous machines of apower system are coupled together by verylong transmission lines, series capaci-tors are used to compensate for the lineinductance and reduce the net impedance.Although this increases the amount ofpower which can be transmitted over thelines, it also results in a natural fre-quency below the fundamental 60-Hertzpower frequency.

These oscillations often damp outwhen excited., however it is sometimespossible for the oscillations to be nega-tively damped. Negative damping arisesprimarily from the induction generatoreffect of the generator at subsynchronousfrequencies. Electromechanical inter-actions will cause pulsating torques whichmay excite torsional natural frequencies.The result can be severe damage to theturbine generator.

A simple model of subsynchronousresonance is depicted in Figure 1. Thesubharmonic natural frequency is given bythe equation:

fssr - fc

where f is the power frequency, x isthe reactance of the series capacitor, andx) is the total series reactance of thesystem. This determines the slip of thegenerator, s, which is negative since f s s r<fo. Thus we see that RR/S in Figure 1is negative. If this negative resistanceis not offset by other resistances in thenetwork, spontaneous oscillations mayoccur and continue to increase in magni-tude. Frequencies typically range from20 to 55 Hz.

SERIES COMPENSATEDTRANSMISSION LINE

GENERATOR INFINITESOURCE

FiGur:i.

SIMPLE MODEL TO ANALYZE SUBSYNCHD. 'U~» RESONANCE

285

Subsynchronous resonance has occured inthe West on several occasions. Two rotorshaft failures at the Mohave Plant in1970 and 1971 were apparently caused bysubsynchronous resonance.

The following approaches have beentaken in order to alleviate subsynch-ronous resonance problems:

1) Blocking filters2) Temporary shunting of series

capaci tors3) Reduce amortisseur resistance

Blocking filters include static,high-Q. filters with supplemental excitercontrol for additional torsional dampingand dynamic filters, applied either Inseries or shunt, which are modulated toresonate with the electrical circuit but180° out of phase with the subsynchronousresonance.

By shorting out the series capacitorsas oscillations begin to grow, the reso-nating circuit is eliminated and theoscillation will quickly diminish. Theseries capacitors can then be reinserted.The obvious problem is that the seriescompensation may not be available whenthe additional power transfer it affordsis required.

Reduced amortisseur winding resis-tance reduces the tendency for self-excited osci 1 lat ion by reducing the magni-tude of the negative rotor resistance, RR.This will, however, reduce the dampingunder large siips.

Returning briefly to the dynamicfilters, a shunt inductive reactance,appropriately modulated, will create aparallel resonant circuit which canactively cancel subsynchronous resonantconditions. The device is essentiaiiy astatic VAR generator with additional con-trols and without capacitors. The induc-tance element and power switching equip-ment is sure to be less expensive thanthe superconducting magnet and its con-verter. A point to be made here is thata simple 1ine-commutated inverter isprobably not usable in a SMES system forthis application. Furthermore, the fre-quencies involved are much higher thanthose considered in the conceptual designof the 30-MJ unit. In summary, it appearsthat subsynchronous resonant systems donot present a very viable application ofsuperconducting magnet energy storage

systems of the type described in Table 1.

TRANSIENT STABILITY

Transient instability represents thegreatest concern, as it can occur on anysystem as the result of a major disturb-ance such as a transmission line fault orthe tripping of a loaded generator. Tran-sient stability analysis is primarilyconcerned with the effect of transmissionline faults on generator synchronization.If the remaining transmission is adequateand the speed and angle variations at thetime the fault is cleared are limited, thesystem is likely to settle into a newsteady-state equilibrium. Loss of synch-ronism as a result of transient disturb-ances will generally occur in the firstfew seconds following the initial disturb-ance, however system alternations result-ing from the transient may precipitate sub-sequent steady-state or dynamic insta-bility which would also result in systemcol lapse.

The time period involved in mosttransient stability analyses is relativelyshort (generally about 1 second or less).It is primarily concerned with the re-sponse of a single unit or plant to acontingency disruption. Because of theshort time period involved, governor re-sponse is usually not important althoughexciter response (especially with highresponse excitation) may play a signifi-cant role in maintaining synchronism withthe system.

Figure 2 illustrates somewhat of aworst-case transient stability situation;that is, a generating unit connected tothe system via a single, long transmissionline. Before a fault occurs, the system

.PIN

INFINITE BUS

-BREAKERS OPEN ANDRECLOSE AFTER 4t SECONDS

FIGURE 2.

WORST-CASE TRANSIENT STABILITY SITUATION

286

is in equilibrium and the input power, Pinis transmitted over the line with a re-sulting phase angle of 6 between the"gen-erator and infinite bus.° If the line issuddenly opened, the power generated canno longer be transmitted and is used toaccelerate the rotor instead. If the lineis reclosed in time, the unit will de-celerate and gain a stable operatingpoint after a few well-damped swings. Iftoo much time elapses, synchronism is lost.The determination of ultimate system sta-bility after major disturbances can bedetermined using the "equal-area crite-rion" depicted in Figure 3.

LINE POWER vs. PHASE ANGLE

. LINE CLOSED

30 60 90 120 150 180

ANGLE - DEGREES

FIGURE 3.

EQUAL - AREA" CRITERION

Several methods of ensuring transientstability have been applied. They include:

1) Fast Response Excitation2) Fast tfalving3) High-Speed Reclosing Breakersk) Independent Pole Operation of

Breakers5) Braking Resistors6) Series capacitor Insertion

By driving the exciter to ceiling, itis possible to maximize the amount of powertransmitted over remaining transmissionlines. One problem with the use of high-respo-ise excitation is its tendency toaggravate dynamic instability problems.Fast vaiving serves to reroute the flow ofmuch of the steam flow around the turbine,thereby eliminating accelerating power,but resets slowly. Both high-speed break-ers and single-pole tripping minimize thedisturbance, the first by re-establishingcircuits as quickly as possible while theother maintains power transfer (unbalanced)over the remaining phases. Braking resis-tors inserted during a fault create anartificial local load, thereby limitingthe angular accelerator and increase inrotor angle. Series capacitors insertedinto remaining circuits when a circuit istaken out of service by a fault increasethe power transfer capability in thoselines, again limiting the acceleration ofthe rotor.

The SMES unit, when used as a tran-sient stability aid, would operate in away similar to the braking resistor. Ex-cess energy resulting from the sudden lossof transmission capacity would be absorbedby the magnet, thereby limiting the angu-lar excursion of the rotor from its steady-state operating point. The SMES unitwould have the added advantage of provid-ing active damping of subsequent swingsafter the fault is cleared. In order tobe effective as a transient stability aid,however, the SMES unit must first meetcertain energy storage and power ratings.Both of these ratings will typically bewell over an order of magnitude greaterthan those provided in the conceptualdesign of the 30-MJ SMES unit, with in-herent cost penalties. In view of theserequirements (particularly power rating)and compared to the alternatives, theapplication of SMES units for transientstability appears impractical.

DYNAMIC STABILITY

Some interconnected systems exhibit atendency toward dynamic instability inwhich negatively damped power oscillationscan lead to the ultimate breakup of thepower system. Although there has been noevidence of dynamic instability in theeastern connections to date, the problemhas been shown to exist in portions of theWest, Midwest, and Canada.

2B7

The time intervals involved indynamic stability calculations are muchlarger than for transient stability.Oscillat.on frequencies ranging from afraction of a Hertz to a few cycles perminute are common. In studying the dy-namic stability of a system, the concernoften relates to oscillation modes involv-ing large groups of machines. The inter-action of automatic voltage regulator,exciter, and speed governor controls is anessential part of such analyses.

The most commonly employed measurefor combatting dynamic instability is thepower system stabilizer. This device isquite inexpensive and has been proveneffective in eliminating or reducing theoccurrence of spontaneous dynamic oscil-lations. It also improves the damping ofoscillations subsequent to transient dis-turbances. The power system stabilizerprovides a supplemental leading frequencyfeedback signal to the exciter throughthe automatic voltage regulator whichcompensates for lags inherent in otherportions of the control system. Althoughthe power system stabilizer has been suf-ficient in overcoming problems of dynamicstability encountered in the past, it isnot as effective at very low frequenciesof oscillation or for high system angleoperation (heavy tieline loading).

Another approach is to simply reducethe gain of the automatic voltage regula-tor, however this tends to sacrifice theperformance of the particular unit undertransient conditions.

The advantages of dc transmissionlines in improving ac system stabilityhave been recognized and these capabili-ties have been incorporated into the con-trols of existing dc lines. There, theyhave demonstrated the ability of the dcline to damp out oscillations by properlymodulating the power flow over the dc lineand to improve transient stability byrapidly changing the dc line loading. Todate, however, the installation of dclines has been justified on the basis ofconsiderations other than improved sta-bility characteristics. But the factorswhich contribute to the economic feasi-bility of dc lines (particularly long dis-tances) are often the same factors whichcause a dynamic stability problem. The"fringe benefit" of improved stabilitymight arise naturally with the expansionof systems in which SMES units might beemployed for system stability applications.

Utilities which contend with this pro-blem are characterized by load centerswhich are widely separated and are suppliedby remote generation with longer angulardisplacements relative to local generation,long transmission lines connecting theloads and generating stations are a commonfeature. The large angular differencesin the generators which are required totransmit power over these long lines re-sult in reduced line capacity and low nat-ural frequencies which are poorly (or pos-sibly negatively) damped. Slight speedvariations which occur continuously undernormal operation excite oscillations inrotor angle and generator load. If thesystem damping of these oscillations isnegative, the oscillation will increase inmagnitude until it is limited by non-linearities inherenet in equipment or con-tinue to the point that synchronism willbe lost if no remedial action is taken.

The 30 MJ SMES unit described in Table1 is appropriately sized for dynamic oscil-lation dampirg applications and would beoperated in a manner very similar to dctransmission where power system stabilizersprovide insufficient damping in lieu of acor dc transmission line additions. Thereare some problems, however. First, thenumber of applications where a 30 MJ unitcould be justified to increase dynamicstability limits will be quite limited (atleast at the present time). Second, theamount of energy passed through the deviceover a given time period will be relative-ly small. When this is considered alongwith the relatively small inductance (2.kHenries), one must seriously question theneed for a superconducting coil for thisapplication.

CONCLUSIONS

Superconducting magnet energy storageis not likely as an economic solution topower system stability problems except invery special cases. The 30 MJ SMES unitsconsidered, when applied for power systemstability, will give important operatingexperience for superconducting bulk energystorage.

The primary applications of supercon-ducting magnet energy storage will proba-bly be in the area of energy storage;therefore, applications should be soughtfor early demonstration of the technologyin this area. When SMES units are usedfor the purpose of bulk energy storagethe stability improvements (transient and

288

dynamic) discussed in this paper will bean additional asset realized by the SMESunit.

ACKNOWLEDGEMENTS

The authors wish to acknowledge thesupport of the Los Alamos ScientificLaboratory which provided funding for thisproject.

REFERENCES

1. H. A. Peterson, N. Mohan, andR. W. Boom, "Superconductive EnergyStorage Inductor-Converter Units forPower Systems," IEEE Trans. PowerApparatus and Systems, Vol. PAS-9A,pp. 1337-13^8, July/August 1975-

2. R. L. Cresap, D. N. Scott7 W. A.Mittelstadt, and C. W. Taylor, "Opera-ting Experience With Modulation of thePacific HVDC Intertie," IEEE Trans.Power Apparatus and Systems, Vol. PAS-97, pp. 1053-1059, July/August 1978.

3. W. A. Mittelstadt, "Four Methods ofPower System Damping," IEEE Trans.Power Apparatus and Systems, Vol. PAS-87, pp. 1323-1329, May 1968.

4. E. W. Kimbark, "Improvement of PowerSystem Stability by Changes in theNetwork," IEEE Trans. Power Apparatusand Systems, Vol. PAS-88, pp. 773-781,May 1969.

5. L. A. Kilgore, L. C. Elliott, andE. R. Taylor, "The Prediction and Con-trol of Self-Excited Oscillations Dueto Series Capacitors in Power Systems,"IEEE Trans. Power Apparatus and Sys-tems, Vol. PAS-90, pp. 1305-1313, May/June 1971

6. E. W. Kimbark, "How to Improve SystemStability Without Risking Subsynchro-nous Resonance," IEEE Trans. PowerApparatus and Systems, Vol. PAS-9~6~,pp. 1608-1619, September 1977.

7. L. A. Kilgore, D. G. Ramey, andW. H. South, "Dynamic Filter and'Ot^rSolutions to the Subsynchronous Reso-nance Problem," Proceedings of theAmer. Power Conf., Vol. 37, PP929, (1975):

fi. R. H. Millan, J. A. Mendoza,C. Cardoza, and A. de Lima, "DynamicStability and Power System Stabiliz-ers," IEEE Trans. Power Apparatusand Systems, Vol. PAS-96, pp. 855-862, May/June 1977.

9. P. Bingen, G. L. Landgren, F. W. Keay,and C. Raczkowski, "Dynamic StabilityTests on a 733 MVA Generator atKincaid Station," IEEE Trans. PowerApparatus and Systems, Vol. PAS-93,pp. 1328-1331*, Sept./Oct. 197**.

10. H. M. Ellis, J. E. Hardy, andA. L. Blythe, "Dynamic Stability ofthe Peace River Transmission System,"IEEE Trans. Power Apparatus and Sys-tems, Vol. PAS-85, pp. 586-600,June 1966.

11. W. H. Croft and R. H. Hartley, "Im-proving Transient Stability by Use ofDynamic Braking," IEEE Trans. PowerApparatus and Systems. Vol. 81, pp.17-26, April 1962.

12. M. L. Shelton, P. F. V.'inkelman,W. A. Mittelstad, and W. J. Bellerby,"Bonnevllle Power AdministrationI'lOO MW Braking Resistor,1' IEEE Trans.Power Apparatus and Systems, Vol. PAS-9t, pp. 602-611, March/April 1975.

13. R. T. Byerly and E. W. Kimbark,Stability of Large Electric PowerSystems, IEEE Press, New York, N.Y.,

923-

289

SESSION VI: UNDERGROUND PUMPED HYDROELECTRIC STORAGE

291

PROJECT SUMMARY

Project Title: Underground Pumped Hydro Storage Program

Principle Investigator: Dr. G. T. Kartsounes

Organization: Argonne National Laboratory9700 South Cass AvenueT-12-3Argonne, IL 60439

Project Goals: The primary objective of the Underground Pumped HydroStorage Program is to enable the electric utility industryto supply power at the lowest possible cost, while elimi-nating the use of premium fossil, fuels for peak-power generation.Successful commercialization of this energy storage schemerequires research and development studies on turbomachinery,lower reservoir contruction, geology, motor/generators, andsystem arrangement and optimization.

Project Status: This program was initiated during FY 1978. A contract wasnegotiated with Ailis-Chalmers Corporation, Hydro-TurbineDivision, for preliminary design studies on single-anddouble stage reversible pump/turbines with wicket gates. Thesestudies were completed and a report is in preparation.

Contract Number: W-31-109-ENG-38

Contract Period: FY 1978

Funding Level: $200,000 B0

Funding Source: Department of Energy, Office of Conservation, Division ofEnergy Storage Systems

293

UNDERGROUND PUMPED HYDRO STORAGE - AN OVERVIEW

S.W. Tain, C.A, Blomquist, G.T. KartsounesEnergy and Environmental Systems Division

Argonne National Laboratory9700 South Cass Avenue

Argonne, Illinois 60439

ABSTRACT

This paper reviews the status of Underground Pumped Hydro Storage (UPHS) for elec-tric utility peaking and energy storage applications. The salient features of major re-cent studies are reviewed. Turbomachinery options and advances in high-head pump/turbinesare discussed. The effect of head, capacity, turbomachinery unit size and type, and otherperformance variables on the cost of a UPHS plant are presented. Market potential, sitingcriteria, lower reservoir construction, and geological related issues are addressed. Theenvironmental impact of a UPHS plant is reduced from comparable facilities, and theseissues and other safety concerns are presented. UPHS is an economically viable schemefor energy storage and peaking applications, where considerable savings in premium fuelsare achievable by the replacement of combustion gas turbines. The technology for UPHS isavailable, but additional research and development is required for high-head turbomach-inery, motor/generators, cavern geology, and system optimization.

INTRODUCTION

THE NEED FOR ENERGY STORAGE

Electric utilities face large varia-tions in the daily electric power demand.During periods of relatively staady demand,power is generally provided by nuclear orcoal-fdred base-load plants, which are eco-nomical to operate. For peak load periods,gas turbine driven generators are used.They are fast responding, easy to maintainand operate, but they have relatively lowefficiencies and burn premium fuel (oil ornatural gas). These undesirable character-istics provide the incentive to seek alter-nate energy technologies for peaking appli-cation.

During periods of low demand (usuallyat night), bar.,e-load plants operate at re-duced capacity. If they are operated atfull-capacity during these periods, theextra energy generated could be stored pro-vided viable energy storage schemes areavailable. The stored energy would thenbe available to meet (at least partially)the peak demands, and thus, reduce the re-liance on premium-fuel fired units.

Solar energy (direct and indirect)requires storage because its availabilityusually cannot be assured. Depending onthe scale of power generation (residential,

community, or utility-size), there is awide range of viable energy storage tech-nologies available.1* This storage is some-what different from the peaking applicationdiscussed previously.

UNDERGROUND PUMPED HYDRO STORAGE

One energy storage scheme that haslong been utilized for peak power genera-tion by the electric utilities is surfacepumped hydro storage (SPHS). In thisscheme (Fig. 1) idle base plant capacityis utilized to pump water from a lowerreservoir to an upper reservoir during off-peak hours. During the peak demand period,power is fed into the power grid by lettingthe water run down to the lower reservoir.In this way, the stored potential energyof the water is converted into kineticenergy which is utilized to run a hydro-turbine coupled to a generator.

SPHS has been in the U.S. since 1953when the U.S. Bureau of Reclamation in-stalled a 9 MW unit at Flatiron.9 Todaymany systems are operational. However, aSPHS system requires a suitable site forthe upper-level reservoir located at suf-ficient elevation in the vicinity of the

294

lower reservoir and also at a reasonabledistance from the load center concerned.Sites with such characteristics are beingrapidly depleted today within the U.S.The concept of underground pumped hydrostorage (UPHS) is designed to minimizethese siting and transmission difficultieswhile still fulfilling the primary missionof generating peak power for the electricutility grid.

CONVENTIONAL PUMPED HYDRO

- INTAKESURGE TANK

through the penstock shaft, and the lowercavern is vented to the atmosphere throughthe vent shaft.

UPPER RESERVOIR

GROUND LEVEL

ACCESS, VENTILATION -AND CABLE SHAFT

VENTILATION''SHAFT

FOWERSTATION-

UNOERGROUND PUMPED HYDRO

MNTAKE

^ PENSTOCK SHAFT

LOWER RESERVOIR

Fig. 2. UPHS Single-Drop Scheme

Fig. 1. Pumped Storage Schemes

Conceptually, UPHS (Fig. 1) is verysimilar to SPHS, and, similarly, does notrequire premium fuel. The major differencefrom SPHS is that the upper reservoir forUPHS is at ground-level while the lowerreservoir (an excavated cavern in general)is located underground. This importantfeature results in the advantages of sitingflexibility and reduced transmission costsand retains the good system reliability andavailability characteristics of SPHS.

A typical proposed UPHS plant willhave a capacity of 1000-2000 MW with 8-10hours of storage, and an overall efficiencyclose to 80%. Presently, there are twobasic plant configurations — the one-dropand tha two-drop schemes (Figs. 2 and 3).

In the one-drop scheme, there is onlyone underground reservoir. The turbomach-inery is housed in the powerhouse locatedbelow the underground reservoir to providepump submergence. Access, cable and equip-ment shafts connect the powerhouse to theaboveground service building. The power-house is linked to the upper reservoir

PENSTOCK -SHAFT

POWER STATION ~

SERVICE BUILDING

GROUND LEVEL

-ACCESS, VENTILATIONAND CABLE SHAFT

-VENTILATION SHAFT

-INTERMEDIATERESERVOIR

Fig. 3. UPHS Two-Drop Scheme

The two-drop configuration utilizessingle-stage, reversible pump/turbines(RPT) with an average gross head of abouthalf the depth at which the lowest reser-

295

voir is situated. A smaller intermediatereservoir is located half way betweenground level and the underground cavern.A powerhouse is connected to each of thetwo underground reservoirs. The interme-diate cavern allows the two power plantsto operate in series without synchronizingthe turbomachinery. The shaft requirementsare similar to the one-diop scheme.

UNDERGROUND RESERVOIR

For both configurations, the lowerreservoir is usually constructed in a room-and-pillar layout with an intersecting gridof tunnels (Figs. 2 and 3). In the Chas.T. Main report,2 which is the latest studyon the subject, a typical UPHS lower reser-voir will cover an area of about 64 x 101*sq. m. ("VL/4 sq. mi.). The size of eachtunnel is about 15-m wide by 25-m high witha crown of 7.5 m radius. The pillars sep-arating the tunnels are 45 m by 45 m. Suchroom-and-pillar construction is expected togive a stable configuration at depths(>1000 m) considered in UPHS schemes.

The three primary methods of shaftconstruction are conventional sinkingmethods, machine boring, and raising bymeans of an elevating platform, followed,if necessary, by downward enlarging. Thetime required for construction of a UPHSplant is 6-8 years. Total time in excessof 10 years may result if planning andlead-time are considered.

UPPER RESERVOIR

The upper reservoir for a UPHS plantcan be a specially-constructed dedicatedreservoir, or natural lake, or one createdby damming a river or stream. Because ofhigh-head plant operation, the requiredupper reservoir volume will be significant-ly less than a conventional pumped storagereservoir. Containment structures for thereservoir will be essentially the same asthose used for conventional pumped storage.There does not appear to be any identifiedproblems with reservoir construction otherthan possible environmental aspects, dis-cussed later. A unique feature of the UPKSconcept is the availability of a large vol-ume of high-quality rock from the under-ground reservoir excavation that can beused for containment construction.

MAJOR STUDIES

Within the last few years, Chas T.Main, Inc.,2 PSEG,wAcres American, InC.,1'13

and Harza Engineering Co.6 have conductedstudies of UPHS. Salient features of thesestudies are summarized in Table 1, andseveral points are worth noting.

(1) There is a considerable spread in therange of head under consideration(from 765 m to 1500 m ) . Generally,the more recent studies tend to con-sider higher head.

(2) The estimated plant cost varies from273 to 405 $/kW. These costs havebeen updated to September 1978 dollarsby using a general inflation rate of8% per annum. In a more precise esti-mate, the inflation rate will be dif-ferent for the various components ofthe plant. The more recent studiestend to give higher cost estimates.The higher costs are due to differentsystem configurations and inflation-related costs.

(3) The turbomachinery considered iseither conventional or slight exten-sions of present technology. Single-stage RPT units for heads up to about765 in are considered. For higherheads, either a two drop scheme isused or multistage RPT units or tan-dem units are employed. The state-of-the-art of the multistage and tandemunits is adequate for UPHS application,but they are inherently more expensive.Economical operation requires some im-portant extensions to turbomachinerytechnology.

GEOLOGY-RELATED ISSUES

The important factors governing UPHSsiting opportunities are site geology, hy-drology and the location of major load cen-ters.' 2' 1 1 In general, the undergroundconstruction, particularly for the cavern,requires a medium with good competent rockwith few discontinuities. Igneous intru-sive and crystalline metamorphic rockstructure will in general satisfy UPHS con-struction requirements. Favorable areaswithin the continental U.S. for UPHS de-ployment are shown in Fig. 4. Of course,within the favorable areas, there may existlocalized regions with unfavorable geologyand vice versa. Since transmission cost isabout $0.34/kW/mi,2 incentives exist tokeep plants relatively close to load cen-ters. It is worthwhile to note that withinthe favorable geological areas such as theMidwest and the Northeast lie some of the

296

Table 1. Major UPHS Studies (adapted from Ref. 2)

TITLE PREFERRED SCHEMEPREFERRED TYPEOF EQUIPMENT

ESTIMATEDCOST <$/kW)

PRINCIPAL PURPOSE

Underground Pumped StorageResearch Priorities (April1976) Prepared by AcresAmerican Incorporated

An Assessment of EnergyStorage Systems Suitablefor Use by Electric Utili-ties (July 1976) Preparedby Public Service Electricand Gas Company, Newark,New Jersey

Siting Opportunities inthe U.S. for CompressedAir and UndergroundPumped Hydro EnergyStorage Facilities(December 1976) Pre-red by Acres American

Incorporated

Underground PumpedHydro Storage and Com-pressed Air EnergyStorage. An Analysisof Regional Marketsand Development Poten-tial (March 1977)Prepared by HarzaEngineering Company

Underground Hydro-electric PumpedStorage. An Evalua-tion of the Concept(June 1978) Prepared>y Chas. T. Main Inc.

Indirectly suggested oneor two drops, approxi-mately 900 m (3000 ft)head

Considered:- Up to 765 m (2500 ft)

for single-stage re-versible units.

- About 1070 m (3500 ft)for two plants in series

In economic computationsassumed 200 MW plant,10-hr storage1070 m (3500 ft) head

Two-drop schemeTotal head 1340 m (4400ft) 2000 MW, 6 to 10-hrstorage

200 MW units: at 700 m(2300 ft) single-stagerunner reversible andat 1000 m (3300 ft)multi-stage runner,reversible

Single-stage rever-sible units of about300 MW

250 MW single-stagereversible units at670 m (2200 ft) head

One-drop schemeTwo-drop schemeTotal head -1200 i10-hr storage

333 MW multi-stageOne-drop: a. Multi-stage RPT; b. TandemunitTwo-drop: two single-stage RPT in series

300 To identify particular as-for 1000 MW pects requiring detailedat 900 m examination during a sub-(3000 ft) sequent comprehensive pre-and 10-yr llninary design phasestorage

<544 To provide the required datafor 10-hr to establish research andstorage development priorities for

energy storage technology

Approxi-mately 273

200 to 366(2000 MW)10-hr stor-age

348 to 405

To categorize the geologyof the continental UnitedStates in accordance withthe potential for the sitingof compressed air energystorage (CAES) and UPHS

To Identify and describeregional markets for UPHSaod CAES and perform geo-logic analysis to determineregional development poten-tial

To Investigate state-of-the-art of UPHS, to evaluatefeasibility and relativeeconomic viability and toIdentify needs for furthertechnological and economicresearch

major load centers within the continentalU.S. These areas are, therefore, primeregions for UPHS applications.

One important issue is the stabilityof the underground structure. This per-tains to the response and long-term beha-vior of the underground cavern when i t issubjected to the cycling effect of chargingand discharging of water over a long periodof time. During normal operation of a UPHSplant, the underground structures are ex-posed alternately to water and air. Theeffect of changing pressure, temperatureand erosion on the stability of the lowercavern is an issue that has to be investi-gated. Necessary inputs for such an anal-ysis will be:

(a) the thermo/physical/mechanical rockproperties of the region in question,

(b) the state of the rock mass such as

(c)

spacing of joints and their orienta-tion and in situ stress state (verti-cal as well as horizontal stresses),

specific details of the water/pumping/discharge cycle, and, finally,

(d) specific configuration of the under-ground reservoir.

With these as inputs, the elastic/plasticresponse of the underground structure maybe analyzed. For site-specific stabilityanalysis, of course, rock properties per-taining to that particular? site will beneeded.

Another important issue is the explor-atory technique used. Since rock mass in-formation is a prerequisite for any stabi-lity analysis, the means of collecting suchinformation is an important issue. However,up to now the principal exploratory tech-

297

AREAS GENERALLYFAVORABLE FORUPHS TAVERNS.

Fig. 4. U.S. Sites Generally Favorable for UPHS Caverns

nique still involves direct drilling ofboreholes. Because rock mass propertiescan vary substantially over the depth rangeinvolved in UPHS project (>1000 m), bore-holes have to be drilled to comparabledepth in order that reliable information isobtained. For example, the in situ stressratio (horizontal to vertical stress) mayvary significantly over the depth rangeconsidered in UPHS schemes. In addition,all methods are accurate only over a shortrange from the borehole. However, thiskind of approach, while expensive, seems tohave no substitute at the moment. Hence, aUPHS project carries an element of riskthat is characteristic of projects that re-quire geo-exploration to provide input in-formation. This risk can be reduced bymore extensive on-site geo-investigationwithin reasonable economic limits.

Other issues such as questions of hy-drological and environmental concerns canbe dealt with through judicious site selec-tion.

ENVIRONMENTAL, SOCIO-ECONOMIC,AND SAFETY ASPECTS OF UPHS

The selection of a UPHS plant willgenerally be determined by the availabilityof competent geologic conditions, the prox-imity of load centers and transmissionlines, and a sufficient water supply.Closely tied to these variables are envi-ronmental considerations. Whereas the en-vironmental effects of a UPHS plant will

vary with the individual site, all loca-tions have certain generic factors such asdisposal of excavated material, water con-tamination, and disruption of the naturalhabitat. Consideration must be given tothe socio-economic impact of the area duringplant construction.

It has been estimated that from 5.5to 7.5 million cubic meters of materialwill be excavated for construction of thelower reservoir and powerhouse. Part ofthis material can be utilized for construc-tion of the upper reservoir containment;the remainder requires either on- or off-site disposal. This is excellent qualityconstruction material and its usage couldbe profitable. Off-site utilization ordisposal can produce air pollution or traf-fic congestion. Disposal by dumping canalter surface drainage, create a visualblight and limit land usage.

The contamination of reservoir surfaceand subsurface water is a factor that mustbe considered. Construction of the lowerreservoir will alter ground water charac-teristics and local aquifers could be pene-trated, diverted or drained. Groundwatercontamination may result from seepage aroundaccess shafts or penstock leaks. Surfacewater drainage and quality will be affectedby roads, buildings and the upper reservoir.Cycling of water between the reservoirscould result in mineralization, solids sus-pension, and an increase in water tempera-ture. Mineralization should be minimal due

298

to the high quality of rock required forcavern construction. Solids suspensionshould approach the level of normal reser-voirs after initial operation and sedimen-tation have occurred.

Water heating from the hotter under-ground rock does not appear to be signifi-cant, but additional investigations arerequired. If the surface reservoir is anatural lake, then water level fluctuations,mineralization, temperature, groundwaterseepage, oxygen depletion, and algae for-mation effects on the aquatic life in thelake must be evaluated. Water dischargefrom the lake will in turn impact the re-gion downstream of the lake. A manmadereservoir will reduce these effects butwill introduce concerns during the diver-sion of water for the filling and replen-ishment of the reservoir.

It is assumed that site selection willexclude areas of historic or archaeologicsignificance, rare or endangered plants orwildlife, planned parks, and valuable min-eral deposits. The disruption of thenatural habitat requires consideration offoliage elimination and displacement ofwildlife. The addition or alteration ofterrain features could produce a visualimpact.

Socio-economic impacts can be bothbeneficial and detrimental to the surroun-ding region. Some beneficial effects in-clude increases in the tax base, employmentand business activities. Detrimental ef-fects include traffic congestion and in-creased demands on service such as police,schools, and medical. These socio-econo-mic factors are not unique to a UPHS plantbut are generally similar to any large con-struction project.

Safety issues are similar for anyplant construction, but a UPHS plant hasthe additional problems associated withunderground mining and operation. Floodingof the powerhouse is a concern because ofthe serious equipment damage and danger tooperating personnel that could result. Thepossibility of having the underground sys-tem subjected to the full hydrostatic headcould also be a problem.

• Careful site selection, planning, ap-propriate design, and the use of a dedica-ted man-made upper reservoir will eliminateor mitigate the environmental impact of aUPHS plant. Alternatives to UPHS such asconventional pumped storage, combustion

turbines, and coal-fired cycling plants,are not as attractive environmentally asUPHS. Conventional pumped storage requiresadditional land area for the lower reser-voir and only limited sites are existent.Combustion turbines and coal-fired plantsintroduce air pollutants and require stor-age and transportation of fossil fuels.Common to all storage schemes is the addi-tional pollution resulting from increasedoperation of the base-load plant.

TURBOMACHINERY FOR UPHS PLANTS

As discussed previously, the twoschemes projected for UPHS are the single-drop and the two-drop systems. A schematicrepresentation of the turbomachinery andpresent head limitations that will be usedfor these systems are shown in Fig. 5.Basically, the equipment is either a rever-sible pump/turbine (RPT) or a tandem unitconsisting of an impulse turbine, usuallya Pelton-type waterwheel, and a multistagepump. Single-stage and multistage unitsare available and two-stage turbine designsare being considered.

HEADLIMIT

SEPARATE-UNITS

1400 m IMPULSE FORTURBINE MODE

MULTI-STAGENON-GATEDPUMP FORPUMP MODE

Fig. 5. Turbomachinery for UPHS Plants

This single-stage RPT has been u t i l -ized extensively for pumped storage pro-jects since about 1954. These units pro-vide the least complicated arrangement andare generally equipped with adjustable

TYPE

SINGLE-STAGEREVERSIBLEGATED

TWO-STAGEREVERSIBLEGATED

MULTI-STAGEREVERSIBLENON-GATED

299

wicket gates which provide regulation ofpower output and good control for startingand synchronizing the unit in turbineoperation. In pumping operation, adjus-table wicket gates are relatively ineffec-tive for regulating discharge but enablevariable-head operation at optimum effi-ciency. One disadvantage of the single-stage turbine is its requirement of deepersubmergence to avoid cavitation; anotherlimitation is the head range. The maximumdesign head for these machines has beansteadily increasing. In the early sixties,operating heads were less than 250 m. Inthe next ten-year period, the maximum headreached 390 m (Robiei-Switzerland). Sincethat time, machinery has been ordered andcommercial operation started for a numberof plants with turbine heads in the 500-mrange. The current maximum head is 620 tnfor the Bajina Basta Power Company inYugoslavia. Therefore, utilization ofconventional single-stage RPT units for a1000-1200-m UPHS plant necessitates a two-drop system.

Continual progress toward higher-head,single-stage RPT units is being made, butSwiecicki estimates that it may take morethan another decade before, if ever, amachine is ordered and designed for a 900-1000-m head. A conservative approach tohead increases is attributed to the reluc-tance of manufacturers to advance thestate-of-the-art too rapidly and becauseutilities are not willing to accept mach-ines which deviate significantly from es-tablished practice. In addition, therehas not been a market for high-head tur-bines that would encourage manufacturersto develop these machines. In the UnitedStates, there are very few conventionalpumped storage sites which could provideheads in excess of 550 m. In Europe, high-head pumped storage plants have used eithermultistage RPT or tandem units. The prac-tical upper limit on the design-head for asingle-stage RPT is estimated as 800-1200 m.Being able to achieve the higher limit ismandatory for the single-drop UPHS scheme.

High-head design poses problems withrigidity of turbine parts, fatigue stresslimits, seals, and the manufacturing com-ponents with narrow passages. Attentionmust be paid to surface profiles and fini-shes to avoid cavitation and head losses.Serious stress and vibration problems ofthe wicket gates and their operating mech-anism will occur with increasing head.Higher heads also require longer andthicker stay vanes, and thicker wicket

gates and runner buckets, all of whichcould reduce turbine efficiency. To coun-teract this potential efficiency reductionrequires careful mechanical and hydraulicdesign of the runner to operate at higherspecific speeds than customarily used to-day. To increase the specific speed, eitherthe turbine speed or capacity must be in-creased. However, turbine speed is some-what restricted by the available synchronousspeeds of the motor/generator and manufac-turers of this equipment are reluctant togo to very high speeds. Efficiency lossesassociated with high-volume flow rates willalso limit the specific speed. In addition,some unidentified technological barriersmay exist which will limit the achievablemaximum head. Resolving these concerns re-quires conducting research studies andmodel tests. This activity is part of theongoing Department of Energy-UPHS program.The Hydro-Turbine Division of Allis-Chalmers Corp., with a contract from ArgonneNational Laboratory (ANL), has completed apreliminary design study of 500 MW, single-stage RPT units for operation at heads of500, 750, and 1000 m. A contract report ispresently being written.

Multistage reversible pump/turbineswithout wicket gates are available for thehead range proposed for UPHS. Some unitscurrently operational or under considera-tion are:

- Chictas-Italy, 1070 m, 4 stages, 150 MW- Edolo-Italy, 1290 m, 4 stages, 142 MW- LaCache-France, 930 m, 5 stages, 79 MW.

The multistage unit has the advantageof a developed pump cycle and requires lesssubmergence than a single-stage machine.Its specific speed per stage can be higherwhich increases efficiency, but this isoffset by additional losses between stages.Since these machines do not have wicketgates and therefore, cannot be regulated,they always generate a block-level of power.Therefore, the system loading must be capa-ble of absorbing this ungoverned gnerationwithout creating stability problems. Inthe pumping mode, these machines are allstarted in the watered condition becauseof technological problems in dewateringand priming. The main obstacle in the de-sign of multistage units appears to be thestructural and metallurgical design of thecasing which tends to limit the size of theunit.

Tandem units consisting of a Pelton-type waterwheel and a multistage pump

300

coupled to the same shaft are utilized inEurope for conventional, high-head pumpedstorage. Such units are:

- San Fiorino-Italy, ]440 m, 140 MW- Rottau-Austria, 1100 m, 200 MW.

The Pelton turbine must operate infree air and is normally situated abovethe highest tailwater. The pump has to bebelow the lowest tailwater level for cavi-tation-free operation. For vertical shaftunits, the shaft length is long, often inexcess of 30 m, and requires considerablesupport bearings. The powerhouse caverntends to be high and narrow which causesstress problems in the walls. To overcomethese disadvantages, a booster pump can beused to provide the necessary inlet pres-sure to the main pump. This technique re-duces the vertical shaft length and isutilized for horizontal shaft units.Operating the Pelton unit under an airback-pressure allows it to be located be-low the tailwater level. This techniquehas been sucessfully used at the Tysso IXstation in Norway. Tandem units have highefficiency over a wide head range, simpli-city of synchronous-condenser operation,ease of starting as a pump, rapid change-over from the pulping mode to the turbinemode, and vice versa. Its disadvantagesinclude high cost and limited capacity ofthe multi-stage pump section.

A two-stage RPT with wicket gates isbeing considered for UPHS application eventhough none of these units are operationalor on order today. Escher-Wyss has com-pleted model tests and performed mechanicaldesign investigations for such a unit. Thetwo-stage RPT with wicket gates is moreflexible than a non-gated design, and re-quires less submergence than a single-stageRFT for the same total head. It is capableof being used for higher heads than asingle-stage turbine, but requires initialdevelopment of a high-head single-stageunit. One disadvantage is the loss of ef-ficiency between stages. Hartmann andMeier7 point out that considerable effortwill be necessary in hydraulic research andmechanical design to bring this type ofmachine to a stage of development where theconstruction of a full-siae prototype willbe feasible. Allis-Chalmers has recentlycompleted a study for ANL on the preliminarydesign apects of such a machine. A reportis presently being written on this study.

This discussion has excluded separatepump and turbine units that have been com-

monly used in a number of European pumpedstorage plants. This scheme does notappear to be economically attractive andwould require the development of a high-head reaction turbine. Hartmann and Meier7

presented an arrangement called the "One +One." This idea basically combines twosingle-stage RPT units in series to forma two-stage machine. Several advantagesof this arrangement were cited, but addi-tional study is required to assess thefeasibility of this scheme. Considerationalso needs to be given to the controlled-flow rate pump/turbine of Gokhman,5 whichappears to be applicable for solar energystorage. This concept utilizes an adjus-table hub and a supplementary adjustablecover instead of wicket gates to obtain arelatively flat efficiency curve over avariable flow range.

UPHS ECONOMICS

Typical cost estimates for severalUPHS plant configurations are listed inTable 2. Included in the item designatedas reservoirs, dams, and waterways are theupper and lower reservoirs and water con-ductors with all the associated shafts ex-cept those connected to the upper servicebuilding. Tlu lower reservoir cost in-cludes excavation, disposal, rock-boltingand shotcrete. It comprises about 25-30%of the total plant cost, and is thereforethe most important cost component. Itemssuch as power plant structures, pump/tur-bines, generator/motors, and miscellaneousequipment are each about 8-10% of the totalcost.

The approximate costs for the turbo-machinery proposed for UPHS plants is shownin Fig. 6. These costs were obtained fromMitchell8 and include turbine governors andpenstock valves. The cost of a two-stageRPT with wicket gates will fall between thesingle-stage and multistage cost. Thesingle-stage RPT is the lowest cost unitwhich provides the incentive for the de-velopment of this type of high-head turbo-machinery. In addition, multistage andtandem units are size-limited and will re-quire more units for a given plant capacitythan the single-stage RPT. Additionalunits require more waterways and gateswhich increases plant cost.

The effect of head on the cost of a500 MW single-stage RPT is shown in Fig. 7.These data are obtained from the recentAllis-Chalmers study3 and show decreasingturbine costs with increasing head. This

301

trend is contrary to the multistage RPTcost data in the Chas. T. Main study2 whichpredicts increasing machinery cost withincreasing head.

Table 3. Cost Estimate Summary onSeveral UPHS Systems '

Item/System

Land and Land Rights

Power Plant Structuresand Improvements

Reservoirs, Dams, andWaterways

Pump/Turbines

Generators/Motors

Miscellaneous Equipment

Roads and Bridges

Contingencies, Engineer-ing, Supervision andOverhead

TOTAL

$/kW

I

2

71

293

76

52

51

6

183

734

367

II

2

89

320

44

49

58

6

194

762

381

III

2

75

297

117

54

52

6

197

800

400

*1) Adapted from Ref. 3. I: MultistageRPT; II: Single-Stage RPT (two-drop);III: Tandem Unit. All three plants are2000 MW, 1200 m head with 10 hrs storage.2) Units in 106 September 1978 dollars.3) $/kW in September 1978 dollars.

200

100

* 50

£ 40

g 30o

20

10

1 1 i i rAND PUMP (TANDEM)

NOTE:NOV. 1975 COST INCLUDINGGOVERNOR AND VALVESI I I 1 1

30 40 50 100 200UNIT SIZE, MW

300 500

Fig. 6. Capacity Effect of Cost of Turbo-machinery for Single-Drop Schemes6

18

16

§14-

12 -

10

SINGLE-STAGE RPT

NOTE: SEPT, 1978 COSTS

250 500 750 1000 1250 1500HEAD, M

Fig. 7. Effect of Head on Single- andTwo-Stage RPT Cost3

Recent studies1'2> s> 8> 10 have shownthat with other factors (such as energystorage capacity and types of turbomach-inery equipment) being held constant, therequired underground reservoir volume de-creases as the net head increases. There-fore, strong incentive exists to go tohigher heads to reduce reservoir cost.However, higher heads (>1000 m) result ina cost penalty from increased constructiontime. An optimum head should be reachedwhere any further increase in head willonly marginally reduce the overall systemcost. The general trend of system cost(in $/kW) versus net head is illustratedin Fig. 8. Estimated costs are directcosts and do not contain the effects ofinterest and escalation during construction;inclusion of these effects can produce po-tentially significant changes to economicconclusions.

RESEARCH AND DEVELOPMENT REQUIREMENTS

Even though the technology for UPHSis somewhat advanced, additional researchand development studies are required toimprove the economics, and eliminate anytechnological uncertainties that would hin-der large-scale commercialization. Researchshould be conducted to improve equipmentand construction techniques for excavatinglarge-diameter deep shafts and large under-ground caverns. Geological studies areneeded to define, evaluate and determinesuitable lower reservoir design and sta-bility criteria with emphasis on stress

302

500

400

200

100 -

NOTE:COSTS AS OF SEPT. 1978 PRICE LEVELTEN HOURS STORAGE, 2000 MW CAPACITY

600

Fig. 8.

900 1200HEAD, M

ISOO

Effect of Head on Single-DropUPHS Plant Cost (adapted fromRef. 2)

conditions of the surrounding rock. Tech-niques should be developed to determinethe underground geotechnical conditionswith surface detection methods.

Design, research, and testing programsare required for the development of large-capacity, high-head, single- and two-stageregulated RFT units. Research is neededto increase the unit size, efficiency, andregulation of multistage pump/turbinesCoupled with the turbomachinery developmentfor UPHS plants, investigations of themotor/generator design requirements shouldbe conducted. System studies to determinethe optimum head, equipment arrangement,unit size, type of turbomachinery, andplant size are required.

In addition, specific site locationsshould be identified "for the near-term con-struction of UPHS plants.

CONCLUSION

UPHS is a feasible energy storagescheme for utility peaking service. Itoffers potentially large benefits, espe-cially in premium fuel savings. Siting ofa UFHS plant will be determined by theavailability of competent geologic condi-tions and the proximity of load centersand transmission lines. Construction tech-niques for the underground reservoir andpowerhouse are well established, but im-provements should be pursued..' Environmen-

tally, a UPHS plant is more attractivethan alternatives such as conventionalpumped storage, combustion turbines, andcoal-fired cycling plants. Careful siteselection, planning, appropriate design,and the use of a dedicated man-made upperreservoir will eliminate or mitigate theenvironmental aspects of a UPHS plant.

Recent studies have indicated thatthe optimum head for a UPHS plant is about1000-1500 m. The highest head, single-stage RPT unit built today is 620 m.Therefore, to utilize available equipmentand operate at a high head requires a two-drop scheme, i.e., two single-stage tur-bines in series with an intermediate reser-voir. High-head multistage RPT units andtandem units (Pelton-type waterwheel plusa multistage pump) are available but arecapacity-limited and expensive. The de-velopment of high-head, single- and two-stage, regulated, RPT units is economicallyattractive, but requires extensive tech-nological development.

The head, unit size, plant size, plantarrangement, type of hydraulic machinery,and costs are not clearly defined. Thus,further system studies arj r1 quired todetermine optimum conditions. Studies arealso required to reduce plant constructiontime, improve mining techniques, and pro-vide rock mechanics criteria for under-ground reservoirs.

ACKNOWLEDGMENTS

The research activities in undergroundpumped hydro storage, which formed thebasis of this paper, were funded by theDivision of Energy Storage Systems, Officeof Conservation, U.S. Department of Energy.

REFERENCES

lSiting Opportunities in the U.S. for Com-pressed Air and Underground Pumped HydroStorage Facilities, Acres American, Inc.,New York (1976)

2Underground Hydroelectric Pumped Storage -An 'Evaluation of the Concept, DOE (Soli-citation 6-07-DR-500),(June 1978).

Chas. T. Main, Inc.

3Degan, J.R., Allis-Chalmers Hydro-TurbineDivision, unpublished studies (1977-1978).

303

"Applied Research on Energy Storage andConversion Photovoltaic and Wind EnergySystem, Vol. I, II, and III, GeneralElectric Space Division under contractNo. NSF C-75-22221 (Jan. 1978).

5Gokhman, A., University of Miami, unpub-lished work under ERDA contract No. EC-77-S-05-5516 (1977-1978).

6'Underground Pumped Hydro Storage andCompressed Air Energy Storage: An Anal-ysis of Regional Markets and DevelopmentPotential, ANL-R-77-3485-1, Harza Engin-neering Co. (1977).

7Hartraann, 0., and W. Meier, Developmentsin High-Head Pumped Storage, Water Power22(3):102-106 (March 1970).

"Mitchell, W.S., Underground Pumped HydroStorage, paper presented at the EngineeringFoundation Energy Storage Conf., PacificGrove, Cal. (Feb. 8-13, 1978).

9Pfaffin, G.E., Future Trends in HydroPumped Storage Equipment, presented atthe American Power Conference, Chicago(April 29-May 1, 1974).

10An Assessment of Energy Storage. SystemsSuitable for Use by Electric Utilities,EM-264, EPR1 Troject 225, ERDA E(ll-l)-2501 Final Report, Public Service Elec-tric and Gas Company, New Jersey (1975).

"Schmid, E.M., W.F. Adolfson, and C.K. Jee,Principal Geological Issues Related toUnderground Pumped Hydro Storage, draftreport contract No. EM-78-C-01-5114 (Sept.1978).

ESwiecichi, I., Trends in Pump-TurbineDesign, International Water Power andDam Construction, 29(1):45-47 (Jan. 1977);39(2)::42-45 (Feb. 1977).

13Willet, D.C., Underground Pumped StorageResearch Priorities, EPRI AF-182 TPS 75-618 Planning Study (April 1976).

304

PROJECT SUMMARY

Pro jec t T i t l e : Underground Pumped Hydro Storage-Single and DoubleStage Reversible Pump/Turbine Development.

P r i nc ipa l I n v e s t i g a t o r : John Degnan

Organ iza t ion : A l l is -Chalmers Corporat ionHydro-Turbine D i v i s i onP. 0 . Box 712York, PA 17405717-792-3511

Project Goals: Extension of the present state-of-the-art technology forhigh head single and double stage pump/turbines to headsbeyond 500 and 1000 meters respectively;

Generation of preliminary machine designs and correspondinghydraulic performance which together wi l l satisfy theoperating characteristics specified in the contract madebetween Allis-Chalmers and the Department of Energy (DOE)j

Projection of the costs for the various alternate machine designsdeveloped through tne study;

Evaluation of the feasibi l i ty of the single and double stagepump/turbine propositions as applied to Underground PunpedHydro Storage (UPHS).

Project Status: I t should be noted that this documentary addresses only thein i t i a l phase of a large development project extending into1982. The total project is concerned not only with the con-ceptual development of mechanical and hydraulic designs but withactual model testing of prototype heads.

The project's f i r s t phase has been completed and fu l lydocumented in a report issued to the Argonne National Laboratory.Completion of the project's next phase includes the preliminarydesign of multi-stage high head reversible machinery for thestorage of hydro-energy. This multi-stage concept as wellas the model testing for the single and double stage unitswhich are addressed in this paper w i l l be developed with fundsappropriated for 1979.

Contract Number: 31-109-38-4301

Contract Period: The f i r s t phase of the overall project was started June 1,1978 and completed in late September 1978.

Funding Level:

Funding Source:

$100,000.00

Argonne National Laboratory

305

EVALUATION OF ONE AND TWO STAGE HIGH HEADPUMP/TURBINE DESIGN FOR UNDERGROUND POWER STATIONS

John DegnanAllis-Chalmers CorporationHydro-Turbine DivisionBox 712, York, PA 17405

ABSTRACT

This paper attempts to summarize the significant design considerations found duringdevelopment of one and two stage reversible hydro machinery for the high head under-ground pumped hydro storage (UPHS) project currently funded through the U.S. Departmentof Energy (D.O.E.). The immediate objectives of the initial phase of this D.O.E.investigation are as follows:

" Extension of the present state-of-the-art technology for high head single anddouble stage pump/turbines to heads beyond 500 and 1000 meters respectively.

To generate preliminary machine designs which will comply with the requiredoperating characteristics for UPHS and also serve as a basis for developmentof model testing procedures and equipment.

To project the manufacturing costs of any proposed hydro-turbine equipmentresulting from the investigation.

' Evaluate the feasibility of the one and two stage pump/turbine propositions foroperation at a rated power of 500 megawatts under heads of up to 1000 and 1500meters respectively.

Examples of the preliminary single and double stage machine configurations areincluded within. Discussion of the special analyses and results leading to thesemachine designs is given and finally, conclusions as to the feasibility and limitationsof each type of machine are drawn according to their expected cost, mechancial andhydraulic performance.

INTRODUCTION

From the outset of the project, it wasbelieved that the maximum head whichcould be accommodated by a single stagemachine was approximately 1000 metersand 1500 meters for double stagemachines. Also it seemed obvious thatthe economics would not justify a larg-er two stage concept where head condi-tions allowed a single stage applica-tion. For this reason, both conceptswere developed within their expectedaJlowable head ranges.

Three single stage pump/turbines weredeveloped with rated net heads of 500,750, and 1000 meters respectively andeach with a rated power of 500 mega-watts. The 1000 meter machine will beused to exemplify the points of discus-sion since this machine is most removed

from current design and operation expe-rience.

Also, four separate double stage designswere investigated. One of these wasrated at 350 megawatts operating at arated 1250 meters head while the otherswere designed for a rated power of 500megawatts operating at rated head of1000, 1250, and 1500 meters respective-ly. The 1000 meter machine will be usedto exemplify significant design consid-erations of the two stage pump/turbineconcept.

The feasibility of the mechanical designsfor these machines was verified usingvery accurate computer models which pre-dict structural behavior and stress con-ditions based on the finite elementmethod of analysis.

306

The hydraulic shape of the water passagesand the performance for the single stagemachines were predicted from existinghigh head pump/turbine model configura-tions and test data. The hydrauliccharacteristics of the two stage machinesrequired more rigorous studies. The per-formance estimates, for instance, involv-ed using known two stage pump performancecurves or a combination of the perform-ance characteristics of two single stageunits which were similar to the respec-tive stages of the pump/turbine understudy. Any expected losses betweenstages were then accounted for. Ofcourse, determination of the actualhydraulic performance can only be real-ized when the model tests are made.

SINGLE STAGE PUMP/TURBINE DESIGN

GENERAL

Figures 1 and 2 show the 1000 M singlestage pump/turbine preliminary design.The major component designs were gen-erated automatically using the IRISprogram. The engineering design cal-culations performed in IRIS proved togive a sensible optimized design. Thisis substantiated through the specialcomputer studies whose results indicatedonly minor geometry changes would benecessary to afford a sound singlestage pump/turbine design.

Flu. 1 Sln* l . «t>||» 1000 F/T - dl.trlNKor nrcllon

rig. 2 until »t«i« 1000 Mttr-41:trlbutor f lu

SPIRAL CASE - STAY RJMG

The most significant criteria used todesign the configuration of these inte-grated components is that their mechan-ical interaction be related so any netmoment generated in the stay ring crownduring the design condition (maximumwaterhammer pressure outside the wicketgates and tailwater head inside the gates)and pump prime (full shut-off head insideclosed wicket gates with maximum static

ne. Stress Distribution! 1I Pump Shut-off *

Fig. 3 Stay ring vane stresses

307

head outside) cause equivalent tensilestresses on the outside and inside diam-eter of the stay vanes respectively.Such a design criteria truly produces anoptimized structure since this techniqueminimizes the stress amplitude section bysection through constructive use of theresidual ring moments. Figure 3 showshow these moments are controlled by thesection's spiral case to stay ringattachment location. Note how the spiralcase attachment point slides along thestay ring cone until the optimizingcriteria are satisfied.

HEADCOVER

The mechanical design of this componentis optimized based on the structure'srigidity or its resistance to distortunder load. Such parameters as maximumhoop stress, maximum angular rotationof the head cover and maximum deflectionat the turbine bearing support arechecked against preset limits in theIRIS design system according to a log-arithmic type head cover analysis. Thestructure is incremented in size accord-ingly until all optimizing criteria aresatisfied. The following boundary con-ditions were fovced on the IRIS designof the 1000 M single stage pump/turbine:

Maximum Hoop StressMaximum Angular RotationTurbine Bearing RadialDeflection

17000 PSI.0008 radians.0500 inch

It will become apparent from the stressanalysis of the distributor section whythese parameters characterize the generalbehavior of the head cover.

DISCHARGE RING

This component draws its sizing from thebasic head cover configuration. Themost important feature of the dischargering is that it is of heavy constructionand can support its own pressure loadingwithout relying on a distributed founda-tion reaction. This full strength dis-charge ring design has the advantage ofallowing only 40% of the vertical headcover load, F , to be realized at themachines' foundation; the remaining. 60%of the head cover load is compensatedby the discharge ring load. Figure 4shows this concept graphically.

DUcMi|« nai vittlcal lewd*

It should be noted that cheaper, lighterdischarge ring designs can be proposedso that the only compensation for theextreme head cover load is in the foun-dation reaction. This huge load, how-ever, becomes increasingly difficult orimpossible to accommodate as pressureheads increase and machines necessarilybecome smaller.

DISTRIBUTOR SECTION ANALYSIS

The 1000 meter single stage pump/turbinedistributor components have been analyzedusing the axisymmetric-two dimensionalrepresentation including the dischargering, spiral case, stay ring, head cover,and connecting bolts (Figure 5).

This analysis will provide the followinginformation:

a. A general understanding of thedeflection and stress pattern inthe assembly.

b. A good evaluation of the interactiveforces between components, for exam-ple, the calculation of the radialforces developed at the stay ringto head cover and stay ring to dis-charge ring bolting flanges.

c. The calculation of the necessarybolt prestress required to eliminatefatigue.

d. Th-s foundation loading and requiredprestress of the anchor bolts.

Ten basic loading conditions were runsimultaneously by the finite element pro-gram . These cases are factorized and

308

superimposed using a post processor pro-gram to represent the following relevantoperation conditions:

Mormal Running Condition. The machineoperates at maximum static hend. Thepressure distribution inside the runnerperiphery follows a forced vortex par-abolic law assuming a" mass of waterrotating at half the speed of the runner.The maximum tailwater head is appliedinside the runner seals.

Pump Shutoff Condition. This loadingcase represents the condition occurringduring a pump start after the releaseof the pressurized air. The pressureacting outside of the wicket gate isequal to the maximum static head. Thepressure acting from the wicket gate tothe runner periphery is equal to themaximum shutoff head. This pressuredecreases parbolically from the peripheryof the runner to the runner seals asdescribed in the first case. The pres-sure inside of the seals is equal to themaximum tailwater head.

Pressure Rise, Wicket Gates Closed. Thisloading case represents the normal tur-bine shutdown operation with the maximumpressure rise and maximum tailwater headacting, respectively, outside and insideof the wicket gate.

Runaway Condition. This loading caserepresents the extreme condition of tur-bine load rejection with the wicket gatesstuck open. The maximum pressure riseacts up to the runner periphery. Thispressure decreases parabolically fromthe runner periphery to the runner sealfollowing a forced vortex law with themass of water rotating at 50% of therunner runaway speed. The maximum tail-water head is used inside of the seals.

Results Interpretation. The use ofdeflection results for the purpose ofrigidity evaluation or field test com-parison has to be based on differentialvalues between two points of the struc-ture. This is necessary in order toestablish a common fixed datum. Forexample, the vertical deflections of ahead cover can be defined as the dif-ferential vertical movement between anypoint on the head cover and the stayring bolting flange interface.

Some of the important information anddesign criteria extracted from the com-puter study are defined in Figure 5 andtabulated in Table 1 and 2. They are:

- "A", axial deflection of top deck atthe inner radius. The value is cal-culated relative to the head cover tostay ring interface at the bolt center-line.

- "B", radial deflection of top deck atthe inner radius. This value has tobe considered very carefully for thisdesign where the turbine guide bearingis bolted directly to the top deck.

- "C", axial deflection of top deck atthe outer radius.

- "D", radial movement of wicket gatetop bearing.

- "E", radial movement Of wicket gateintermediate bearing.

- "F", gap opening at head cover to stayring seal interface. This value ischecked to assure that "0" ring extru-sion will not occur.

309

Table 1.

Node

A*BC*DEFGJ*aBe

Table 2.

KLMN

*referenced

HEAD COVER DEFLECTIONS AND ANGULAR

Generating

.10018

.04012

.01082

.03379-.03069.02388.06635.11848.00062.00045.73116

Pump Prime

.15990

.06165

.02055

.05196-.04820.04053.10420.18468.00096.00070.73310

ROTATIONS (cm & radians)

Pressure Rise

.03608

.01732

.00077

.01459-.01348.00853.02963.04760.00024•00020.82353

FOUNDATION LOADS AND COMPONENT INTERACTIVE FORCES

2,830,3361,518,0214,573,6064,335,834

to datum

3,428,0731,286,2297,047,6737,223,523

3,068,8202,638,5062, 005, 5941,122,144

Runaway

.14961

.05861

.01941

.04958-.04561.03988.10232.17294.00089.00067.75144

, KG

4,135,6662,204,4686,701,2366,500,989

- "G", distributor height growth. Thisvalue is particularly important atpump start and at shutdown when theadditional clearance has to be filledby the movable gate end seal to limitleakage and wire drawing.

• "J", axial deflection at the shaftseal bolting flange. This is usuallythe largest axial deflection recordedin a bead cover.

"a", head cover top deck angular rota-tion. This is one of the most impor-tant head cover design criteria. Itsmagnitude is inversely proportional tothe head cover rigidity. Experiencehas shown that a should be lower than0.001 at the normal running conditionto insure a satisfactory performancein a straight Francis turbine. A moreconservative value of 0.0009 radiansis used for pump/turbines.

"8", wicket gate bearings angularrotation. The same criteria used for

"9", is the ratio of gate bearingangular rotation to head cover topdeck angular rotation. The magnitudeof 6 is an indication of the efficiencyof the transfer, by the ribs, of theshear load between the head cover topand bottom deck.

9 = 1.0 for an infinitely rigidribbing

Foundation Load and Components Inter-action. The component interactive forcesand foundation loading are defined inFigure 5. These are the radial forcestending to separate the stay ring fromthe head cover "K" and also, from thedischarge ring "L", plus the inner andouter foundation reactions "N" and "M".These values are tabulated in Table 2.

WICKET GATE

A critical balance exists between thehydraulic and mechanical requirements ofa guide vane system. It is essential toassure that neither the hydraulic per-formance nor the mechanical considera-tions of long life and reliability becompromised.

The wicket gate system for the 1000 meterpump/turbine was also generated by theIRIS program. The preliminary designof the wicket, gate is achieved by incre-menting the gate geometry until allstatic and dynamic criteria (stress,deflection and torsional frequency) aresatisfied. A schematic representationof this optimization process used byIRIS is shown in Figure 6.

310

MCLINIHUT STRESS AKALtttlINITUL SUN * ICAF GEOMETRY•EARIHC REACTIONS:TORQUE!

LINKAGE OPTIM1I*TION tSERVOMOTOR SUING

Ist>ESS, OEFLECTIONS I RCAC*DON CALCULATIONS FORDIFFERENT CONDITIONS

CHECK LIMITING DESIGN CRITERIA

- STRESSES- DEFLECTIONS• SLOPE * f 9 E M I W S• STEM ANGULAR ROTATION- lEARING PRESSURES

j

INCRCKEHTKL GEO-HCTRV CHANGt ONLEAF ON STEHACCORDING TOUMSATW.tB OfSIClCRITERIA

—-czml

\ FIHAj QESICH • OUTPUT

g. 6 Icii pro^ru q«t optimisation

TAIL SEAL COfTTACt LVtl _ ,

-TIUUULATIQSAI. CM3TM1KI

-THIUST COXSTUIVT

—TtANStATKHAL CONSTRUCT

_ S O S t SEAL COfTrXT LIKE

^ossnwwT

WICKET GATE STRUCTURAL ANALYSES

Since the classical Beam theory used inIRIS does not adequately represent thebehavior of the leaf and stem to leafintersection, a three-dimensional finiteelement static analysis of the gate wasconducted. Figure 7 shows the mathemat-ical model. The structural response tothe following loading cases has beenevaluated:

Gate Squeeze Condition. The gate stemis subjected to the maximum torque avail-able at closed position while the gateleaf is subjected to the maximum tran-sient head and the intergate seal reac-tions which are presented as contactpressures in Figure 8.

Shear Pin Failure Condition. The gatestem is subjected to the maximum torquerequired to break the shear pin at ornear the closed position. This conditionoccurs when an obstruction becomes lodgedbetween any two gates preventing furtherclosure of the gates. The highest stress-es result from the blockage which islocated at the upper tail end of the leaf.Since full servomotor torque is availableto these two gates alone, they must beprotected by insertion of a shear pin intheir linkage mechanisms. The pin wasdesigned to break at 225% of normalservomotor torque.

Fatigue Analysis. Once all the signifi-cant static loading conditions areanalyzed, evaluation of the fatique lifeis done. The Miner's cumulative damagerule is typically used which is based onthe linear summation of the fractions offatigue damage expressed in terms of thecycle ratio (n./Nf)

where n. = estimated if of cycles ofeach mode of loading

311

# of cycles to failure asdetermined from materialspecifications. This is afunction of the mean andcyclic stresses occurringat potential sites of fail-ure.

gives solid evidence to the integrity ofits conceptual design.

Z(n±/Nf) = 1 failure is hypoth-esized

A conservative approach in using theanalysis is to limit the accumulateddamage to an order of magnitude lessthan the predicted level so that Z(n./N^)< 0.1 indicates a reliable design.

HYDRAULIC PERFORMANCE

In addition to the mechanical develop-ment, the hydraulic performance of themachines have been evaluated. Figure 9shows the performance as a pump,

CONCLUSION

These studies indicate that both thewicket gate and the components of thedistributor section will, in fact, behaveaccording to the criteria used to sizethem. The analyses show, for instance,that the 1000 meter machine will:

* develop foundation loads which havemagnitudes of only 40% of the respec-tive head cover loads (18,927,000 Kgat Pump prime)

' that the maximum hoop stress developedin the head cover is 17,970 psi

' the axial stress at the inside of thestay vane during pump prime and at theoutside of the vane during the tran-sient design conditions are approx-imately equal at each spiral case sec-rtion.

Also, it was evident that, despite theextreme conditions of operation assumedin the wicket gate analysis, the maximumaccumulated damage was 0.000154 for aperiod of 50 years or less than 0.002%of the damage required for fatigue fail-ure. Subjected to normal design loads,a near infinite life can be expected forthese gates.

The general stress and deflectionresponse of this machine to its loading

i: M u

uoo

1050

LOW

tit. S fmf p«t[01MK. 1000 HUr but ll»|U ftM* «

Fig. 19 D.O.E. two atage 1000 meter P/T - distributoranalysis finite clenent nodel

312

TWO STAGE PUMP/TURBINE DESIGN

GENERAL

Figures Hand 12 show the 1000 meterdouble-stage pump/turbine preliminarydesign. Some of the basic componentdesigns such as the impellers, headcover, gate and spiral case configura-tions have been initially sized by theIRIS program. Other components suchas the stay ring, intermediate waterpassage, and the shafting have beendesigned by careful layout coordinatedwith the IRIS designed components.

At the outset of the development project,little was known as to the interactivebehavior of the components making up thismachine. In order to determine this, astatic stress and deflection analysis wasperformed on an intermediate design sub-jected to its normal loading during gen-eration mode. Although the study wasvery preliminary in nature, it providedthe required information to develop afeasible solution.

DISTRIBUTOR SECTION

An intermediate design of the 1000 metertwo stage pump/turbine has been analyzedusing an axisymmetric-two dimensionalrepresentation. Figure 10 depicts themathematical model developed for thefinite element structural analysis. Theanalytical results provided for:

' a general understanding of the deflec-tion and stress pattern in the assembly

• a good evaluation of the interactiveforces and moments between components.This information is essential to prop-erly design connection of the compon-ents.

M». 11 Ito •teg* 1000 Mter r/T ~ distributer Mctloa

the calculation of the bolt prestressrequired to eliminate flange slippageand bolt fatigue.

foundation loading required to designthe powerhouse.

The proposed design as shown in Figures11 and 12 incorporates many modificationsbased on the above analysis.

SHAFTING SYSTEM DESIGN

The double stage machines were initiallydeveloped so that each stage generatedequal amounts of power, however, a basicproblem was brought to light concerningthis design: The critical speed of theshafting system is greatly influenced bythe amount of overhung runner mass andits distance from the turbine guide bear-ing. It was apparent that it would beadvantageous to design the second stagerunner to carry as much of the impellersmass as possible due to its nearness tothe guide bearing. This was accomplishedby designing the first and second stageto develop 40% and 60% of the powerrespectively. This concept was verifiedthrough critical speed dynamic analysesperformed on the proposed pump/turbineimpeller, shafting and bearing withassumed shafting and generator systems.The modification caused the fundamentalcritical speed of the shaft system(lateral mode) to be increased from 8.3Hz to 13.0 Hz or to 73% above the normalspeed.

313

Another subject of notable considerationwas the seal clearance of the first stageshafting. Again, due to the long over-hung length of the turbine shaftingdownstream of the guide bearing, largeradial deflections could occur duringtransient conditions which must beaccommodated without compromising propersealing and thus performance. It isproposed that the interstage seal bedesigned so that during transient loadingconditions it will serve as a secondaryguide bearing. A carbon impregnatedmetal material has been- developed whichresists corrosion and has varying degreesof self-lubrication and load carryingcapacity. If contact should occur dur-ing short transient periods between thisstationary seal material and the rotatingshaft, the seal can indeed function as aself-lubricated guide bearing.

WICKET GATE

The same design criteria considered onthe single stage machines apply to thiswickfit gate design. Again, finite ele-ment stress, deflection, and fatigueanalyses were performed to verify theinitial sizing. According to thesestudies, the wicket gate design presentsno problems which would adversely effectthe feasibility of this two-stage con-cept even when subjected to the limitinghead of the 1500 meter machine.

HYDRAULIC PERFORMANCE

In addition to the mechanical develop-ment, the hydraulic performance of themachines has been evaluated. Figure 13shows the performance of the machine asa pump.

CONCLUSION

Although further development work isindicated to design double stage machineswhich would have the same degree ofoptimization as is built into the singlestage machine, the studies to date aremore than sufficient to prove the fea-sibility of the two-stage concept froma mechanical standpoint for heads between1000 and 1500 meters.

FEASIBILITY

The preliminary designs and hydraulicevaluation of these machines give witnessto their useful potential.

The manufacturing cost analysis producedresultc showing that for the same applica-tion, the two stage machine costs are 55%greater than the single stage unit. Itshould be pointed out that project con-struction, machine efficiency, submergenceand maintainability all must be weighedand evaluated to determine the feasi-bility of such a project. Current con-struction rates, for instance, probablyhave the greatest influence on thefeasibility of a project. The amount ofexcavation for the project site is depend-ent on the required submergence of thehydraulic machine. This submergencerequirement is inherently greater for asingle stage unit than a two or multi-stage machine.

REFERENCES

1. Chacour, S. and Graybill, J. D.,"IRIS, a Computerized High HeadPump Turbine Design System," ASMEPaper No. 76-WA/FE-12, December 1976.

2. Chacour, S., "A High PrecisionAxisymmetric Triangular ElementUsed in the Analysis of HydraulicTurbine Components," ASME PaperNo. 70-FE-19, Reprinted in theTransactions of the ASME, Journalof Basic Engineering, December 1970.

3. Chacour, S. and Degnan, J., "Structur-al Optimization of High Head Pump/Turbines," CEA, March 1977.

314

PROJECT SUMMARY

Project Title: Multistage Turbine-Pump with Controlled Flow Rate

Principal Investigator: Alexander Gokhman

Organization: Department of Mechanical EngineeringSchool of Engineering and ArchitectureUniversity of MiamiCoral Cables, FL 3.3124Telephone: (305) 284-4848

Project Goals: The goal of this project was to conduct a conceptual design ofone and two stage turbine-pumps with controlled flow rate foruse with pumped hydro storage power plants. Both the hydraulicand mechanical designs of the turbine-pumps were to be evaluated.

Project Status: The project was completed on September 30, 1978,.and a finalreport was prepared and submitted to DOE/STOR in October 1978.

The hydraulic analysis shows that this new machine has the uniqueability to regulate the flow rate even in the pump mode of operation.This unique feature makes the new turbine-pump especially attractivefor above or below ground pumped storage plant applications tostore the energy produced by plants utilizing non-controllablenatural energy sources such as solar energy, wind energy, etc.

The new turbine-pump with controlled flow rate is more complicatedand, consequently,, more expensive than conventional unit for thesame operating parameters. However, the new machine does not requirean operating valve since the outer cylinder of the movable upper bandof the distributor acts as a cylindrical valve. Therefore, it is notcertain that the new turbine-pump will cost more than a conventionalunit which requires a costly operating valve.

The comparison of the new machine and a conventional unit requireshydraulic experiments and technical design by leading firms in thefield of hydraulic machinery. Even if the new machine proves to bemore expensive than a conventional unit with a regulating valve, theadditional cost will be offset by the energy savings due to flowregulation.

Contract Number: EC-77-S-05-5517

Contract Period: June 1, 1977 -Sept. 30, 1978

Funding Level: $74,582

Funding Source: Department of Energy, Division of Energy Storage Systems

315

MULTISTAGE TURBINE-PUMP WITH CONTROLLED FLOW RATE

Alexander Gokhman, Nail OzboyaUniversity of Miami

Department of Mechanical EngineeringCoral Gables, Florida 33124

ABSTRACT

The presented paper shows the results of the preliminary part of the development ofthe new hydraulic machine "Multistage Turbine-Pump with Controlled Flow Rate." The hy-draulic analysis of the turbine-pump shows that this new machine has the unique featureto regulate the flow rate even in the pump mode (the conventional turbine-pumps cannotcontrol the flow rate in the pump mode). This unique feature makes the new turbine-pumpespecially attractive for the pumped-storage plant applications to store the energy pro-duced by the plants utilizing non-controllable natural energy sources such as solar energy,wind energy, etc. Obviously, during the process of storing energy, the conventional ma-chine will cause the waste of energy produced due to its lack of ability to regulate theflow rate. The new turbine-pump with controlled flow rate is more complicated and conse-quently more expensive than the conventional one for the same parameters (head, power) ascan be concluded from the conceptual design of the new machine. However, the turbine-pumpwith controlled flow rate does not require an operating valve, since the outer cylinder ofthe movable upper band of the distributor (in case of multistage machine, the distributorof the first stage) acts as cylindrical valve. Therefore it is not certain that new tur-bine-pump which does not need the operating valve will cost more than conventional turbine-pump with the operating valve (It is well known that the high head valve for pov;er plantsis very expensive). Even in the case that the new machine is more expensive than the con-ventional one with the operating valve, the additional cost of the new machine will beoffset by the energy savings due to the unique feature of flow rate regulation.

INTRODUCTION

Modern turbine-pumps utilized forpumped storage are divided into two groups.The first one is one-stage turbine-pumps.The largest head for such machines is 800 m(Hitachi (Japan)). Allis-Chaltners (U.S.A.)advertised the conceptual design of a one-stage turbine with a 1000 m head, but otherleading companies such as Escher-Wyss(Switzerland), Hydroart (Italy) and Nyer-pic (France) consider the limit for headfor a one-stage machine to be around 500 m.One-stage conventional turbine-pumps havethe ability to regulate power in turbinemode, however in pump mode they can onlywork with nominal power or not work at all.Multi-stage machines cannot regulate pow-er in either mode, but they seem to be morereliable especially for high heads owingto the absence of guide gate apparatus withadjustable vanes. In a one-stage turbine-pump working as a pump the guide gate ap-paratus become the sources of strong vibra-tions, because the flow after the runner isunsteady and causes pressure pulsationsaround the guide vanes, which are not ri-

gidly fixed. Therefore all conventionalpumps, one-stage or multistage, have somedisadvantages in addition to lacking theability to regulate power in pump mode.Neyrpic (France) is trying to develop a onestage turbine-pump, but their proposal doesnot look very promising. The absence ofpower regulation in pump mode is a big dis-advantage in certain conditions, e.g. whenthe nominal power of a unit is comparableto the power of the system. Lack of abil-ity of multi-stage turbine-pumps to reg-ulate power in turbine mode is even a big-ger disadvantage, because the machine worksin turbine mode only during peak time withsharp changes in power.

The ability to regulate power is vi-tal to pumps working with non-controlledoutput electrical plants, e.g. solar plants,because it leads to inevitable power losses,except for those few moments when the pow-er of the plant matches the sum of capaci-ties of several pump units. These lossescan be as high as 95% of the capacity of a

316 -

single unit, because the only available wayof regulation in such a case is a completeshut down of one or more units. Howeverthere exists a possibility to eliminate theabovementioned disadvantages of convention-al turbine-pumps in both turbine and pumpmodes. Power regulation can be achieved inboth modes by changing at will the heightof the water passage in tho runner and thedistributor, the latter having rigidly se-cured vanes. '

This work accomplished under the De-partment of Energy contract was devoted tothe conceptual design of such a machine.According to this contract the researchersalso had to develop all the necessary hy-drodynamic programs for this task, so theaccomplished work consists of the properhydraulic analysis of the turbomachine withvariable water passage height in the dis-tributor, the conceptual designs of one-stage and two-stage turbine-pumps of thistype accompanied by all the necessary hy-draulic, force and stress calculations.

The goal of this entire project is toinvestigate the feasibility of machineswith this new principle.

DESCRIPTION OF THE DESIGN

ONE-STAGE TURBINE-PUMP

The one-stage turbine-pump was de-signed for a 450 m head, 48 MW capacity and1.8 m diameter.

The one-stage turbine-pump with con-trolled flow rate (TPCFR) comprises thescroll casing (23) fixed inside of the se-condary support concrete structure (24),the distributor placed inside of the scrollcasing and secured to the scroll casingbody, the runner placed inside of the dis-tributor and secured to the turbine shaft(9), and the draft tube (30) attached tothe lower ring (17) of the distributor(see Fig. 1).

The distributor comprises the upperring (21), lower ring (17), guide vanes(16) welded to the upper and lower rings,and movable upper band (18). The movableupper band (18) is made up of the outerhorizontal flange, outer cylindrical part,flat bottom part and inner cylindricalpart, all forming a solid body. The lowerring is formed by the outer cylindricalpart, flat top part and inner cylindricalpart. The horizontal flange of the upperring (21) of the distributor is fastened to

the horizontal flange of the scroll casingby means of studs (26) . The outer cylinderof the lower ring is fastened to the verti-cal inner cylinder of the scroll casing bymeans of studs (25). As seen from Fig. 1,the fastening of the distributor to thescroll casing by the horizontal flange andvertical outer cylinder gives the distri-butor absolutely fixed position and goodrigidity. The outer cylinder of the mov-able upper band (18) is placed inside ofthe inner cylinder of the upper ring (21).The movable upper band is driven by thevertical servomotors (37). The servomotorguide rods (19) are fastened to the flatbottom part of the movable upper band (18)on one end and to the servomotor rod on theother end by the servomotor rod connector(39). The rubber seals (20) are placedbetween the vertical inner lower cylinderof the scroll casing and the outer cylin-der of the movable upper band. There areadditional rubber seals (22) between theinner cylinder of the upper ring and in-ner cylinder of the movable upper band.

The turbine cover (32) is fastenedrigidly to the upper horizontal part of thescroll casing by means of studs (33) andto the lug (45) of the upper ring by meansof studs (46). It is clear from Fig. 1that the turbine cover, the horizontalflange of the upper ring (21) and thehorizontal flange of the scroll casing (23)are pressed together. This design pro-vides the turbine cover with fixed andrigid position.

The servomotor guide rod (19) whichmoves the movable upper band (18) isguided through the bronze bushing which istightly pressed in to the upper ring ofthe distributor.

The servomotors are mounted on thehorizontal plate (44) which is welded tothe turbine cover (32) and the support(38). The servomotor rod is also guidedby the bushing placed in this horizontalplate (44). There are 4 servomotors andthe diameter of the servomotor piston is200 mm. The pressure of the oil in theservomotor is 70 kgf/cm2.

The outer cylinder of the movable up-per band (18) has enough height to closethe water passage completely. The rubberseal (20) between the scroll casing andthis cylinder will prevent the leakage ofwater between them. In order to preventthe leakage of water between the horizon-tal bottom of movable band (18) and the

317

I

318

lower ring (17), a rubber seal (47) is in-stalled in the groove of the lower ring.This rubber seal is pressed by the adjust-able metal ring (48) which is fastened tothe lower ring (17). It is obvious thatthis seal is replaceable. This designgives the upper movable band the abilityto function as a cylindrical valve.

The labyrinth seal (43) for the run-ner is mounted to the inner cylinder of themovable band (18).

The runner comprises the lower ring(2), upper ring (4), runner blades (1)welded to the upper and lower rings,streamlined hub (3), hub extension (14),connecting ring (5) and upper cover disk(6) of the runner.

There are 15 runner blades (1) andeach blade is designed for the minimumheight of the water passage ((bo)opt), theupper part of the blade is a vertical ex-tension of the blade cross-section atbo = (bo)opt. The runner blades (1) arewelded to the lower ring (2) of the run-ner. The streamlined hub (3) of the run-ner is then placed on the lower ring (2) ofthe runner, the runner blades passingthrough the slots of the streamlined hub.The upper ends of the runner blades (1) arewelded to the upper ring (4) of the runner.The connecting ring (5) is welded to theupper cover disk (6) of the runner and theyform a solid piece which is placed on thestreamlined hub (3) of the runner and fas-tened to it by means of cap screws.

The guide pivot (49) is secured tothe upper cover disk (6) on the upper sideby cap screws and screwed to the stream-lined hub (3) on the lower side and itpasses through the hole in the upper ring(4). While the guide rod guides thestreamlined hub (3) and the upper coverdisk (6) assembly during its vertical move-ment, it also gives additional strength tothis assembly. The alternate guide pivots(49) are left with holes in them in orderto direct the flow from the space betweenthe turbine cover (32) and the upper coverdisk (6) to the downstream of the runnerblades (1), therefore reducing the excesspressure in that space. The leakage ofwater in that space to outside is prevent-ed by the labyrinth seal (35) between thehollow turbine shaft (9) and the seal sup-porting disk (34) which is mounted to theturbine cover (32).

The streamlined hub (3) of the runner

is attached to the guide rod (8) by means ofstuds (13). The streamlined hub extension(14) in turn is mounted to the streamlinedhub (3) by means of bolts (15). The bolts(11) secure the upper ring (4) of the runnerto the hollow shaft (9).

During the regulation of the flow rate,the cross-sectional area of the water pas-sage is changed by the vertical movement ofthe guide rod (8) inside the hollow shaft(9) and the subassembly of the streamlinedhub (3), hub extension (14), connecting ring(?) and the upper cover disk (6) of the run-ner. The lubricating oil provides thesmooth sliding of the guide rod (8) in thehollow shaft and the bushing (10) guides therod (8). The leakage of oil between theguide rod and the hollow shaft is preventedby the rubber seals (12) mounted to the up-per ring (4) of the runner.

The vertical movement of the guide rod(8) is provided by regulating the pressureof the oil in the servomotor on the upperand lower sides of the piston (7) mountedto the guide rod (8). The oil supply to theservomotor in the hollow shaft is similarto that used for adjustable blade turbines.

Guide bearing (41) is the conventionaloil-lubricated type bearing.

The draft tube (30) is attached to thesupport disk (28)' which is in turn mountedto the lower ring (17) of the distributorby means of cap screws (29). The labyrinthseal (27) attached to the lower ring (2) ofthe runner and inserted into the grooves ofthe draft tube flange reduces the water lossfrom the upstream of runner to the drafttube.

The governor of the machine is similarto the governors of conventional Francisturbines. The only difference is that theservomotors of the distributor have to besynchronized with the servomotor in thehollow shaft of the turbine.

TWO-STAGE TURBINE-PUMP

The layout of two-stage TPCFR is shownin Fig. 2. The two-stage TPCFR is designedwith the stages having the same hydraulicparameters, i.e., the same geometry of thewater passage and the same diameter of therunner.

The main differences in the designs ofthe mechanism of flow rate regulation for

319

320

one-stage and two-stage TPCFR are in therunners, shaft, shaft support and secondstage distributor.

The first stage streamlined hub (3)of the runner is not driven directly by theguide rod (8) in the hollow shaft (9), butby means of a special connecting cylinder(71). The connecting cylinder (71) is madefrom two halves and connected to the uppercover disk (66) of the second stage runner.

The second stage runner is essential-ly the same as the one-stage machine run-ner, except for two differences. The firstdifference is that the upper cover disk(66) has a longer cylindrical part and anadditional inner disk which is secured tothe connecting cylinder (71) by means ofscrews (73). The connecting cylindertransfers the vertical motion of the uppercover disk of the second stage runner tothe streamlined hub (3) of the first stagerunner. The second difference is that thestreamlined hub (63) of the second stagerunner is not connected directly to theguide rod (8) but to the carrying arms(76 and 93).

The design of the upper part of theturbine shaft (9) above the first stagerunner is similar to that of a one-stagemachine. However, the diameters of thepiston inside the turbine shaft and theflange of the shaft are larger in order toprovide a larger force by the servomotorin the shaft. Also the flange for connect-ing the first stage runner to the turbineshaft (9) is wider in the two-stage ma-chine. The diameter of the shaft belowthis flange is smaller, since the torqueon the shaft between the two stages ishalf of the torque on the upper part. Thelower flange of the turbine shaft is usedto attach the second stage runner to theshaft and the diameter of this flange issmaller because of assembly requirements.The lower end of the turbine shaft istightly pressed into the opening of theupper ring (64) of the second stage runner.There are two bushings (10 and 50) on theturbine shaft located between the runners.They direct the movement, of the connectingcylinder (71) and each bushing is made oftwo half cylinders. The turbine shaft (9)is hollow like in the one-stage machineand the guide rod (8) secured to the piston(7) of the servomotor is placed inside theshaft (9). The diameter of the shaft hole •and consequently the diameter of the guiderod (8) are 320 mm. above the first stagerunner and 240 mm. between the runners.

The diameter of the guide rod below the se-cond stage runner is 168 mm. It passesthrough the bushing (51) mounted to thehole of the shaft. This bushing directs themotion of the guide rod (8). The verticalmovement of the guide rod (8) is transferredto the streamlined hub (63) of the secondstage runner by means of two carrying arms(76 and 93) which are inserted into therectangular holes in the guide rod (8). Theupper arm (76) is placed in the groove ofthe streamlined hub (63) and secured to itby the screws (77). The lower arm is per-pendicular to this upper arm (76) and is al-so fastened to the hub (63). Both arms passthrough the openings in the turbine shaftwall so that they can move up and down withthe guide rod (8) in the range of regulation(in our case 90 mm). The arms (76, .93) arealso secured to the guide rod (8) by meansof screw (78). The rubber seals (12) onthe turbine shaft above the opening for theupper carrying arm (76) prevent the leakageof oil. These seals (12) are pressed by thering (94) which has two partial flanges inthe opening for the arms. The labyrinthseal (72) between the upper cover disk (66)of the second stage runner and the innercylinder of the upper ring (52) of the se-cond stage distributor is mounted on theinner cylinder of the upper cover disk (66).The hollow shaft (9) has a solid extension(79) which is secured to the turbine shaftby the screws (80). Since there is almostno torque acting on the shaft below the se-cond stage runner, the shaft extension (79)is this and the connection to the shaft (9)does not have to be very strong. The onlytorque is due to the friction in the waterlubricated lower guide bearing (81). Thisbearing is necessary in order to eliminatethe shaft beat (the shaft of the two-stagemachine is approximately 3 tn. long comparedto the 1.2 m. long shaft of the one-stagemachine). The lower guide bearing (81) ismounted inside of the metal base (82) and awater pipe (85) supplies the clean water forthe lubrication and cooling of the bearing.

The second stage distributor comprisesthe lower ring (60), upper ring (52), sta-tionary guide vanes (86), movable guidevanes (55) and movable upper band (54). Thestationary guide vanes (86) are welded tothe lower and upper rings (60 and 52) andthey form a rigid structure. The lowerring (60) is pressed tightly in the cylin-drical recess of the lower turbine cover(58) and secured by means of studs (87).The lower turbine cover (58) itself istightly pressed into the recess in the hor-izontal flange of turbine housing (23) and

321

secured to it by means of bolts (59), theheads of which are in the secondary con-crete structure (24). The lower ring (60)is also secured to the lower turbine cover(58). This arrangement gives a fixed du-rable position to the second stage distri-butor. The movable upper band (54) hasslots through which the stationary guidevanes pass. In the second stage, the mov-able upper band (54) does not function asa valve since it is sufficient to have onevalve for the machine (the movable upperband (18) of the first stage distributor).Therefore the outer cylinder of the movableupper band of the second stage distributoris shorter than that of first stage and itgoes inside of the groove in the upper ring(52). The inner cylinder of the movableupper band (54) has similar design as thefirst stage and at the top position it goesinside the other groove of the upper ring(52). The second stage distributor hasalso four movable guide vanes (55). Thesemovable guide vanes (55) are uniformly dis-tributed between eight stationary guidevanes (86). The lower ring (60) has sever-al slots under each movable guide vane (55)which permit it to go down inside the ring.The movable guide vanes (55) are casted to-gether with upper (88) and lower (57) jour-nals. The upper journal (88) passesthrough the hole in the movable upper band(54) and a nut (89) presses the upper endface of the guide vane (55) to the movableupper band (54). The lower journal (57) isconnected to the second stage distributorservomotor rod (90) by means of connector(91). The lower journal (57) passesthrough the bronze bushing (62) in the low-er turbine cover (58) and the upper journal(88) is guided by the bushing (63) in theupper ring (52) of the second stage dis-tributor. There are four second stageservomotors (56) mounted to the lower tur-bine cover (58) by means of brackets (92).These servomotors (56) provide the verticalmotion of the movable upper band during theregulation period.

HYDRAULIC ANALYSIS

The flow rate and thus the power ofTPCFR can be regulated by moving the ad-justable hub and band along the axis of themachine (in case of a multi-stage machinethis has to be done simultaneously at eachstage). This can be shown in the turbinemode for a simultaneous movement of theadjustable hub and band as follows:

In our analysis

Wn

'11

H

i s the height of a guide vaneat any time,l s Vb0min'is the radial coordinate inthe cylindrical system,is the value of r for thepoint 0 of intersection of theoutlet edge of a guide vanewith an arbitrary streamlinein the water passage,is the value of r for thepoint 2 of intersection of theoutlet edge of a blade with anarbitrary streamline in thewater passage for a certainvalue of L_,is the length of a segment a-long the outlet edge of ablade connecting point 2 andthe point of intersection ofthe outlet edge with the low-er ring,is the tangential component ofthe absolute velocity,is the meridional component ofthe absolute velocity,is the radial component of theabsolute velocity,is the relative velocity,is the speed of the turbine(rpm),is the specific speed,

is the efficiency of the tur-bine,is the head of the turbine,

u=7r-n/3O is the angular velocity of therunner (1/sec),

U=wr is the corresponding velocity,g is the acceleration of free

fall,(3Q is the angle between the vec-

tor of absolute velocity V.and the tangential unitvector § Q at point 0,

By is the angle between the vec-tor of relative velocity $.and the tangential unitvector e 2 at point 2 and

E 2 i s the angle between the

streamline and a vector per-pendicular to the outlet edgeat point 2.

The flow rate according to Fig.3 is

Q=J 2-ir-r2-Vm2'cos(e2).dJl2. (1)

322

Vu0 = Vr0 / t a n (V

Fig. 3. Cross-sectional view of runnerand distributor.

Fig. 4 shows the development on theplane of cross-section of the outlet edgeof the blade by a surface of revolutionformed by an arbitrary streamline in themeridional flow. From Fig. 4 it can beseen tnat

(2)

where U,=wr . From (1) and (2) one ob-

tains

(7)

(8)

)2= 2-ir-

•tan(62)'d!>2,

and from Euler's equation one has

(3)

Fig. 4. Kinematics of flow at runneroutlet.

By combining equations 3, 4 and 8,one receives

where V ,'r, is the moment of absolute ve-ul 1

locity about the axis of the turbine beforethe inlet edge of a blade. In fact

V .r =V «r (5)

because hydraulic losses between the guidevanes and the inlet of the runner are neg-ligible and the flow is axisymmetrical.The flow in the distributor is uniform andfrom Fig. 3 it follows that

(6)

cos(e2)'tan(62)«dJ!,2)

(l+(l/bo-tan(eo)))-j cos(e2)

0

•tan(62)'d«,2) (9)

If the generalized mean value theorem3

is employed and both sides are divided by

D12-H°*S, Equation 9 becomes

and since

323

(l/((cos(e2)-tan(e2))(2)-L2))

+l/(bo-tan(3o))), (10)

-where r^r

-^ L

2= L2 / Di a n d

and superscripts (1) and (2) indicate themean values. The Equation 10 can be writ-ten as

/((l/((cos(e2)-tan(B2))(2)

• (l+a/bo))))+(l/tan(eo))),

where A=^2min~b0min" lt i s c l e a r f r o m t h l s

Equation that if X^O and n^is constant,

Q.1 increases with increasing bn, but at a

rate slower than that for the increase of

VGeometrical parameter A characterizes

the rate of deviation of the water passagewithin the runner from strictly radial toradial-axial. Only for a water passagepurely radial within the runner, Q.. is

proportional to b-, because in this case

5Omin=i:2min (see " } ' 3) and *2 and

(cos(e2)'tan(e_)) are two exact con-

stants.Clearly the construction of a radial

machine allows a design with b\, . =0, inUmin

which case Q.... can be changed from 0% to

100%. The formula for Q 1 for the pump

mode differs from Equation 11, because in

pump mode V I ' T ^ O and 2= bo"

in the pump mode

•(cos(e2)-tan(e2))(2)-b0

where and (cos(e2)-tan(6 ))(2) relate

to the outlet edge in the pump mode, i.e.,

the inlet edge in the turbine mode. There-

fore in the pump mode Q_. is always propor-

tional to b Q, when it changes from b _ .

to £„_ . From Fig. 3 it is clear that forUmax

a purely radial pump the design allows

bOmin=°-The analysis of the power losses, in

the runner, distributor, draft tube (suctionpipe), scroll casing, during the transitionfrom scroll casing to the distributor, andfriction losses at the upper and lowershrouds showed that TPCFR has almost flatcurves n=n(N/Nmax) for both turbine andpump modes in comparison with Neyrpic andconventional reversible machines (Figures5 and 6).

o W

100

90

80

70

60

SO

{{•ConstantTurbine Mode

10 20 30 40 SO 60 70 30 90 100

i is)

100

I

90

80

70

60

50

R'ConstantFurap Mode

0 10 20 30 40 SO M 70 80 90 100

K/Nnux (70

The o n e - s t a g e turbine-pump w i t h c o n t r o l l e dd i scharge by changing the h e i g h t of the runnerand s t a y r i n g (N-U3 MW. H-450 m, N-750 rpm,D-1.80 m>

Neyrpic one-stage turbine-puop vi th controlleddischarge by distributor (N-3S MU. «>438 m.K-750 rpm'. D-1.70 m)

Fig. 5. Forecast topograms for one-stageTPCFR.

The drastic decrease of efficiency inNeyrpic turbine-pump for the powers smallerthan N can be explained easily. The re-

gulation of power in this machine is pro-vided by changing the vane angle in the dis-tributor. However, in the pump mode theflow enters the distributor after passingthrough the runner. Consequently, thechange of vane angle cannot change the flowcharacter in the runner, but only increasesthe head losses in the distributor.

324

100

80

70

60

50

100

1 (T.)90

80

70

«0

H«ConstaneTurbine Mode

"it 51$ SO 7i3 St 94 ffl

H-ConstancPump Mode

50 Is—ib—2for-~3fr—str—5b—6tr- rrt—at!—stnN/Nmox (7.)

Two-atage turbine-pump with controlled discharge(N-96.MW, H-900 m, N»75O rpm. 0-1.80 n)

Conventional two-atage turbine-pump(N-96 MH. H-900 m, U-750 ryn, D-1.80 m)

Fig. 6. Forecast topograms for two-stage TPCFR.

RESULTS

The comparison of TPCFR with the con-ventional machine gives the following re-sults.

In the turbine mode, both the conven-tional turbine-pump and TPCFR have approx-imately the same, ability to regulate thepower with high efficiencies as can be seenin Fig. 5. In the pump mode, althoughTPCFR can effectively regulate the power,the conventional turbine-pump can only op-erate at one point with high efficiency orregulate the power only by drastically de-creasing the efficiency' (Neyrpic scheme) asseen in Fig. 5. The cavitation factor forTPCFR is approximately the same as that ofthe conventional machine since both areradial-axial machines. From the pulsationspoint of view, in the turbine mode, bothmachines will have the same pulsation le-

vels. In the pump mode, both machines dohave strong pulsations in the distributor,since the flow after the runner is unsteady.However in the conventional machine theguide vanes are not rigidly fixed, sincethe upper and lower journals of the guidevanes have to rotate freely in the bushingsmounted to the upper and lower rings.Therefore the pulsations in the distributorof the conventional machine causes strongvibrations. This is the inherited disad-vantage of the conventional turbine-pumpwhich drastically increases with the in-crease of head. In the turbine-pump withcontrolled flow rate, the vanes are fixedto the upper and lower rings (they arewelded to these rings). Consequently thepulsations of flow will not produce vibra-tions, since the vanes and the upper andlower rings form a heavy rigid, monolithstructure. The pulsations cannot producevibrations of the movable upper band either,since they will be damped by this ring.Consequently, from this point of view, theturbine-pump with controlled flow rate ismore reliable. The weight calculationsshowed that the weight of the distributorsof the conventional machine and TPCFR areapproximately the same, (the new machinedoes not have the gate ring with servomo-tors, but it has the upper movable bandwith servomotors). The runner with shaftfor the new one-stage machine is approxi-mately 70% heavier than that of the conven-tional machine. However, TPCFR does notrequire the operating valve which mightweigh as much as four times the weight ofthe runner of the conventional machine.Consequently the conventional one-stagemachine with the operating valve has to beheavier than the one-stage TPCFR. It isclear that the cost of labor for the pro-duction of TPCFR is higher, therefore it isdifficult at this stage to predict whichmachine will be more expensive. However,even if the cost of the new pump is accept-ed to be higher than that of the convention-al pump by 20%, it is easy to show that thenew pump is economically better than theconventional for storage of solar electricalplant energy, because of its ability to re-gulate the flow rate in the pump mode. Inorder to participate during peak time inthe utility grid with certain power, thesolar electrical plant with pumped storagehas to store a certain amount of energyduring sunny days.

From Fig. 7 it is clear that new pumpscan store energy of the solar plant contin-uously without waste, but the conventionalpumps working by steps inevitably waste en-

325

ergy. This waste causes the increase ofinstalled power in the solar electricalplant in order to accomplish the same re-quirements of energy.

? (MW)

S 9 10 11 i2 13 !'• 15 If. t fh>

a. Storage Plane vlth Tuo I'uapswltb Controlled Flow Fate

Output of Solar Electrical riant .

( 9 Iff 11 12 13/ 14 iy\ 16 t (h)

>tj .J tj tijb. ScoraRe Plant with Conventional

Punpa

Fig. 7. Pumped storage of solar elec-trical plant energy.

In the following comparison, theseassumptions were accepted.

The solar electrical plant has to worktwo hours during the peak time with power

of 1109 MW, the ratio of sunny days in thisregion is 80%, the efficiency of storage isTI =0.8. The typical graph of insolation

for the latitude 25°N was used for compar-ison (Fig. 7). The result of the calcula-tions is that the solar plant equipped withtwo new pumps in combination with threeconventional pumps (in order to have con-tinuous regulation, two of the new pumpswere sufficient) will have the installedpower of 500 MW. The solar plant equippedwith five conventional machines after opti-mization has to have an installed power of552 MW. If the cost of 1 kWe of solar e-lectrical plant is taken as $2000 and thecost of the conventional pump is $100/kWe,the cost of turbines is $100/kWe and thecost of construction is $100/kWe, the solarplant with storage equipped only with con-ventional machines will cost $1242.5/kWe.The cost of the solar plant with storageutilizing two new machines will be $1150.4/kWe. Consequently the scheme with new ma-chines- costs 8% less in this particularcase.

The analysis of feasibility and pros-pects leads to the following conclusions.

It is feasible to use the turbine-pumpwith controlled flow rate at the high headconventional pumped-storage plants in com-bination with conventional machines. Twoor three machines with controlled flow ratewill provide the pumped-storage plant withthe ability to regulate the power continu-ously and with high efficiency in both pumpand turbine modes.

The most attractive application of thenew machine is the utilization of pump withcontrolled flow rate in storage systems ofsolar-electrical plants, wind-electricalplants, etc., since this pump eliminatesthe energy losses which are inevitable dur-ing operation of these plants.

It is appropriate to begin the labora-tory experiments of the turbine-pump withcontrolled flow rate, since the hydraulicanalysis and conceptual design confirmedthe positive features of this machine.

There is an immediate demand of solarand wind electrical plant storage systemsfor rural areas of the U.S.A. In order tosatisfy this urgent demand it is possibleto start the technical design and produc-tion of small machines of this new type.The power of these machines should be nomore than 1 MW. and the diameters no more

326

than lm. These machines will successfullyoperate under a head around 100m.

ACKNOWLEDGEMENTS

The authors gratefully acknowledge thefinancial support of the Department ofEnergy and the continuous support and guid-ance of Dr. George C. Chang, Chief, Ad-vanced Physical Methods Branch, Divisionof Energy Storage Systems, D.O.E., duringthe accomplishment of the project.

The authors also express their grati-tudes to Dr. George Kartsounes, Dr. CarlBloomquist and Dr. Shiu-Wing Tarn of Ar-gonne National Laboratory for their contin-uous leadership and assistance during theproject and especially the preparation ofthe final report, and to Dr. Selim Chacourof Allis-Chalmers, Hydroturbine Division,for his valuable suggestions concerningthe design of the one and two-stage ma-chines.

REFERENCES

1. Radial-axial Hydraulic Turbine withDouble Governing, A. Gokhman, U.S.Patent No. 3 240 469, March 15, 1966.

2. Multi-Stage Turbine-Pump with Govern-ing, A. Gokhman, Patent ApplicationNo. 618 964, October 2, 1975.

3. Differential and Integral Calculus,R. Courant, Interscience Publishers,New York, 1964.

327

SESSION V H : COMPRESSED AIR ENERGY STORAGE

329

Project Title:

PROJECT SUMMARY

Compressed Air Energy Storage--Reservoir-StabilityCriteria for Temperature and Pressure Cycling

Principal Investigator: W. V. Loscutoff, J. A. Stottlemyre

Organization: Pacific Njrthwest LaboratoryPO Box 999Richland, WA 99352Telephone: (509) 946-2768

Project Objectives: Develop design and stability criteria for long-termoperation of storage reservoirs subjected to temperaturepressure and humidity fluctuations of a CAES olant.

Project Status: This project is divided into the fell owing four phases:

(1) Define preliminary design and stability criteriabased on a comprehensive state-of-the-art survey;

(2J Establish numerical models to study behavior ofreservoirs and obtain additional design and stabilitycriteria for reservoirs;

(3) Perform laboratory experiments to obtain supplementarystability criteria;

(4) Perform field tests to establish final design andstability criteria.

The project status is as follows:

(1) Preliminary design and stability criteria have beenformulated for compressed air energy storaae inaquifers and salt cavities. Criteria for Hard rockcaverns will be formulated by February 1979.

(.2) Numerical analyses of the behavior of acquifers, saltcavities and hard rock caverns are proceeding accordingto the program clan.

(3) Laboratory studies of aquifers are currently underway. A contract has been signed with Louisiana StateUniversity to perform laboratory studies of saltcavities. Proposals are being evaluated to select acontractor to perform laboratory studies of hard rocks-pectmens.

(4) A program plan for field studies of salt cavitiesis being formulated. Program plans for field studiesof aquifers and hard rock caverns are to be preoaredduring FY-1979.

Contract Number: EY-76-C-06-1830

Contract Period: FY-1978, continuing

Funding Level: $800,000 B0*

Funding Source: Department of Energy, Division of Energy Storage Systems

331

•Included in this project are the following:

L. W. Wiles, "Fluid Flow and Thermal Analysis for CAES in Porous RockReservoirs," page 337

J. R. Friley, T. J. Doherty, "Thermo Mechanical Stress Analysis of PorousRock Reservoirs," page 349

J. A. Stottlemyre and R. P. Smith, "Potential Air/Water/Rock Interactionsin a Porous Media CAES Reservoir," page 355

332

COMPRESSED AIR ENERGY STORAGE PROGRAM OVERVIEW

Walter V. LoscutoffPacific Northwest LaboratoryRichland, Washington 99352

ABSTRACT

The DOE compressed air energy storage (CAES) program consists of a group of inter-related studies directed at developing a new technology to improve the cost and efficiencyof electrical power utilization. The program has two major thursts — ReservoirStability Criteria Studies and Advanced Concepts Studies. The Reservoir StabilityCriteria Studies are aimed at accelerating the near-term application of conventionalsystems. The Advanced Concepts Studies are directed at development of systems that requirelittle or no firing of the turbines with, premium fuels. The Pacific Northwest Laboratory,as the lead laboratory in the CAES Program, assists DOE, Division of Energy Storage Systemsin program planning, budgeting, contracting, managing, and reporting.

In this overview, we summarize the CAES program, outline the specific tasks andidentify the performers, indicate progress on current tasks, point out anticipated activ-ities, and examine significant milestones.

INTRODUCTION

The Pacific Northwest Laboratory isthe lead laboratory for the Department ofEnergy's Compressed Air Energy Storage(CAES) Technology Program. As such, it isresponsible for assisting the Departmentof Energy in program responsibilitiesfor planning, budgeting, contracting,managing, reporting and disseminating in-formation.

The CAES Technology Program is a seriesof interrelated studies being performed bynational laboratories, universities, andindustry to improve the cost and efficiencyof electrical power utilization. It has twomajor thrusts — CAES Reservoir StabilityCriteria Studies and Advanced CAES ConceptsStudies.

The CAES Reservoir Stability CriteriaStudies are directed at accelerating near-term application of conventional CAESconcepts by the utilities industry. Itsmajor tasks are to establish design andstability criteria for reservoirs to beused by CAES plants. The CAES AdvancedConcepts Studies are directed at developingadvanced CAES systems for the future thatwill require little or no firing of theturbines with premium fuels. The effortsare concentrated upon CAES systems withthermal energy storage (CAES/TES) and onCAES systems with coal-fired fluidized bedcombustion (CAES/FBC). Other concepts

considered include coal gasification,liquefaction, MHD systems, solar-thermalaugmentation, and nuclear waste heat ofdecay.

In order to put the CAES TechnologyProgram into the proper perspective, letme briefly describe the overall DOE CAESprogram. There are two distinct programson the CAES concept within DOE as shownin Figure 1: the Demonstration Program,co-sponsored by EPRI and managed by threeutility companies, and the TechnologyProgram with PNL as the lead laboratory.The Demonstration Program seeks to estab-lish the feasibility and availability ofCAES to a utility, leading to a demonstra-tion plant. The Technology Programobjectives are to accelerate the commer-cialization of CAES and explore advance-ments that will make CAES totally indepen-dent of any premium fuels such as naturalgas and oil. It is responsive to thequestions raised by the DemonstrationProgram and works with the DemonstrationProgram to maximize the efforts of thetwo.

CAES TECHNOLOGY PROGRAM FOR FY-1978

At this meeting, we shall presentreports on progress in major tasks under-taken during FY-1978. Because of spaceand time limitations, several tasks willnot be presented here either because theywere relatively minor or because they have

333

DOE CAES

PROQRAM

EPRI CAES

PROGRAM

TECHNOLOGY

CAES RESERVOIR STABILITY PROGRAM OUTLINE

DEMONSTRATION

MSSPSI

PEPCO

Fig. 1. CAES Technology and DemonstrationPrograms

been initiated only a short while ago.The most important in the latter categoryare the tasks established to perform fieldstudies in a salt cavity, to study thetechnical and economic feasibility of coalfired fluidized bed systems integrated withCAES, and to perform laboratory studies ofsalt specimens under simulated CAES condi-tions. These tasks have been initiatedand we expect activity to begin duringFY-1979. The other major tasks will bepresented by the principals involved. Inthis overview, I would like to outline therelationship that exists between the varioustasks.

A. Reservoir Stability Criteria Studies

The objective of this program is toestablish design and stability criteriafor CAES reservoirs. The program outlinedin Figure 2 consists of four phases:

. state-of-the-art studies

. numerical modeling

. laboratory studies

. field studies

The three types of air storage reser-voirs being considered for CAES are:

. salt cavities

porous rock reservoirs

. hard rock caverns

The past year has seen a number ofmilestones reached in terms of accomplish-ments and new starts. These include thefollowing items:

Develop preliminary design andstability criteria for air storage

Fig. 2. CAES Reservoir Stability CriteriaStudies

in porous rock via state-of-the-artsurvey (PNL-Stottlemyre);

Develop preliminary design andstability criteria for air storagein salt cavities via state-of-the-artsurvey (LSU-Thoms and Martinez):

. Initiate study to develop preliminarydesign and stability criteria forair storage in hard rock caverns(RE/Spec-Gnirk)

. Start numerical studies of behaviorof salt cavities subject to CAESconditions (SGI-Serata);

. Perform numerical studies of thebehavior of porous rock reservoirs(PNL)

. Start numerical studies on behaviorof hard rock caverns subject to CAESconditions (RE/Spec-Gnirk);

. Perform laboratory studies on effectsof CAES conditions on porous rock(PNL and Wisconsin-Pincus)

. Start laboratory studies on effectsof CAES conditions on salt cavities(LSU-Thoms)

. Develop draft program plan for fieldtests in salt domes (PNL-McSpadden);

A summary of the foregoing items isshown in the milestone/accomplishment chartshown in Table 1 together with anticipated

334

Table 1. CAES Technology Program Mile-stones/Accomplishments

— — . Fiscal *ear

* . Reservoir Design and Stability Crttert*1. Preliminary design and stability

criteria (state-of-the-art-survey)a. Aquifersb. Salt cavitiesc. Hard rock caverns

2. NuMrical modeling». Aquifersb. Salt cavitiesc. Hard rock caverns

3. Laboratory studies». fajuifersb. Salt cavitiesc. Hard rock caverns

4. Full scale field testinga. Aquifersb. Salt cavitiese. Hard rock caverns

© complete test program plan<Z> init iate testing program

5. Aquifer field optimization6. Advanced CAES Concepts Studies

1. Plufdiicd Bed Combustion (CAES/FBC1$ Initiate feasibility study© Initiate demonstration program

2. Adiabatic Systems (CAES/T£S)© Initiate demonstration orouram

3. Equipment Evaluation and Oeveloownta. Turbinesb. Compressorsc. 1C expander/compressor engine

4. Coal 6is1f1tltton

5. Evaluation o'f Hex Conceptsa. Solar augmentation of thermal bedsb. Coal-driven HH;c. Nuclear waste heat

1978

n

.

tripas

s

•7

A

7

1980

aVB

a

0

7

completion dates where applicable.

Additional major action items in theReservoir Stability Studies include thefollowing:

. Start laboratory studies of hardrock specimens subject to CAES (RFP);

. Develop program plans for field testsof porous rock reservoirs and hardrock caverns;

. Initiate field testing activities;

The ultimate objective of this por-tion of the program and the final mile-stone is to establish design and stabilitycriteria for reservoirs used for com-pressed air storage. This is to beaccomplished by 1981.

B. Advanced Concepts Studies

The objective of this portion of theprogram is to eliminate the dependence ofCAES plants on either natural gas or oil.Advanced CAES plants could eliminate theneed for premium fuels. Plants currentlybeing considered for demonstration discardthe heat energy of compression. Thermalenergy storage concepts are being studiedwhich would extract and store the heat ofcompression from air pumped into the air

reservoir. This heat would then be re-turned to the air before it is expandedin a turbine, thus eliminating any needfor fuels. Another approach is to lookfor other fuels to provide the requiredheating of the air. One attractive con-cept is coal-fired fluidized bed combus-tion integrated with a CAES plant. Thesevarious options are outlined in Figure 3.

Fig. 3. Advanced CAES Concepts Studies

During the past year, a number ofactivities have been initiated leadingto several important milestones. Themilestones include the following:

. Complete evaluation of advancedturbomachinery for CAES (ANL-Kart-sounes);

Start technical and economic feasi-bility study of CAES/FBC systems(UTRC-Giramonti);

. Start solar augmentation study ofCAES/TES systems (MIT-Flynn);

. Start feasibility study of CAES/TESsystems (PNL-McKinnon, MIT-Flynn);

Start design of internal combustionexpander/compressor engine (ANL-Kartsounes);

. Examine MHD/CAES and nuclear waste/CAES concepts (PNL-McKinnon);

. Perform numerical study on optimaldesign of aquifer reservoirs (ANL-Ahrens).

A summary of the foregoing issues isshown in the milestone/accomplishmentchart shown in Table 1 together withanticipated completion dates when applicable.

135

Future major activities in the AdvancedConcepts Studies project include thefollowing:

. Initiate technical and economicfeasibility analysis of advanced CAESfor a utility;

. Complete evaluation of technical andeconomical feasibility of advancedCAES equipment;

. Initiate demonstration program foradvanced CAES.

RELATED CAES ACTIVITIES

In addition to the DOE activitiesdiscussed above, there are several otheron-going projects in CAES. A brief summaryof the major projects is given below.

. The CAES plant at Huntorf was commis-sioned in August, 1978, and commercialoperation of the plant began in October.Since the basic features of this plant,owned by Northwest deutsche Kraftwerke ofWest Germany,, have been described else-where, only highlights of the plant aregiven here. This is a 290 MW peakingplant with a two hour discharge and aneight hour charge time. The air isstored in two salt cavities havinga total volume of 300,000 cubic meterswith pressures up to 70 bars. Theheat rate of this plant is 5500 Btu/KWh. Plant construction began inMay 1975.

. The Kansas Utility Group, consistingof six private utilities, has com-pleted a statement of work to EPRIfor a Phase-2 study that will look atthe economics of CAES in Kansas. Thestudy will examine a detailed loaddistribution model of a compositesystem for the utilities and evaluatethe available options that includecombustion turbines, small coal-firedunits, and CAES. This study followsthe Phase-1 study, performed byBlack and Veatci . which was a geo-logic assessment of CAES in Kansas.If CAES is found to be economicallyfeasible, Phase-2 is expected to befollowed with preliminary CAES plantdesign.

. The International Research andTechnology Corporation, assisted byDames and Moore, is performing astudy, under a contract with the

Department of Energy, on risk insur-ance for CAES storage reservoir degra-dations.

A joint study by the Central Elec-tricity Generating Board (CEGB) ofEngland and EPRI is looking at meansof improving the performance of com-pressed air storage schemes by theuse of thermal energy storage (TES).Two approaches are being examined.In one approach, adiabatic storage,the TES is used to eliminate anyneed for fuel. In the other approachhybrid systems, TES is used to reducethe premium fuel consumption by theturbines. This study is more compre-hensive arid complements the studiesfunded by the Department of Energy.

336

FLUID FLOW AND THERMAL ANALYSIS FOR CAES IN POROUS ROCK RESERVOIRS

L. E. Wi lesCo-Author C. A. Os te r

Pacif ic Northwest LaboratoryRichland, Washington 99352

ABSTRACT

The analysis described in this report is a continuation of work initiated at PNL inFY-1977. A computer code was developed in FY-1977 to define the hydrodynamic and thermo-dynamic response to simulated mass cycling in a CAES dry porous media reservoir. Thecode was based on one-dimensional radial transport of mass and energy. The capabilityof the code has been extended in FY-1978 to include a one-dimensional, radial flowanalysis of the effects of vapor and liquid phase water on reservoir performance.Parametric analysis included consideration of a range of injection temperatures, injec-tion humidities, and residual water levels. Potential consequences of the presence ofwater related to deliverability, thermal energy recovery, working fluid recovery, andstorage volume are evaluated. Also, two-dimensional modelling has been developed for dryreservoirs to include the effects of gravity, vertical heat losses and the effects ofstratified permeability.

INTRODUCTION

The analysis of the hydrodynamic andthermodynamic response to mass cycling inCAES porous rock reservoirs is intended toprovide design guidelines for the efficientand stable operation of the air storagereservoir. The performance of the reser-voir depends on reservoir material proper-ties, reservoir geometry, and operatingconditions. The influence of many of theimportant parameters was investigated atPNL in FY-1977. That analysis made useof a computer cede based on one-dimensional,radial transport of mass and energy in adry porous rock reservoir.1

The capability of the code was extendedin FY-1978 to permit the analysis of theeffects of water on reservoir performance.Potential problems or areas of concern re-lated to the presence of water in thereservoir that could be evaluated with aone-dimensional model were identified.These include deliverability, thermalenergy recovery, working fluid recoveryand storage volume. It was the objectiveof this study to quantify the magnitudesof these effects. This was done by evalu-ating the influence of injection tempera-ture, injection humidity and residual watercontent on reservoir performance.

While the performance of the bulk ofthe reservoir and the influence of indivi-dual parameters can be reasonably charac-

terized by the one-dimensional modelling,the delineation of the influences of thevertical boundaries, potential stratifica-tion of permeability, and gravity effectsrequires two-dimensional modelling. Codedevelopment and analysis to this end wasalso done at PNL in FY-1978.

Summaries of these two efforts are pre-sented in this paper, which is divided intotwo sections. The first section deals withwater in the reservoir. The second sectioncovers the development of the two dimen-sional code.

I. THE EFFECTS OF WATER ON RESERVOIRPERFORMANCE

THEORETICAL DEVELOPMENT

BASIC ASSUMPTIONS

In general, the hydrodynamic andthermodynamic behavior of a CAES porousrock reservoir can be characterized bya set of conservation equations thatdescribe the flow of an air-vapor mixtureand liquid water through a rigid porousmaterial. A number of assumptions mustbe made such that the resultant set ofequations does not include terms irrele-vant to CAES in porous rock reservoirs.These assumptions are:

. the air-vapor mixture is homogeneousand behaves as an ideal gas

337 -

. the rock porosity, density, and heatcapacity are constant;

. the rock is immobile;

. binary diffusion in the air-vapormixture is negligible;

. inertial effects are negligible;

. the air-vapor mixture, liquid, androck are in thermal equilibrium;

. kinetic energy is negligible; and

. viscous dissipation of energy isnegligible.

REDUCED EQUATIONS

Additional assumptions were made tofacilitate this initial investigation ofthe effects of water on reservoir behavior.These include:

. the transport of mass and energyoccurs only in the radial, horizontaldirection;

. the liquid water is immobile; and

. Darcy's law can be used to write thefluid velocity in terms of the pres-sure gradient to eliminate themomentum equation.

With these assumptions the governingequations reduce to the following:

. Conservation of mass in the reservoir

Conservation of energy in thereservoir

. Conservation of mass for the vaporcomponent

•33t vf r

3P» .(2)

. Conservation of mass for the liquidcomponent

SP ) = -in1; and (3)

A definition of nomenclature is given inTable 1.

Table 1. Nomenclature

c<ih

k

K

m1

M

P

r

S

t

T

u

z

heat capacitygravitational accelerationenthalpyeffective permeabilitythermal conductivityevaporation ratemass of working f luid 1n reservoirf luid pressureradial coordinatewater content, fraction of pore volume

timetemperatureinternal energyvertical coordinate

GREEK

»

L

U

porositydensityviscositypercent of mass cycles

SUBSCRIPTS

a

s.

m

r

sV

z

a i r

liquid waterair-vapor mixtureradial coordinotesolidwater vaporvertical coordinate

BOUNDARY CONDITIONS

The reservoir geometry adapted tothis problem is shown in Fig. 1 where theregion of influence of a single well isapproximated as a cylindrical disc.

At the well boundary the mass flowrate is continuously specified. Duringthe reservoir charging cycle the temper-ature and humidity are specified. Duringreservoir discharge, or when the reservoiris closed, the temperature and humidityat the well are obtained from localequilibrium conditions.

338

RADIAL FLOWIN POROUS MEDIA

OUTERBOUNDARY

Table 2. Reservoir Parameters

Fig.

SINGLE WELL RESERVOIRDIAMETER

1. Geometry of the 1-D ReservoirModel

The symmetrical arrangement ofadjacent wells, which would be chargedand discharged at about equal rates, sug-gests that the outer boundary of thesingle well reservoir be insulated tothe transport of mass and energy.

SOLUTION

A detailed outline of the solutionmethod of the governing equations isprovided in the FY-1978 Progress Reportto be published by PNL.

ANALYSIS OF THE MOISTURE PARAMETERS

The parameters that were analyzed withrespect to dehydration and thermal develop-ment in the air storage zone are:

. injection air temperature;

. injection air humidity; and

. residual water content.

The foundation for the analysis ofthese parameters is a set of referencevalues for the reservoir geometry, materialproperties and operating conditions. Thevalues given to these parameters are shownin Table 2. These values were fixed through-out this analysis.

The reference values and the range ofvalues investigated for the moisture para-meters is given in Table 3.

The analysis was based on a reservoirhaving unit vertical thickness. Thus,wherever appropriate, the specification of

Parameter Reference Value

GeometryReservoir DiameterNell Diameter

PropertiesPorosityPermeability

Rock Thermal Conductivity

Operating ConditionsNominal PressureMass Flow Rate

Injection

Extraction

Initial Reservoir Temperature

400 f t (122 m)

7 In (18 cm)

20*

500 md

50 atm (S070 kPa)

" • " ' s e C T t < ° 1

<>•»<> I e £ t

!00°F (38°C)

Table 3. Moisture Parameters

Reference Value Range of Values

Injection Temperature

Residual Water as Percent

of Pore Volume

Injection Humidity

450vF(232°C)205

0 (dry air)

100-450(38-232)0-40

dry air - saturated

air at the Injection

temperature andpressure8

The Injection humidity is limited by the maximum water vipor available Inatmospheric air. This was considered to be 0.0429, the units being mass ofwater per unit mass of dry air; i .e . , the saturation condition of lOOOF (38°C)air at atmospheric pressure. This limit applied to injection temperaturesaboie 276°F (I36°C).

per foot (per meter) is used to imply theunit of vertical measurement.

A complete specification of operatingconditions includes a schedule for theinjection and extraction of mass to andfrom the reservoir. To develop thisschedule the following conditions wereapplied to the reference reservoir:

. 40% of the cycled mass is stored inthe reservoir over the weekend;

. the weekdays are characterized byalternating injection and extrac-tion periods of 10-hours each,separated by 2-hour intervalswhen the reservoir is closed;

. the time averaged reservoirpressure is the discovery pres-sure; and

. the percent of mass that iscycled is 14.8%.

339

The reservoir pressure is the spatialaverage pressure. The percent of massthat is cycled is calculated from

M

A _ max (5)

For these conditions, the weekly variationof reservoir pressure for the referencereservoir is shown in Fig. 2. The discoverypressure was assumed to be 735 psi (5070kPa).

mmIMitlU

mm ;mmSul

j| MIDWAY |

/

/

/

SIMMY | i

\ »m

AV\v 1m

BHDAY | IIESMr

l\l\f\«sa, \/ \

u \M M \

|MgwsMr|iMMSMr 1 am*

sultant condensation. Thus, a dehydratedregion surrounding the well will grow duringmass injection. During reservoir dischargethe mixture will gain heat as it nearsthe well, thus, gaining moisture as itmoves radially inward. Net dehydration ofthe reservoir occurs by the eventual extrac-tion of this moisture.

The process of dehydration for variousradial locations is shown in Fig. 3. Themodel predicts that the percent of porevolume filled with liquid water, which isinitially uniform at 20%, never exceeds 25%.This result for the reference reservoirsatisfies the assumption of zero liquidmobility which required that the watercontent should never greatly exceed theresidual level. More importantly, themodel suggests that for dry air injection,pore plugging should not adversely affectthe deliverability.

25

1IK0FMDC.** >

Fig. 2. Weekly Cycle of Reservoir Pressure, £Beginning Friday at 8:00 p.m. *

15 -

Changes to the moisture parameters willaffect the dehydration rates and the rateof thermal development of the reservoir.However, over the range of the values con-sidered the characteristic behavior of theprocesses of dehydration and thermal develop-ment are identical. Thus, the descriptionof these processes, as interpretted fromthe results obtained from the computer model,are presented in this section. Thesedescriptions will be applicable, in general,to all of the results of the parametricstudy of the moisture parameters.

Dehydration

When heated dry air is injected intothe reservoir it will exchange heat withthe surrounding material by sensible heat-ing of the rock and liquid water and byevaporating some of the water. The endresult is that the air-vapor mixture willbe saturated and will be in thermal equili-brium with the rock and liquid water. Whenthe air-vapor mixture moves radially out-ward it will continue to loose heat to thesurroundings. The mixture will maintain astate of saturation by virtue of the re-

A

--1

I2&.5fl-RADIUS

1(8.69 ml

3&2-ft-RA0IUS(1L6 ID)

i1 1

\ 47.frfl-RADlUS

\

i \

10 -

5 -

0 1 2 3 4 5

TIME, yrs

Fig. 3. Dehydration for Three Locations(5-Yr History)

Evaporation will occur near the welluntil complete dehydration of this regionis achieved. Gradually, the dry regionexpands. The model implies that a sharpinterface exist between the dehydratedzone and the region still containingliquid water. The "dry front" is a termused to define the location of this hypo-thetical interface. When the water contentreaches zero for a given radial locationthe dry front is defined to exist at thatlocation. By this definition the growthof the dry front is shown in Fig. 4.

340

r. MOPMn H

TIKOFKSK. On

Fig. 4. Radius of the Dry Front for theReference Reservoir

By the process of dehydration, watermay be removed from the well bore regionalthough it is not necessarily removedfrom the reservoir. After 5 years ofcontinuous reservoir operation the modelshows that the total water removed fromthe reservoir represents only-5.5% of theoriginal mass of water. The net increasein reservoir storage volume is a mere 1.1%.Thus, the dehydration process is quiteslow and should not, in general, be expect-ed to significantly increase the potentialstorage volume.

Thermal Cycling and Thermal Growth

When the simulation of reservoiroperation begins the reservoir is at auniform temperature of 100°F (38°C). Airis injected into the reservoir at 450 F(232 C) and is extracted at the equilibriumtemperature adjacent to the well. Figure5 shows the temperature variationsoccurring in the near wellbore region duringthe first week of reservoir operation.While large temperature swings occur at thewell boundary, thermal cycling is observedto be nearly non-existent within 14.0 ft(4.28 m) of the well center. The volumetricthermal capacity of the rock mass is largeenough to contain the injected thermalenergy very near the well. Net heating ofthe reservoir beyond 14.0 ft (4.28 m)occurs primarily by conduction.

The development of thermal cyclingpredicted by the model, is shown in Fig. 6,where the maximum and minimum weekly extrac-tion temperatures are plotted as functionsof time for a dry reservoir and for thereference reservoir. A significant reduc-tion in thermal cycling at the well occursin the first year of operation. Beyond oneyear the changes occur more gradually.

Fig. 5. Thermal Cycling During the FirstWeek of Reservoir Operation

500(260)

a-sr

4oo

AMAXIMUM WEEKLY

EXTRACTION TEMPERATURE

MINIMUM WEEKLYEXTRACTION TEMPERATURE

RESIDUAL WATER,

PERCENT OF PORE VOLUME

0 *

20%

200(93)

2 3

TIME, yrs

Fig. 6. Effect of Thermal Growth on Extrac-tion Temperatures

By examination of Fig. 6 it can beconcluded that a significant percentage ofthermal energy injected into a reservoircan be recovered during reservoir discharge.The average thermal energy recovery overthe first 5 years of operation of the refer-ence reservoir is about 81%. The dashedlines in Fig. 6 represent the result for adry reservoir. The average thermal energyrecovery for the dry reservoir is about82%. This comparison suggests that thepresence of residual water will have onlya small effect on temperature cycling aridthermal growth.

341

INJECTION TEMPERATURE

This portion of the analysis estab-lishes tha difference in the dehydrationrates associated with the spectrum ofpossible injection temperatures. Also,the coexistence of high temperature andliquid water can dramatically increasethe potential for adverse geochemicalreactions. Thus, it is important toquantify when such conditions might occur.A reasonable lower limit of injectiontemperature was considered to be 100°F(38OC). An upper limit of 450°F (232°C)was chosen as this represents the approxi-mate maximum outlet temperature that canbe tolerated by currently available cen-trifugal compressors.

In Fig. 7 the radius to the dryfront is plotted as a function of timefor each injection temperature.

INJECTION TEMPERATURES, °F(°C»

Fig. 7.

20 30

TIME, wks

Effect of Injection Temperatureon Growth of the Dry Front

In Fig. 8 the rates of net reservoirdehydration are shown. A time dependentdecrease in the dehydration rate, which ismost apparent for high temperature injec-tion, characterizes the results. Duringinitial operation the dehydration in thenear well region is influenced by theperiodically high temperatures presentdue to thermal cycling. When the dry

INJECTI ON TEMPERATURES, °F (°C)

100(381

10 20 30

TIME, wks50

Fig. 8. Effect of Injection Temperatureon the Net Reservoir DehydrationRate

front moves beyond the region of thermalcycling the dehydration rates slow as theheat necessary for evaporation is avail-able primarily by conduction. Anothercause of the decrease in dehydration ratewith time is that an increasing portionof the injected air-vapor mixture isstored within the expanding dry region.That portion of the air is subsequentlyextracted without having encountered anyliquid water.

INJECTION HUMIDITY

Unless it is otherwise defined thereference to humidity in this report isabsolute humidity where the units are massof water vapor per unit mass of dry air.

The moisture content of the injectedair is limited by the humidity of theambient air that is to be compressed.Saturated air at 100°F (38°C) and 1 atm(101 kPa) of pressure will have an ab-solute humidity of 0.0429. For a pressureof 50 atm (5070 kPa) an absolute humidityof 0.0429 can exist only if the tempera-ture of the mixture is above 276°F (136°C).Below this temperature the humidity of theinjected air-vapor mixture cannot exceedthe saturation conditions dictated by

342

the injection temperature and the reser-voir pressure. This is shown in Fig.9.

0.06

£ 0.04 -

in 0.02 -

LIMIT OF 0.0429,SATURATED AIR AT1 ATM (101 kPa) AND100°F(38°C)

100(38)

200 300(93) (149)

TEMPERATURE,

4001204)

°F(°C)

500(260)

Fig. 9. Saturated Absolute Humidity at50 atm (5070 kPa)

The effect of injected moisture onthe dehydration at a radius of 14.0 ft(4.27 m) is shown in Fig. 10. The dehy-dration is obviously slowed by the inclu-sion of humidity in the injected air stream.Also, local pore volume water content cansignificantly exceed the residual levels,which indicates that pore plugging may bea potential problem.

60 -

1 40 -

20\ 0.0V

l \

INJECTIONHUMIDITY

1 I

\

0.0429 \

1 \10 20 30

TIME, wks

40 50

Fig. 10. Effect of Injection Humidity onthe Dehydration of the RegionCentered at a Radius of 14.0 ft(4.27 m)

Reservoir dehydration rates are shownin Fig. 11. Although the negative dehydra-

5

ifif

M

RIS

ER

! It

Fig

200(298)

0

-200(-298)

-400(-595)

C

. 11

-___________^^

DRY AIR INJECTION

_ ^ _ _ _ _ _ _ _

^ INJECTION HUMIDITY - 0.0429

{i i i I 1

10 20 30 40 50

TIME, weeks

Effect of Injection Humidity onDehydration

tion rates shown in Fig. 11 suggest thatthe injection of moist air into the reser-voir can result in the increase of themass of liquid water in the reservoir, theregion near the well bore can still bedehydrated. The progression of the dryfront with time is shown in Fig. 12.Although the dry fronts appear to moveoutward rapidly, conservative reservoirdesign will likely dictate that considerableeffort will be given to minimizing theinjection humidity.

Fig. 12.

50

Effect of Injection Humidity onGrowth of the Dry Front

The results obtained from the modelsuggest that for injection humiditiesabove those for which the net dehydra-tion is negative, the temperature atthe dry front is characterized by the dewpoint temperature of the injected air-vapor mixtu1"5. This is an important ob-servation because, if the existence ofliquid water at this temperaturewas known to produce adverse geochemicalreactions, then some dehydration ofthe injected air stream would benecessary.

343

The effect of injected moistureon thermal growth in the reservoir isshown in Fig. 13. When dry air entersthe reservoir the subsequent evaporationwill inhibit the thermal growth. Incontrast, the injection of moist airresults in a net increase of liquidwater in the reservoir due to conden-sation, and the thermal growth will beenhanced because of the deposition oflatent heat.

•SCKVOIItllUIIJS.lt M

Fig. 13. Effect of Injection Humidity onThermal Growth

RESIDUAL WATER CONTENT

For reasonable values of residualwater content the thermal capacity of themass of water is not so large that it cansignificantly affect the thermal behaviorand dehydration in the reservoir. Itsgreatest impact would be to reduce thestorage volume. To obtain given mass flowrates to a given size reservoir, a greaterpercent of pore volume initially filledwith water will result in proportionallyhigher pressure cycling.

The primary effect of residual wateron the dehydration can be deduced by con-sidering Figs. 14 and 15. Although thethermal development occurs at nearly iden-tical rates, the reservoir having moreresidual water has a higher dehydrationrate. Because the dry front moves radiallyoutward at a slower pace for the case hav-ing more residual water, the dry front isin a region of the reservoir where thetemperature is higher. Since the tempera-ture at the dry front is greater, theextracted humidity will be greater. Thus,the net reservoir dehydration rate isgreater even though the temperature of theextracted mixtures would be about the same.

RESIDUAL WATER,PERCENT Of PORE VOLUME

50

io 20 30

TIME, wfcs40 50

Fig. 14.

30

Effect of Residual Water onDehydration Rate

eIU

S.RA

DI

(9)

20(61

10(31

0

RESIDUAL WATER,PERCENT OF PORE V O L U M E ^ ^ —

20* ^ ^ - " ^ ^ .

A'/

1 1 1

•

- — •

. — — ~ ~ '

1 1

20 30TIME, wks

40 50

Fig. 15. Effect of Residual Water onGrowth of the Dry Front

CONCLUSIONS AND RECOMMENDATIONS FOR FURTHERSTUDY

The results of the computer analysisprovide insight as to how the presence ofresidual water in an aquifer CAES reservoirmay affect the performance of the reservoir.More importantly, perhaps, the results canbe used to suggest how the presence of theresidual water will affect the areas ofconcern outlined in the introduction.

Deliverability. It appears that high tem-perature injection may not be necessary inorder to avoid reduced deliverability dueto reduced relative permeability in thevicinity of the wellbore. For a broadrange of conditions the computer modelpredicts that comparatively rapid dehy-dration of the wellbore region is achieved.

Injection humidities may proveto be very important with regard to theintegrity of reservoir deliverability.Calculations show that pore plugging can

344

result. Although a liquid mobilitymodel was not included in the analysisthe potential plugging problem is prob-ably real. Conservative reservoir designwould, therefore, dictate that relativelydry air be injected to the reservoir todecrease the potential for pore plugging.

Thermal Energy Recovery. In general, thethermal capacity of the residual liquidwater in an aquifer CAES reservoir shouldbe small comparad to the total thermalcapacity of the rock in the reservoir.Thus, variations in the residual water con-tent will not have a large effect upon theoverall thermal development of the reservoir.Thermal energy recovery will be affectedby the humidity of the injected air stream,although the effect will be small.

Working Fluid Recovery. The injection ofdry air resulting in net evaporation willincrease the working fluid available inthe reservoir. Net condensation of mois-ture injected into the reservoir resultsin a loss in the amount of working fluid.Thus, it may be necessary to weigh theimportance of thermal energy recoveryagainst working fluid recovery. However,it is suggested by the model that there isrelatively little to be gained or lost inthe balance between the two. It wouldseem to be more important to consider thepotential pore plugging problem.

Storage Volume. The presence of residualwater in the pore structure of the reser-voir rock reduces the potential storagevolume. The process of dehydration shouldnot, in general, be expected to signifi-cantly improve this situation. In fact,if the injected air is not relatively dry,the deposition of moisture in the reservoircould reduce the storage volume.

Further Study. These conclusions regardingthe potential problems associated withresidual water are subject to the assumptionsmade in the numerical model. The two mostlimiting assumptions of the model are zeroliquid mobility and instantaneous saturationof the air-vapor mixture. Adjustments ofthe governing equations can be made toinclude liquid mobility. Improvementsto the prediction of the rate of evapora-tion will require inputs from an experi-mental program scheduled for FY-1979.

II. TWO-DIMENSIONAL ANALYSIS OF DRYRESERVOIRS

THEORETICAL DEVELOPMENT

BASIC ASSUMPTIONS

The basic assumptions applicableto this problem are given in Section Iunder the same heading.

REDUCED EQUATIONS

The additional assumptions applied tothis problem include:

. the transport of mass and energy occursin the radial and vertical directions;

. the reservoir and the air are dry;and

. Darcy's law applies.

With these assumptions the governingequations reduce to the following:

Conservation of mass

kr

(6)

Conservation of energy

s s

1 3 / uF 3? (paha

, 1 3r zr

••- 3 r ' 3z PaV"

3?(7)

BOUNDARY CONDITIONS

The geometry adapted to this two-dimensional analysis, is shown in Fig. 16.The radial boundary conditions are similarto those specified for the one-dimensionalanalysis. At the well boundary, however,pressures are specified that were computedby the one-dimensional dry model of FY-1977to provide approximately constant massflow rates.

The boundaries between the porouszone and the caprock and basement rock areclosed to mass transfer. Heat is lost tothese regions by conduction.

SOLUTION

Details of the solution method are

345

GEOMETRY

3 ^ OVER- = £ 3

WELLBORE DIAMETER

CAPROCK

POROUSROCK

BASEMENT ROCK

-RESERVOIR DIAMETER'

Fig. 16. Geometry of the 2-D ReservoirModel

available in the FY-1978 Progress Reportto be published by PNL.

ANALYSIS OF VERTICAL EFFECTS

The intent of this analysis is toquantify the influence of vertical bound-aries, stratification of permeability, andgravity on the performance of a dryreservoir.

The reference conditions given inTable 2 were used in this analysis.Those conditions apply to the storagezone. Additional parameters are givenin Table 4. Material properties such aspermeability and thermal conductivity wereassumed to be isotropic. The weeklycharge and discharge cycle as described inSection 1, was applied. The overburdenwas assumed to be a thermal insulator.The geothermal gradient was taken as0.02°F/ft (0.128C/m).

ANALYSIS OF THE REFERENCE RESERVOIR

The analysis of the reference reser-voir permits an evaluation of the influenceof the vertical boundaries and gravity onthe performance of the reservoir. Thisanalysis can also be used to judge theusefulness of the one-dimensional modeling.

Temperature profiles are shown inFig. 17 for a homobeneous, dry reservoir.This result was obtained after completionof the final charging cycle of the first

Table 4. Additional Reference ConditionsApplied to the 2-0 Analysis

Parameter

Geometry

Caprock Thickness

Storage Zone Thickness

Basement Rock Thickness

Properties (Basement Rock and Caprock)

Permeability

Porosity

Operating Conditions

Inject ion Temperature

Well Pressures

Reference Value

37.5 f t {11 m)

]00 ft (30 m)

50 f t (15 m)

0

0

450°F (232°C)

Computed by code developedin FY-1977 to approximateflow rates given in Table 1

S

5.38-H- RADIUS

\

i ]}-lt RADIUS

\

l.OP-ll- RADIUS

1 B0-I1-RADIUS

30011491

ItMPlR'T^t. "F I°CI

400TOI

Fig. 17.Temperature Profiles in theHomogeneous Reservoir

week of simulated reservoir operation.The temperature profiles indicate thatvery little thermal energy is lost to thecaprock or basement rock. Since there isno convection in the caprock or basementrock then the model dictates that thermalenergy transport depends entirely on con-duction, which is extremely slow. Theresult also indicates that buoyancy effectsare not significant in the homogeneousreservoir. Heating of the rock occurssomewhat uniformly in the radial direction.Thus, there are no significant verticaltemperature differentials to create anoticable buoyancy component.

The thermal development and pressurecycling in the bulk of the reservoir agreeswith that predicted by the one-dimensionalmodel. Thus, the one-dimensional model canbe used to characterize the performance ofthe bulk of a homogeneous reservoir withequal accuracy but at a much lower cost.

346

STRATIFIED PERMEABILITY

A horizontal layer (i.e., a cylindri-cal disc) having a permeability of 50 mdwas defined to exist between zones havingpermeabilities of 500 md. The layer was15 ft (4.6 m) thick and was centered half-way between the caprock and basement rock.

In Fig. 18 vertical temperature pro-files are shown after completion of thefinal charging cycle of the first week ofsimulated reservoir operation. The resultsshow that thermal development of the lowpermeability zone is significantly dimin-ished. However, comparison of Figs. 17and 18 show that thermal development ofthe 500 md zones, and the caprock and base-ment rock is not noticeably affected bythe presence of the low permeability layer.

V V.i.oen-RADius

\

5.»H»«OIUS^J ).1)HR»OIOS J I. W-n RADIUS-.

J 1 )

tures will follow successful completionof that effort.

In addition, temperature and pressureprofiles computed by the two-dimensionalcode are necessary for evaluation of thepotential for failure of reservoir inte-grity due to thermo-mechanical stressbehavior. Thus, data from the code willbe used as input to the stress analysis.

REFERENCES

1. G.C. Smith, J.A. Stottlemyre, L.E.Wiles, W.V. Loscutoff, H.J. Pincus,FY-1977 Progress Report: Stabilityand Design Criteria Studies for Com-pressed Air Energy Storage Reservoirs.PNL-2443, Battelle, Pacific North-west Laboratory, Richland, WA 99352,March 1978.

Fig. 18. Temperature Profiles tn theStratified Reservoir

CONCLUSIONS AND RECOMMENDATIONS FOR FURTHERSTUDY

A two-dimensional model was developedto analyze the performance of dry, nonhomo-geneous reservoirs. Initial analyses wereperformed to evaluate vertical heat lossesand gravitational effects, both of whfchappear to be small. Also, the results showthat the one-dimensional model developed atPNL in FY-1977 is an equally valid tool fordescribing the performance of the bulk ofa homogeneous reservoir.

The capability to analyze nonhomo-geneous reservoirs and other verticaltransport effects has been demonstrated.Continued code development is planned forFY-1979 to improve the efficiency of thecalculations. A parametric evaluation ofpossible nonhomogeneous reservoir struc-

347

THERMO MECHANICAL STRESS ANALYSIS OF POROUS ROCK RESERVOIRS

J.R. FrileyT.J. Doherty

Pacific Northwest LaboratoryPO Box 999

Richland, Washington 99352

ABSTRACT

Stress levels in a generic porous rock compressed air energy storage site wereinvestigated. Site loadings considered thermal, pore pressure, and in situ effects.Loading conditions simulated site operation for approximately one year of reservoircharging and discharging. Structural concerns focused primarily on tensile stress andfatigue stress magnitudes. In addition, the reservoir structural response due totensile fracture of the cap rock region was investigated.

For the conditions studied, fatigue stress levels were observed to be greatest inthe porous material. These stresses decreased with age of reservoir operation and withdistance from the well bore. Tensile stresses were observed to be greatest in thecap rock. While the tensile stress magnitude will most likely depend on site dependentoverburden characteristics, mechanisms thought to cause this tensile behavior werepostulated.

While lack of site dependent information precluded interpretation of localizedfatigue and fracture stress levels, it is felt that several salient features of caprock and porous rock structural behavior in a porous rock reservoir were illustrated.

INTRODUCTION

Structural behavior of a porous rockreservoir subjected to loading conditionsof compressed air energy storage operationat elevated temperatures is, as yet, notfully known. No such operational facilityexists today. In addition, scale modeltesting of such a concept has not beencarried out at this time. In order togain some insight regarding these unknownsand perhaps to aid in performing testing ina more enlightened fashion, structuralanalyses with assumed site geometry andmaterial properties have been carried out.This paper discusses these initial investi-gations.

BACKGROUND

Structural features of a porous rockreservoir can best be described withreference to Figure 1. Air is compressedand forced down the well bore duringperiods where excess generating capacityis not needed. Non permeable cap rockand base rock provide for vertical con-tainment of the air while radial contain-ment is provided by means of interstitialwater contained in the outer regions of

- •?.••«

0

0

TYPICAL '.;ffiISOTHERMVT-;;

''SOT

° o. OVERBURDEN•

mCt CAP ROCK

f. ; l ' POROUS ROCK

lf%f/ BASE ROCK

Fig. 1. Cross Sectional View of a TypicalPorous Rock Site

the porous body. During peak demand, airis drawn from the porous zone to producepower.

Structural loading of the cap, porous,and base rock masses is caused by severalsources. In situ loading is caused byhorizontal and vertical stress fields ori-ginally existing in the rock forms. Thevertical stress level is approximatelyequal to the stress resulting from theoverburden loading. The horizontal compo-

nent depends largely on the tectonic acti-vity of the site region. Both vertical andhorizontal stress components are likely tobe compressive.

The structural behavior due to airpressure in the porous zone will depend onseveral factors. These factors include:the rock porosity, the spatial dependencyof the pressure, and the micro elasticcharacteristics of the porous rock fabric.

Another significant structural loadingmethod is that of non-uniform temperaturedistributions within the rock massesaffected. Heat from the compressed airflows primarily radially in the cap rockand the porous rock. In the cap rock, thisheat flow is by conduction. In the porousmaterial, however, heat is transported byconduction and by flow effects as well. Asa result of this difference, thermal effectscan be expected to decay more rapidly inthe cap rock than in the porous rock, andisothermal lines will appear/some what likea bottle as shown in Figure 1.

The radial temperature gradient inboth the porous and nonporous rock has apronounced effect on rock stresses. Inthe cap rock, hot areas near the well borelocation tend to expand and thus stretch thecooler outer regions. This characteristictends to induce compressive radial and loopstresses near the well bore. The coolerouter rock resisting the expansion willincur tensile loop or circumferentialstress levels.

In the porous rock, radial temperaturegradients will tend to cause the same struc-tural behavior as in the cap rock. Sincethermal gradients in the porous material areless severe, however, the stress levels willlikely be less pronounced.

In addition to radial expansion andcompression, another thermal response modeexists. As the large central region ofporous rock heats up, vertical expansion isattempted. As a result, upward loading onthe central portion of the cap rock willoccur. This will tend to produce flexureor bending stress behavior. As a result,tensile stress will form on the top of thecap rock and compressive stresses on thebottom. The magnitude of this flexuralstress trend will depend upon:

. the axial stiffness of the porouszone

. the flexural stiffness of the caprock

. the compliance of the overburden abovethe cap rock

Loading conditions discussed thus farare essentially independent of daily cyclicfluctuations of charging and discharging.Cyclic stress levels induced by alternatingperiods of charge/discharge could havesignificant structural effects if fatiguestress limits of the materials involvedare exceeded.

PROJECT DESCRIPTION

In order to gain insight into struc-tural behavior, a generic site descriptionwas hypothized and analyzed numerically.Since structural effects are likely tochange from initial charging to steadystate operation, various times duringthe site operation were studied. Thesetimes included initial start up through oneyear of simulated operation.

The structural analysis was precededby an independent thermal/flow analysisdiscussed in Reference 1. The theoryon which the structural analysis was basedassumed that the thermal and flow responseswere uncoupled from the mechanical behavior.This assumption tends to be valid for prob-lems in which the pore structure is stiffwhen compared to the flow medium which, inthis case, is air. Further discussion ofstructural behavior of porous bodies sub-jected to a flow environment may be foundin References 2, 3, and 4.

Both analysis phases of this study(the thermal/flow as well as the struc-tural) assumed axisymmetric models. Thegeometry of this generic site is shown inFigure 2.

The analysis procedure on which thestructural behavior was based made use ofthe poro-elastic formulation^ and thefinite element method^. The elastic stressstrain relationship in this case takesthe following form when using indicia!notation.

(1)

(2)(1-B)P+<*T

350

CAPHOCK

tin m

POROUSHOCK

doom

MtEROCK

• - MODEL AXIt

Ml M>OVtMUNDEN /. " <

UMDWia H Hhull.

11i

too*

AXIAL (VCftnCAU RCSTttAlttr

Table 1. Rock Mechanical Properties

Fig. 2. Model Geometry and Finite ElementMesh Used for Structural Analysis

a.. stress tensor

P pore pressuref rock porosity&•• 6. Kronecker delta

elasticity tensor

total strain tensor

free strain

Poison's ratioYoung's moduluscoefficient of thermal expansiontemperaturebulk modulus ratio of porousrock to interpore material

(.25 assumed after ref. 1)

Equilibrium equations take the form:

1 , = 0,0. (3)

hi

vEaT

Where (, j) denotes partial defferentiationwith respect to the j*" coordinate direc-tion. The above equations were formulatedby using a modified version of the finiteelement computer code AX1S0L5.

Material properties used for thestructural analysis were chosen to repre-sent typical values of shale for the capand base rock forms and typical values forsandstone for the porous rock. Numericalvalues used were taken from Reference 6and are shown in Table 1.

Structural analyses were performedfor various loading conditions of temperatureand pore pressure encountered during thefirst year of site simulation. In situeffects were simulated by using an over-

E (psD

Cap Rockand Base Rock

4.65 x 106

0.25

5.6 x lO"6

Porous Rock

«.35 x 10*

0.25

5.6 x lO"6

20%

burden loading of 562 psi which is equivalentto a depth of about 560 ft. Horizontalstress will occur due to the radial res-traints at the outer portion of the model.These features are illustrated in Fig. 2.

RESULTS

Due to space limitations, only struc-tural results corresponding to the firstand last weeks of site simulation will bepresented here. Figures 3 and 4 showisothermal plots of these two conditions.These figures illustrate the radial growthof thermal effects with increased simulationtime. Pore pressure loadings for both ofthese cases were about 690 psig and quiteuniform throughout the porous media witha variation of only about 25 psig.

-«-250

•160

-.1-260

Fig. 3. Thermal Distribution After OneWeek of Simulation (°F)

Plots of maximum tensile stress andmaximum shear stress for the two casesconsidered are shown in Figures 5 though8. These results show the effect of theincreased thermal zones in all three rockforms. In particular, the differences intensile stress results between week 1 andweek 53 indicates the degree of cap rockflexure caused by vertical thermal growthof the heated porous zone.

351

Fig: 4. Thermal Distribution After 53Weeks of Simulation (°F)

Fig. 7. Contour Plot of Maximum ShearStress (psi) After One Week ofSimulation

Fig. 5. Contour Plot of Maximum TensileStress (psi) After One Week ofSimulation

1400 1000

.200

400 200

Fig. 6. Contour Plot of Maximum TensileStress (psi) After 53 Weeks ofSimulation

Fig. 8. Contour Plot of Maximum ShearStress (psi) After 53 Weeks ofSimulation

The magnitude of daily fatigue orcyclic stress level was investigated byanalyzing the model for stress differencesat the minimum and maximum loading condi-tions occurring during the daily cycle.For these conditions, temperatures at thewell bore were 250°F and 450°F. Correspond-ing input pressure values were 690 psig and735 psig, respectively.

Thermal response to daily cyclingwas observed to occur in the vicinity ofthe well bore. Pressure changes in theporous zone, however, were somewhat uniformand followed the well bore pressure veryclosely.

Fatigue stresses (shear stress) areshown in Figures 9 and 10. As can be

352

780

BOO

260R=12'

• i

1

"—— __^

R=3'

i i i

0 10 20 30 40 60WEEKS - RESERVOIR AGE

Fig. 9. Fatigue Stress Trends on theBottom Region of the Cap Rock

1600

1000 -

600 -

60

Fig. 10.

10 20 30 40 SO 60RESERVOIR TIME - (WEEKS OF AGE)

Fatigue Stress Trends in theCentral Region of the Porous Rock

seen, fatigue stresses tend to be greatestnear the well bore and diminish considerablywith increasing radius. In addition to thistrend, it can be observed that fatiguestresses tend to be greatest during initialperiods of operation and tend to diminishwith time.

The analyses discussed thus far assumedthat material behavior was elastic; thatis, no allowance of material cracking wasconsidered. Rock is notorious for itslack of tensile strength and it is likelythat the tensile stress shown in Fig. 6will result in some localized cracking.

In order to gain some insight intothe effects of cracking, another sequence ofanalyses was performed. In this sequence,temperature and pressure loading was appliedand material stiffness in the direction ofmaximum tensile stress was assumed to vanishif that stress exceeded some thresholdvalue. These new material properties werethen used with later load conditions.

This incremental procedure was carriedout for several time steps during the oneyear simulation. As a cracking threshold,300 psi was assumed to be representativeof shale tensile strength.

Patterns of damaged rock resultingfrom the analysis procedure mentioned aboveare shown in Figures 11 and 12 for thefirst and last week of site simulation,respectively.

Fig. 11. Region Affected by TensileFracture (300 psi) After OneWeek of Simulation

Fig. 12. Region Affected by Tensile Frac-ture (300 psi) After 53 Weeks ofSimulation

CONCLUSIONS

The stress results presented herewere arrived at by analyses based uponseveral important assumptions. The degreeto which these results depict realitydepends of course on the adequacy of

353

these assumptions.

Particularly important in this regardis the assumption dealing with cap rockshape and overburden stiffness. Theassumptions dealing with flat cap rockand no overburden stiffness will mostlikely yield results which show largecap rock flexure when compared to a modelwhich assumes a domed cap rock with stiffoverburden. Thus, the severity of tensilestress levels and the degree of cracking isprobably conservatively estimated by theanalyses performed.

The stress levels in the immediatevicinity of the well bore should be inter-preted with the knowledge that localizedeffects due to the well casing and itsattachment to the rock forms were notmodeled. Global structural trends of sitebehavior were the objective here, notlocalized effects. The consequence offatigue stresses in the well bore vicinitywill be influenced not only on the fatiguecharacteristics of intact rock samplesbut also on the extent of fracturingwhich occurs during the drilling process.

REFERENCES

1. Wiles, Larry E., "Analysis of CAESPorous Rock Reservoirs", Proceedingsof the First Annual DOE ContractorsReview Meeting for Mechanical andMagnetic Energy Storage, Luray, VA,1978.

2. Lubinski, Arthur, "The Theory ofElasticity for Porous Bodies Display-ing a Strong Pore structure",Proceedings of the Second U.S. Congressof Applied Mechanics, 1954.

3. Mordgren, R.P., "Strength of WellCompletions", 18th U.S. Symposium onRock Mechanics, Johnson PublishingCo., Boulder CO, 1977.

4. Stagg, K.G., and Zienkiewicz, O.C.,Rock Mechanics in Engineering Prac-tice, John Wiley & Sons, N.Y., 1968.

5. Wilson, E.L. and Jones, R. M.,"Finite Element Stress Analysis ofAxisymmetric Solids with OrthotropicTemperature Dependent MaterialProperties" Air Force Report No.BSD-TR-67-222, September 1967.

CRC Handbook of Tables for AppliedEngineering Science, 2nd Edition, CRCPress, Cleveland, OH, 1973.

354

POTENTIAL AIR/WATER/ROCK INTERACTIONSIN A POROUS MEDIA CAES RESERVOIR

J.A. Stottlemyre and R.P. SmithPacific Northwest LaboratoryRichland, Washington 99352

ABSTRACT

There appears to be motivation for storing elevated temperature (50-300°C) compressedair in underground porous media reservoirs. The feasibility of this concept dependsto a major extent, on the potential physical and chemical response of the reservoirto any new environmental conditions imposed by Compressed Air Energy Storage (CAES)operations. For example, changes in absolute and relative permeabilities due toelevated and time varying temperatures, interstitial fluid pressures, humidity condi-tions and free oxygen would be of specific interest.

This paper briefly discusses ten potential reservoir damage mechanisms: 1) dis-aggregation and particulate plugging, 2) thermo-mechanical plugging, 3) corrosion,4) clay swelling and dispersion, 5) matrix consolidation, 6) water evaporation andmineral precipitation, 7) oxidation reactions, 8) hydrolytic reactions, 9) mineralsolutioning and precipitation, and 10) fluids incompatibility. These topics are discussedin what is considered their order of importance with respect to the CAES concept.Fatigue failure of well casings and cement materials is also considered on exceedinglyimportant item, but is not elaborated on in this paper. Finally, some explanation cfcurrent and planned experimental and theoretical investigations is presented.

INTRODUCTION

The economic and technical feasibilityof Compressed Air Energy Storage (CAES)technology depends to a major extent onlocating, developing,.and maintaining anadequate subsurface reservoir. The mannerin which such a reservoir may respond,physically and chemically, to new environ-mental conditions associated with CAESoperation is one of the fundamental unknownsfacing the implementation of this promis-ing new technology.

The operating conditions of primaryinterest include: maximum mean pressurewhich is controlled by reservoir depth(0.11 bars/meter), 2) maximum temperaturewhich is apparently restricted to 300-350 Cby standard well bore casing and groutingmaterial limits, 3) maximum loading andunloading rates which are as yet undefined,and 4) maximum inlet air humidity andsuspended solids concentrations. Otherindependent variables, which become con-stant once a site is selected, include:1) the physical, chemical, and mineral-ogical characteristics of the reservoirrocks, 2) groundwater constitutent concen-trations, and 3) the physical and chemicalcharacteristics of the confining strata.

In addressing how CAES operatingconditions might alter the reservoir over30 to 40 years, the following properties(dependent variables) should be stressed:1) absolute and relative permeabilities,2.) effective porosities, 3) bulk compres-sibilities, and 4) the corrosiveness ofany interstitial fluids. In the studyat PNL, a series of assumptions have beenmade to facilitate a description of thesystem. These assumptions are: 1) allphases will attain rapid thermal equilib-rium, 2) a heat pillar will be formedaround each wellbore and will change onlyslowly with time via thermal conductionand vapor transport, 3) a certain percen-tage of water (the residual content) willremain in the air storage zone subsequentto initial bubble development, 4) theresidual water will be gradually butsteadily displaced by evaporation withwarm undersaturated air, 5) the residualwater will be immobile in the liquid phase,6) there will exist a high impedance tothe flow of condensate back towards thewell bore, 7) the reservoir will be rapidlytransformed from a reducing environment toone abundant in free oxygen and carbondioxide, and 8) reservoir heterogeneityand anisotropy will dictate that generalizedand simplified solutions should not be

3 (p 3k

considered as absolute.

This paper briefly discusses tenpotential reservoir damage mechanisms.These mechanisms are ordered in terms ofexpected importance and where warranted,arguments are presented to support orreduce the applicability to CAES reservoirs.The potential damage mechanisms are pre-sented in three categories: Category Iincludes: 1) rock disaggregation andparticulate plugging, 2) thermo-mechanicalplugging, 3) corrosion, and 4) clay swel-ling and dispersion. Category II includes:1) rock consolidation and subsidence, 2)residual water evaporation and mineralprecipitation, 3) oxidation reactions, and4) hydrolytic reactions. Category IIIincludes: 1) mineral solutioning andprecipitation and 2) fluids incompatibility.

Figures 1 and 2 are presented to lendperspective to the discussion. The micro-graphs are of Berea sandstone from anoutcrop in Ohio.

n i t MNOSTONtIUNW.TUUOI

•HE* MNOSTONI{30tfC.I2«lors.*73Hrll

' 0.3pm '

Fig. 1. Micrograph of Unaltered BereaSandstone

Fig. 2. Micrograph of Altered BereaSandstone

Particulate Plugging

Particulate plugging refers to theformation of a "filter cake" of small micronsize particles at the sandface or theformation of barriers to flow due tobridging of particles in channel restric-tions within the reservoir rock itself.Figure 3 conceptually portrays this bridg-ing phenomenon. There are a multitude ofpotential sources for such particles.The inlet air stream can contain suspendedinorganic (dust) particles or micro-organisms. Any mineral scale or corrosionproducts on the inner walls of pipes or wellcasings can result in sloughed particles.The rock itself is a good potential source.Dislodged grains or fragments of grains,frayed clay plates, interstitial cementmaterials and precipitates depositedduring evaporation of the residual waterserve as examples.

One conclusion, based on very limitedexperimental data, is that for at leastelevated temperature CAES concepts, parti-culate plugging may be the most importantphenomenon investigated. Figure 4 presents

356

MimcutMc nuoomo

tMmiTMtMUMIDMCItHIMMNn

mmcunW § CQMOMON MOOUCTS

Fig. 3. Conceptual Representation ofParticulate Plugging

•EREA SANDSTONEIN A WATER VAPOR-AIR ENVIRONMENT

(436 Hre. IMBanl

II

V"

** -'

sot

H1

300*C

Fig. 4. Disaggregation in a Water Vapor-Air Environment

two initially identical cores of Galesvillesandstone (an actual natural gas reservoirrock from Illinois). Both samples we-eexposed to an atmosphere of air and watervapor at 134 bars for a period of two weeks.The only difference was that the core onthe left was reacted at 50°C and that onthe right at 300°C. Although the testconditions were such that definitive conclu-sions cannot be drawn, it is apparentthat substantial grain dislocation andsuperficial erosion occurred at the highertemperature.

To study this phenomena in more de-tail, a fluid flow system is requiredwhereby gas and/or liquid can be ventedthrough a large diameter core under care-

fully controlled pressure, temperature,and humidity conditions. Figure 5 showsa distribution of particles that werepumped into a Berea sandstone. In thiscase, the flowing fluid was distilledwater. Figure 6 shows the incurredpressure drop across the core and theeffluent and particle discharge downstreamfrom the core. u' Nearly total pluggingwas the result. Backflushing was only ofmoderate assistance due to the near randomdistribution of pore and channel sizes inthe sample.

8EKCA NUMBER 1

CONCENTRATION- 112,000 countslnlMEAN-4.25| iMODE-4 .35MM E D I A N - 4 . B M

25 -

Fig. 5.

PARTICLE DIAMETER, microns

Particle Size Distribution forBerea SandstoneO)

It is concluded that: 1) investiga-tion of particulate plugging may be veryimportant for potential CAES reservoirs,2) filtering of the inlet air may benecessary, 3) sand screens may be necessary,4) filter cake formation is not likely, butbridging may occur, and 5) elevated tem-perature and cycling may aggravate theproblem, but low temperature CAES conceptsshould not be considered immune.

Thermo-Mechanical Plugging

In a recent study on the effects oftemperature or relative and absolute per-meabilities of Berea and Boise sandstone,it was concluded that the permeability towater was reduced substantially and the

357

SO 100 150 200 250 300 350 400 450 500

THERMO-MECHANICAL PLUGGING (REVERSIBLECOMPONENT)

SO 100 150 200 250 300 350 400 ISO 500

700

600

500

L, *mi

! 300

rm -

wo -

EEREA CORE NUMBER 1

i—>—-i r i t

r--r-i T iT-rrri m100 150 200 250 300 350

TIME iminutesl

400 450 500

Fig. 6. Pressure and Effluent ParticleConcentration Versus Time forBerea Cores'1)

irreducible water content increased sub-stantially in going from room temperatureto 80°C. "For t!ie five Boise sandstonecores on v/liicli displacement experiments wererun, average (liquid) permeability decreasedfrom 2,050 md at room temperature to 884 mdat 80°C. This is an important new observa-tion not made previously, to our knowledge,in the petroleum literature."(2)

Figure 7 graphically presents the re-sults of the tests. The main points are:1) these results are for liquid water, 2)permeability to oil showed substantiallyless temperature sensitivity, 3) air

1000

500

0

N . BOISE SANDSTONE

BEREA SANDSTONE

i 9

—o

11OO

TEMPERATURE ( X I

Fig. 7.Effect of Temperature on AbsoluteLiquid Permeability of BereaSandstone(2)

permeabilities were not measured and nointuitive conclusions should be drawn basedon a comparison of densities and viscosities,4) the permeability reductions were rever-sible, i.e. recovered as the temperaturewas decreased, 5) the cause is probablythermally-induced increases in rocktortuosity and is not due to increasesin compressibility or decreases in poro-sityw), and 6) similar experiments shouldbe conducted on an CAES candidate rock.Note that since the effects are reversible,only measurements under elevated tempera-ture and pressure will reveal the decreasedpermeability. Post-test measurements atroom temperature are insufficient.

Corrosion

In a CAES reservoir, corrosion couldbe a serious problem if the near-wellboreregion is not maintained totally free ofliquid water. Table I presents the re-lationship between corrosion rates andvarious reservoir conditions for geothermalconditions.^ ' Free oxygen will aggravatethe problem. The potential results ofcorrosion include: damage of casing mate-rial and surface pipes, blowouts or rapiddepressurization and the formation ofparticles that may subsequently plug thereservoir.

Clay Swelling and Dispersion

Many sandstones are known to be watersensitive, i.e. permeability to waterdecreases as the water salinity decreases.

358

Table 1. Effect of Various Parameters onthe Corrosion Rates of Iron BasedAlloys in Geothermal BrinesW

CLAY PRODUCTION, SWELLING AND DISPERSION

l-actorWater Content 'Oxygen ConcentrationSalinityAcWty

TimeTemperature*Silica Concentration

Direction Effect on Corrosiont

f

t

•related to scaling phenomena

In tfie CAES concept, water vapor will beinjected with the airstream and/or will beproduced from residual water in the warmernear well bore zones. As the air and vapormove farther out into the reservoir wherethe temperatures are decreasing, condensa-tion and dilution of existing formationwaters may occur. This could result in theswelling of residence clays. This swelling,in of itself, should not plug the reser-voir since the effected zones will be areasonable distance from the wellbore.However, clay plate fraying and particledispersion may result in particulateplugging much closer to the wells sinceair would sweep particles in that direc-tion during the production phase of theCAES operating cycle.

Figure 8 conceptually portrays thisphenomenon and figure 9 shows the signifi-cant amount of interstitual clay that canexist even in natural gas reservoir rocks.Only bentonite clays tend to cause severeproblems. Berea sandstone has less thanVI by weight of bentonite clay. Figure 10shows how Berea sandstone permeability towater decreases with decreasing watersalinity. Air permeability should not beas dramatically affected, but particledispersion from zones containing residualwater ma., be a general particulate pluggingproblem. Fortunately,pretreating a reser-voir with hydroxy-aluminum compound tendsto significantly inhibit the problem.C5/Testing of potential CAES candidate rocksand treatment when necessary should beconsidered for CAES projects.

Matrix Consolidation

For Berea sandstone, the dry bulk compres-sibility decreases by approximately 20percent in going from 20°C to 200 C at4000 psi effective stress.(3) However,this decrease is much less then the tremen-

SOURCE: ORMMALRESERVOMCONTENT HVDROLYTKREACTIONS

CAUSE: DECREASED SALINITY DEWATERINO

TREATMENT: HVOROXY-AWMINUM

SWELLING CLAY PARTICLE PLUGGING

fresh water fresh water

Fig. 8. Conceptual Representation of theClay Plugging Phenomena

dous scatter in compressibility magnitudesfrom sandstone to sandstone. Although thepotential for matrix consolidation andsubsidence cannot be arbitrarily dismissedat this point for elevated temperaturestorage, it is considered a lower priorityissue.

It is being suggested that rock sof-tening due to the production of clays fromother silicates and carbonates may be thecause of observed reduction in penatrometricstrength with temperature. Berea sandstoneexhibits as much as a 30 percent decrease inpoint load strength after five days ofsteam treatment at 300°C.(°) Fortunatelyammonia pretreatment of the reservoirs seemsto totally eliminate the rock softeningproblem. " Such a treatment might be consi-dered for CAES projects.

359

SAMPLE -6

Fig. 9. Micrograph Showing InterstitialClay in Mt. Simon Sandstone fromthe Media Natural Gas Storage ••Field in Illinois.

Evaporation and Precipitation

Residual water may coexist with ele-vated temperatures for a period of time.The dissolved solids and adjacent rockmineralogy may be altered. Subsequentevaporation of the water will result indeposition of the dissolved solids. Know-ledge of the composition and friabilityof the residue is necessary since this maybe a potential source of mobile particles.Naturally, this will be a function of theinitial salinity of the aquifer.

Oxidation Reactions

Subsurface reservoirs tend to be re-ducing in nature. The CAES concept callsfor the introduction of free oxygen andcarbon dioxide. Iron and manganese mineralswill oxidize and precipitate as solidresidues. In general, metal oxides areless dense and less soluble than metallic

20 30 10

THROUGHPUT (ml/cm?)50

Fig. 10. Water Sensitivity (Clay Pluggingof Berea Sandstone in the Pre-sence of Progressively DecreasingSalinity Waters(5))

Salts. The quantity of iron and manganeseis such that precipitation-type pluggingshould not be a problem especially sinceprecipitation will primarily occur in zonesalready occupied by residual water andtherefore never available to the air anyway.However, once again, the solid residue maybe a source of mobile particles. It issuggested that reservoir rich in iron ormanganese be studied carefully and thatdewatering of near well bore zone be expedited.

Hydrolytic Reactions

Hydrolytic reactions refer, in thiscase, to the production of clay formingminerals from feldspars and carbonates.These reactions require liquid phase waterand in general reaction rates can be roughlyconsidered to double for a 10°C temperatureincrement. Potential consequences includethose already mentioned for clays, e.g.swelling, dispersion and rock softening.It is being suggested that hydrolytic reac-tions may occur for a short period of timein a CAES reservoir but that evaporationof the necessary liquid phase water may out-pace the slowly advancing temperature frontand therefore hydrolytic reactions should beof secondary concern. Reservoir hetero-geneity and anisotropy dictate that theproblem not be totally dismissed from con-sideration, however.

360

Mineral Solutioning and Precipitation

Silica, carbonate, and sulfate scalingare severe problems for the geothermalindustry. Changes in temperature oracidity can drastically affect the solu-bility of these minerals and the responseto temperature is diametrically opposedfor silica and calcium carbonate. Theproblem is being considered secondary forCAES, however. This is because the onlyavailable liquid phase water that may beheated to any degree is the residual con-tent left after air bubble development.This residual or irreducible water is es-sentially immobile in the liquid state.Furthermore scaling, if any, should occurin zones initially occuppied by waterand not available to the air phase anyway.The only possible problem appears to bethe sloughing of scale deposits and theassociated production of mobile particles(formation fines).

PNL Experimental Program

Current experiments to investigateair/water/rock interactions consists ofautoclave tests and thermal propertiestests. A flow system capable of ventingliquids and/or gases and suspended solidsthrough large diameter cores under con-trol red. temperature, pressure, and humiditycondition? is in the design phase.

Table 2 presents the preliminary ex-perimental strategy for autoclave tests ona site specific candidate rock and compatiblegroundwater. Analytical techniques include:optical petrography, x-ray fluorescence,x-ray diffraction, electron microscopy,quantitative shade differentation, ionchromatography, plasma-arc spectrometry,atomic absorption, and standard wetchemistry. Permeability measurements aremade with standard permeameters. Auto-clave tests at 300°C have resulted in somereduction in liquid permeability, and amore moderate reduction in air permea-bility. General quantitative conclusionscannot be made at this time based on thelimited number of experiments and thesite specific nature of the problem, how-ever. Tests to date have been on Bereasandstone as a control sample and onGalesville sandstone (Media natural gasstorage field) as an example reservoirrock.

The next phase of the experimentalprogram will include design and fabricationof a fluid flow test facility for large

Table 2. Screening Autoclave Experimentsfor Site Specific Sandstone

» FOR OAUSVHU ROCK AND OROUNOWATER

diameter (10-30 cm) rock cores (Figure 11).The facility will be designed to permit:1) axial flow of gases and/or liquidsthrough the core, 2) introduction ofsuspended solids, 3) temperature from 50-500°C, 4) confining pressures from 60 to400 bars, 5) uniform temperatures or im-posed temperature gradients, and 6) variousinclinations of the core barrel to investi-gate drainage, imbibition, and advectiveheat transport. Naturally it would beideal to have the system triaxially loadedinstead of hydrostatically loaded but thisdoes not appear to be within the projectscope.

PRESSURE VESSELA W

HEATER

commiiONVALVE

GASSOURCE

tTANDP

•O I(T) UNIFORM TEMPERATURE

© rEMFWATURt PROFILE

® CONFINING PRESSURE

® UNIFORM FLUID PRESSURE

© FLUID PRESSURE PROFILE

© HORIZONTAL TO * > ANGLE

AFFLUENTCOLLECTION

AND MONITORING

CIRCULATINGOPTION

Fig. 11. Conceptual Representation of theFluid-Flow Test Apparatus

The following tests are examples ofthe planned flexibility of the facility:1) porous media evaporation rates andprocesses, 2) thermomechanical permeabil-ity degradation, 3) particulate plugging,4) rock disaggregation, 5) geochemicalplugging, 6) clay swciiing and dispersion,7) corrosion, 8) bulk compressibility,9) geochemical reactions (hydrolytic andredox), 10) mineral solutioning and pre-cipitation, 11) fluids incompatibility,12) liquid mobility, drainage andimbibition, 13) advective heat transfer,

361

and 14) chemical pretreatment.

Theoretical Efforts

A computer code based on thermodynamicequilibrium states has been used to predictchanges in rock mineralogy and liquidconstituent concentrations. Reactionsrates are not considered. The resultsfor Berea sandstone reacted with capatiblegroundwater at 100, 200, and 300 C arepresented in Table 3.

Table 3. Total Dissolved Solids Concen-trations and Mineral WeightPercents for Berea Sandstone at100, 200, 300°C (EquilibriumSimulation)

Groundwater Constituent

A)

K

«1

Ca

"91*1.2

1*1+3

SI

C

%H,

Mineral

CalclttQuartz

Corundun

Kyanite

Pyrope

Alternate TA

Pyrolusite

Treaollte

Chlorite

Pnlogopite

Adulerta

Zoisite

Eoldote

rajc

Paragonlte

Ca-Hontwrill initeLow Alfaite

Ootontte

Kaolinite

AnorthUe

Microcline

Original

6.3E-4

7.1E-4

J.JE-3

5.7E-4

-

-

Original

8.0E-03

8.5E-01

-

-

I.OE-02

3.0E-02

-

1.0E-02

-

2.OE-O2

2.OE-03

4.0E-02

2.0E-02

I.0E-02

Concentration (Holts/Kg H:>0)

100°C

Z.3E-06

5.4E-03

2.1E-02

2.2E-O5

1. IE-OS

2.9E-I7

1.9E-07

S.OE-04

9.6E-03

2.3E-02

3.8E-3S

10QOC

1.5E-0?

1.0E-00

7.1E-0?

1.5E-0!

1.2E-04

4.8E-04

I.OE-02

1.5E-04

1.4E-02

7.1E-03

2.8E-02

8.5E-O3

1.0E-0!

-

-

-

-

iOO'C

9.3E-04

1.3E-04

1.4E-03

1.5E-06

2. )£-O7

2.1E-16

3.U-0S4.7E-03

1.4E-O3

1.5E-O2

7.4E-2J

zxfic

9.4E-01

S.iE-02

3.4E-03

_

1.0E-02

8.3E-06

3.1E-04

2.7E-O2

1.4E-02

1.6E-02

1.0E-02

1.4E-04

1.7E-02

K5E-07

-

-

3D0°C

2.BE-03

3.0E-O4

4.5E-03

l.lt-OB

2.7E-07

3.7E-I5

1.3E-03

1.1E-02

1.6E-04

l .U-02

I.7E-21

300°C

9.2E-01

-

7.6E-02

-

-

8.7E-O3

-

3.1E-02

9.SE-03

1.6E-02

I.OE-02

-

-

-

1.6E-02

-

-

Some important points noted in theCAES Reservoir Stability work thus farinclude:

Critical CAES parameters. initial rock physical and chemical

properties. initial groundwater constituent

concentrationsirreducible water content

. inlet air temperature, pressure,humidity, and suspended solidsconcentration

Critical formation properties. absolute and relative permeabilities

corrosion potential. caprock threshold pressure (macro)

Most likely damage mechanisms. disaggregation. particulate plugging. thermo-mechanical plugging. thermal fatigue of well casings

and cement grouting

Primary theoretical and computer studyareas. thermo-mechanical stress analysis. heat, and mass transfer analysis. particle transport in porous media

geochemical equilibrium reactions

Primary laboratory scale study areas. autoclave air/water/rock tests

thermal properties identification. thermal fatigue analyses

particle transport in porous media. desaturation of porous media. thermo-mechanical plugging tests in

a flowing system. geochemical alterations in a flowing

system

REFERENCES

1. Donaldson, E.C. and Byron A. Baker,"Particle Transport in Sandstones"Society of Petroleum Engineers Conf.,Denver, CO 1977.

2. Weinbrandt, R.M., H.J. Ramey, F. J.Casse, "The Effect of Temperature onRelative and Absolute Permeabilityof Sandstones", Society of PetroleumEngineers Journal, SPE 4142, 1975.

3. Somerton, W.H. and A. K. Mathur, "Effectsof Temperature and Stress on FluidFlow and Storage Capacity of PorousRocks", Proceedings of the 17th Sympo-sium on Rock Mechanics, 1977.

4. Shannon, D.W., "Corrosion of Iron-BasedAlloys Versus Alternate Materials inGeothermal Brines", PNL-2456, 1977.

5. Reed, M.G., "Stabilization of Forma-tion Clays with Hydroxy-Aluminum Solu-tions", J of Petroleum Technology, July1972.

6. Day, J.J., B.B. McGlothlin, and J.L.Hewitt, "Study of Rock Softening and Meansof Prevention During Steam or Hot WaterInjection:, Society of PetroleumEngineers Journal, SPE 1561, Feb, 1967.

362

PROJECT SUMMARY

Project T i t le : State-of-the-Art Review and Formulation of StabilityCriteria for Underground Caverns Used for CompressedAir Energy Storage

Principal Investigator: Dr. Paul F. Gnirk

Organization: RE/SPEC Inc.P. 0. Box 725Rapid City, SD605/343-7868

57709

Project Goals: (1)

(2)

(3)

Survey, evaluate, compile, and document those, physicalchemical, petrophysical, thermal, and mechanical charac-teristics and properties of rock types and geologicalformations for which the CAES concept is applicable, forconditions of elevated temperatures and pressure andmoisture presence;

Formulate a preliminary set of design and stabilitycriteria for underground CAES caverns with expectedlifetimes of 30 years;

Identify areas of analytical, laboratory, and field scaleresearch that are necessary for establishing a consistentand applicable set of final stability and design criteria.

Project Status:

Contract Number:

The survey of literature has been effectively completed, andthe rock properties appropriate to the CAES concept are beingevaluated and documented. The formulation of a preliminaryset of design and stability criteria for underground CAEScaverns is currently in progress. Finally, specific areas ofresearch need are being formally identified and described.

Sepcial Agreement No. B-51225-A-L/Prime Contract EY-76-C-06-1830.

Contract Period: June 1, 1978 - Nov. 30, 1978

Funding Level: $24,900.00

Funding Source: Battelle Pacific Northwest Laboratories

363

PRELIMINARY DESIGN AND STABILITY CRITERIA FOR CAES HARD ROCK CAVERNS

Paul F. Gnirk and Debra S. Port-KellerRE/SPEC Inc.P. 0. Box 725

Rapid City, SD 57709

ABSTRACT

The primary objectives of this study have been to compile rock properties data fromthe literature; to formulate a preliminary set of design and stability criteria forunderground CAES hard rock caverns; and to identify areas of required research. The rockproperties compilation has concentrated on assembling and evaluating those physical,chemical, petrophysical, thermal and mechanical characteristics and properties of igneous,metamorphic, and sedimentary rock types for which the CAES concept is applicable. Thissurvey has indicated that the data base for hydrological and thermomechanical propertiesis generally inadequate for jointed rock under conditions of elevated temperature andincreased moisture presence. In general, the formulation of preliminary design andstability criteria for CAES caverns has been based on information available for largeunderground caverns utilized for mining purposes.

INTRODUCTION

Conceptual design studies have beenrecently conducted in the U.S.A. to iden-tify the potential for using mined cavernsin hard rock to store compressed air foruse in electric utility load-levelingoperations 1 3. The concept involves com-pressing air during off-peak demand periods,storing it in underground caverns, andheating and expanding it through turbinesto generate power. The cavern storageconcept is generally categorized as(1) Compensated: Constant pressure withvarying volume (wet system with hydrauliccompensation), and (2) Uncompensated:Varying pressure with constant-volume(dry system). The useful life of a CAES(Compressed Air Energy Storage) cavern istentatively set as 30 years, overapproximately 7,500 cycles.

Milne, et. al.1 consider a referencecavern design with a volume of 385,000 mat a depth of 690 m, a 20 hour storagecapacity, and a storage pressure of about66 atm. Mailhe, et. al. describe a caverndesign with a volume of 300,000 m , com-pression/production durations of 7-8 hours/6 hours, and a working pressure of about25 atm. Wittke, et. al. 5 discuss a con-ceptual CAES design with a volume of100,000 m^ and a storage pressure of about50 atm. Depending upon the temperature ofthe inlet air and the heat exchange betweenthe air and rock, the above situations may

give rise to cavern air temperatures ashigh as several hundred °C. The cyclicnature of the pressure-temperature/air-water fluctuations during CAES cavernoperation could conceivably perturb thestress field/strength characteristics inthe rock structure and lead to unacceptableglobal and/or local instabilities.

The specification of CAES caverns inhard rock immediately categorizes thesituation to one involving igneous(granitic), possibly metamorphic, andcertain sedimentary (limestones, marbles,dolomites) rocks. The categorization"hard rock" is an implication of relativelygreat strength and resistance to largeductile and creep deformation. Thus, wemay eliminate from consideration such rocktypes as salt, shale, and the like. Con-versely, however, this categorization mustnecessarily include the influence of joints(or planes of weakness), permeability (andgroundwater presence), and elastic/brittlemechanical behavior (or fracture initi-tion) on cavern stability. Vhe cyclictemperature and pressure fluctuations inthe CAES concept also introduce the possi-bility of elastic/ductile/brittle rockbehavior with "fatigue" and possiblytransient "creep rupture" contributions.

With regard to establishing stabilityand design criteria for underground hardrock caverns used for CAES, the goals ofthis current study have been to:

364 -

(1) Survey, evaluate, compile anddocument those physical, chemical, petro-physical, thermal and mechanical charac-teristics and properties of rock types andgeological formations for which the CAESconcept is applicable, for conditions ofelevated temperature and pressure andmoisture presence;

(2) Formulate a preliminary set ofdesign and stability criteria for under-ground CAES caverns with expected life-times of 30 years;

(3) Identify areas of analytical,laboratory, and field scale research thatare necessary for establishing a consistentand applicable set of final stability anddesign criteria.

This paper summarizes the work accomplish-ed to date in satisfying the above goals.

GENERAL GEOLOGIC CONSIDERATIONS

INTRODUCTORY REMARKS

CAES in mined hard rock caverns willrequire extensive geological and geotech-nical assessments for any potential sitingarea. Although each particular site willbe a unique geologic environment with itsown peculiarities, certain characteristicsmust be evaluated for each of the severalgeologic environments that may be suitablefor CAES, including: (1) regional andlocal petrography (mineralogy and fabricof the cavern host rock) and stratigraphyor relations of various rock formations toone another, and (2) primary and secondarystructural features and/or those ofregional as opposed to local extent. Thesetwo general geologic characteristics andtheir significance in potential cavernstability will be briefly discussed forsituations that could be common to igneous,metamorphic, and sedimentary host rocks.

COMMON GEOLOGIC CHARACTERISTICS

The complete petrographic descriptionof a rock includes its primary mineralogiccomposition and that of associated mineral-ization or discontinuities. Also includedare its macroscopic and microscopictextural and fabric features, includinggrain size, grain shape, grain orientationor lineation, foliation, schistosity, andmicrofracturing. These factors essentiallydetermine the strength, elastic, thermal,and hydrological properties of the intactrock and also its weathering characteristics- all critical factors in the cyclic

pressure/temperature and moisture environ-ments of a CAES cavern. In addition,contacts (such as bedding and plutonboundaries) between various rock types inthe rock mass (whether gradational orabrupt) must be carefully defined todetermine the extent of competent rockand to indicate possible zones of weakness.

The nature and extent of structuralfeatures of a potential host rock massmust also be defined - both primaryfeatures (those formed at the time of rockorigin) and secondary features (thosesuperiir-osed on original features). Itshould be emphasized that such featurescan appear in all rock terranes (a geologicarea considered in relation to its suit-ability for a specific purpose) and includemajor and minor faults, folds, and jointsystems, to name a few. These featuresmay define the major strength, elastic,thermal, and hydrological properties ofthe rock mass as a whole. Joint systems,faults, and fracturing associated withfolding, in particular, may produce crit-ical zones of weakness and may alsoindicate an active seismotectonic situation.Likewise, such features, particularlyintense folding, may indicate regions ofextremely high horizontal in situ stress.

IGNEOUS ROCK

Igneous rocks are those rocks formedby solidification of molten or partiallymolten rock material either at some depthin the earth (plutonic or intrusive rocks)or on the earth's surface (volcanic orextrusive rocks). Compositional classi-fication of igneous rocks is generallybased on the percentages of quartz and thepercentages and types of feldspar mineralsin the rock. Textural classification,although somewhat complicated, is basedmainly on predominate grain size. Intrus-ive igneous rocks are generally phaneritic- coarse to medium grained (grains greaterthan 1 mm in size or easily viewed with theunaided eye), while extrusive igneousrocks are generally aphanitic - finegrained (grains less than 1 mm in size).Figure 1 includes a topical igneous rockcompositional classification scheme withaccepted names for phaneritic and aphan-itic compositional equivalents.

Intrusive igneous rocks (of allcompositions) formed at depth occur inmasses (plutons) of all sizes and shapes.They are generally considered to be themost favorable geologic terrane for CAES

365

caverns for several reasons: (1) they maybe found in masses sufficiently large toaccommodate the CAES system; (2) theintact rock is generally homogeneous andisotropic; (3) the petrography, particu-larly the interlocking crystalline texture,imparts generally high strength and favor-able elastic, thermal, and weatheringproperties to the rock; and (4) porosityand permeability are generally low,obviously favorable characteristics whenconsidering cavern air and water leakage.

ICHEOUS HOCK CLASSIFICATION

ESSENTIAL

M t f M L S

D U U T Z

>1QX

OuARTI

FlLOSMlHOIDSRlPLACtOllARTI

FttDOniUMElYPOIUSIW

M I S B I B 1 J WDOniUMELtPlAfilKLASE] riMIOCLASCFClDSPM I FELDSPAR

DIUH PLAGIOCLASt'*!*—INHftntDIAtt*

b U M i f i - GRANODIORITC(jttaOUTE - DMtTC)

6R0U*

SUNITC(TRACHYTE)

GROUP

DlMtTC

( A H D E I H E )

6iour

FfLDSPATHOIOAL S«E«1ti(lILOSPATNOIOAL t U C H i l l )

5 R O I #

CALCI IH••—FLAC—»

IRONMB

fHSHESttffniMCMLS

SAIMO

(IAIALO

Fl LDJPA1 HOI DALC*if*o

(FELDIPAIHOIOALSttALI CROUP)

PfllDOTlTE

GROUP

KMmmic IIKK CLASSIFICATION

1£

o

XOCKKAHE

Sun

PHYUITI

SCHIST

GMIIII

Pkl*COM6LOMUT[

OuARTIITt

cawsniw

TfT!

s| 4

DcroitncD ru&nEHis OF AM*

• « • vtn

0AIARTZ

C A L C I T E OK DoLonitt

TEXTURE

VERY FIMI C«*ttti

Fi-t GKAIKI

CO*IJE GRAINS

COAKI GRAINI

FINE TO Count

SEDIHEKIin ROCK CIASSIF1CMIBI

i

ROCK I1AIC

CONGLOHtRATC

V A > I M S

Ty»E)

SLITSTOKC

SHALC

VANIOUS

llMCSTME

UmDOLONUI

CHEIT

RDC< SALT

CCHPOSITIOM

FnAWtCNTS OF ANV IDCI

TtPt

OuANTI, VARIOUS MOUNTS Of

' U f i i ' M , • « • »ACMfMT»

OtMitrz JIKD CLAY HIWEIULS

C o u n t (CACOJ)

DOLOHITE CUV; ICOjlj

CKKLCEDONV (SlO2>

CIPSUN (CAS04 • 2H?0)

HAL l i t (HACL)

TF.XTORE

COARSt ROUNDtD t « A I « > 2 m

COAKSE ANCULM (RAIMS » 2 MM

REDIUH C R A I N I U/1G TO 2 wi>

FlHt GRAINS (1 /256 ID 1/16)

fllCkD TO COARSE CRYSTALS, fOSSILAND FOSSIL fflAGMENTS

S I M I L A R TO LIMESTONE

CRVfTOCRVSTALLINE, OENSC

f l « t TO COARSE CRt5T*tJ

FIG. 1. GENERIC ROCK CLASSIFICATION (AFTER JACKSON,

TRAVIS', HAMBLIN i HOHAHD^, AND MCKENZIE, ET.AL.^)

Extrusive igneous rocks (of allcompositions), formed at the surface or atvery shallow depths in a variety of ways(such as nonviolent lava flows and violentvolcanic eruptions), are generally lessfavorable for CAES caverns for severalreasons: (1) they are often found in massesof limited extent; (2) the rocks may behighly vesicular, with high porosity andpermeability and generally low strength.The most favorable igneous extrusive is

probably one formed by massive, nonviolentflow of lava onto the earth's surface. Ifsuch a flow occurs in sufficient mass tobe a cavern host, other characteristicscited as favorable in igneous intrusiveswould also generally apply.

METAMORPHIC ROCK

Metamorphic rocks are igneous, sedi-mentary, or metamorphic rocks which havebeen altered by temperature and/or pressure,Metamorphic petrography is extremely compli-cated; however, primary classification canbe based on the presence of laminatedstructures in the rock resulting fromsegregation of different minerals intolayers (foliation). Foliated metamorphicrocks include slates, phyllites, schists,and gneisses. Nonfoliated metamorphicrocks include quartzites and marbles.

Nonfoliated rocks, such as quartzitesand marbles, when they occur in suffi-ciently large masses, may be as suitablefor caverns as igneous intrusives. Suchrock is generally homogeneous and isotropicand has good strength characteristics,particularly the quartzites. Marble,however, may occur less frequently in largemasses, have somewhat lower strength, andbe subject to undesirable weathering andsolution effects since it is primarilycalcium carbonate. The desirability ofmarble for other commercial purposes wouldalso tend to eliminate it as a potentialhost rock.

The major difficulties in CAES cavernstability in foliated rocks would probablyarise from the foliation itself, or fromschistosity (parallel grain alignment).Foliation and schistosity planes may causethe strength, elasticity, and thermalproperties of the rock to be anisotropicand highly variable. In addition, weather-ing of such rocks may be accelerated alongplanes of anisotropy and by the presenceof easily altered metamorphic minerals(chlorite, talc, etc.). This presentstudy has included all types of metamorphicrocks, but concentrated on competentquartzites.

SUMMARY REMARKS

The most favorable and most likelyhost rock types for CAES in mined hardrock caverns appear to be igneous intrusiverocks. Other possible hosts includeigneous extrusive rocks, dolomites andlimestone, and massive metamorphic rocks,

366

particularly quartzites. Possibilitiesalso exist for locally favorable conditionsin other rock types.

In the general sense, considerablegeologic literature exists concerningpetrography and structure of all favorable(and unfavorable) geologic terranes dis-cussed here for CAES. We emphasize, how-ever, that each individual site consideredwould require its own detailed geologicalinvestigation.

GEOTECHNICAL PROPERTIES

The geotechnical properties of a rockmass include the hydrogeological charac-teristics (porosity, permeability), geo-thermal gradient, in situ stress state,and joint characteristics. The jointcharacteristics are generally quite sitespecific and depth dependent as regardsspacing and orientation.

HYDROGEOLOGY

The literature survey of hydrogeologyand case studies of various undergroundfacilities, including underground storesand mines, suggests important hydrogeolog-ical considerations and criteria to be metfor CAES hard rock cavern siting. Primaryconsideration must be given to such factorsas: (1) the hydraulic characteristics ofthe intact rock and of the rock mass as awhole; (2) groundwater behavior; and (3)groundwater chemistry.

General Hydrogeological Considerations.The important hydraulic characteristics ofa rock or rock mass include primary andsecondary porosity and permeability (syn-onymous here with hydraulic conductivity).Porosity naturally determines how much freevolume is available to contain fluids suchas air and water. Porosity of intact rock,due only to pores or voids within therock matrix, is said to be primary, whereasthat porosity due to fractures, fissures,faults, or joints in a rock mass as awhole, is said to be secondary. Likewisepermeability, or the ability of a rock totransmit air or water, may be either primaryor secondary. Primary permeability dependson effective porosity, or on the assemblageof connected void spaces in intact rockthat allow fluid flow. Secondary permeabil-ity depends on the transmission of fluidsthrough discontinuities in a rock mass.Low porosity, and more importantly, lowpermeability are essential in CAES hostrock masses in order to (1) minimize

groundwater inflow during construction andoperation of the cavern, and (2) tominimize air leakage from the cavern duringoperation.

Groundwater behavior is also a prim-ary consideration, particularly in regardto stability and consistency of thesaturation zone depth, or that depth atwhich voids and fissures are filled withwater under hydrostatic pressure. Theoptimum condition for successful CAEScavern operation is that the cavern belocated entirely in the saturated zoneand that after initial constructiondisturbances, the upper boundary of thesaturated zone (phreatic surface) remainstable. Water-filled voids and discontin-uities should effectively retard cavernair leakage, while low permeability(although under saturated conditions)would prevent excessive water inflow tothe cavern.

Finally, groundwater chemistry andits stability may or may not be important,depending on rock type, in successfulcavern operation. Unusual chemistry orchemical changes, perhaps caused by tem-perature fluctuations within the cavern orby other external factors, could causeaccelerated degradation of the rock massand subsequent cavern instability - eitherundesirable surficial wall effects or grossstructural instability. In addition,undesirable chemical constituents occurringnaturally in the water or later due tocavern operation, could pose difficultiesin successful equipment operation.

Particular Hydrogeolosical Considerations.Table 1 summarizes the magnitude of per-meability generally found in various rocktypes, both for intact rock and rock masses.It is evident that igneous intrusives, someigneous extrusives (particularly flowbasalts, rhyolites, or trachytes), highgrade metamorphics, and some limestonesand dolomites (those devoid of karstfeatures) have the most favorable perme-abilities for optimum CAES cavern oper-ation. Most extrusive vesicular igneousrocks, some quartzites and marbles, andmost clastic rocks have generallyunacceptable permeabilities.

Generalization of the stability ofgroundwater conditions is much moredifficult, owing to the very site specificnature of the problem. Considering avail-able literature, the prediction of stabil-ity on the basis of rock type is virtually

367

impossible. Other factors external torock type and to the cavern situation,such as precipitation, runoff, and with-drawal for consumption, are also signifi-cant.

TMU 1

UKB OF HTMAUUC MC« KOrtltlES

ROCK TYPE

IWEOUI tlNTftUIIYt)

ISNEDUS <ERT«USIVE)

SID1HINTAIV

(CHEMICAL)

UDIMCNTAJtY

(CLASTIC)

KETMOMHlC(FOLIATED)

K1WKMPHK(NOMfOLlAHD)

tCKMt «wos m muwnicCONDUCTIVITY

M U M * * : UWl (CCOMWlr: SCMULLV IONftUT VAIIEt MEATLT MHNOIIK ON 0ISCOMTINUIT* OIHtHIIOMS MO MttvCHCV

FtlHARV: LOO, IXCCPT tO* SOME VEIICULAIMCKIi lECOWUtr: LOW TO NJCM, DtUWB-IH6 ON VEIICULAH NATURE AND OH OIICON-T1NUITT DIHtRIIOMS AND FREOUEKCV

MlftUtr; LM 10 H IWl tECONDMV: LON

TO HICH, M'EKDINC Oil MEIENCE Of KMIFEATURE* ANB ON onco*Ft»»«m D I M » -tlOHS AKO FflEQUCNCYM I M W : ««RALLV HUH fO* SANDSTONES,W I C C f U , AND COKSlOMCIUTCt, LOW rotSilUTONEt AM IHALElj SICOHOAIIVi LOM

MINARI: 19*) HCOHDAIIV: LOH TO HI6HAMD HICHLf DlltECTIOHAL. OE'IHOING ONKHISTOSITT, fOLMTIOK M6 Cf«MDl JIHUITlCI

nitVJtr: LOmj HCONAUV: LOM TO MICH

DEPENDING OH DIIC0NTINUITIC1 AXD

SOLUTION FIMUKEl

poncsm(X)0.01

TO

11.2

0.5TO

54.0

*o.oTO

35.0

0.65TO

25.0

0.2TO

1.7

- 0 . 0

TO

3.0

CMEAMUn(c?>

TO

10-12

TO

io-l°

io-»sTO

io-i»TO

10"«

HO DATA

AVAILULC

M DMA

AVAIUIU

(f.T.WLICmurinn

(TVS)

• M-«TO

lo-HTO

Id"'

TO

io-w

io-»TO

10-5

u-iiTO

lo-«

NO M T *

AVAILASLC

of discontinuities under pressure) withdepth, and with confining pressure inlaboratory experiments, but few data exist.The permeability-depth relation is not yetwell defined for variations of rock types.Finally, literature concerning groundwaterchemistry changes and subsequent rockstability, particularly as it relates totemperature changes, is also inadequate.

ROCK MASS PERMEABILITV (M/SEC)

(f9 lO"8 10'7l(f9 lO"8 10'7 10"6 10"6 10"5

Chemical changes in groundwater, how-ever, are naturally more dependent on rocktype than on groundwater fluctuations.From general geochemical knowledge, car-bonate rock types (limestones, dolomites,and marbles) are more soluble than eitherigneous, metamorphic, or nonchemicalsedimentary rocks, primarily becausecarbonates dissolve more readily thansilicates (especially in elevated temper-ate environments). Under normal condi-tions, groundwater in noncarbonate ter-ranes generally has a low ion concentra-tion, unless other mineralization ispresent, such as ore sulfides. Ground-water chemistry is highly dependent onclimate, too, particularly temperature.For this reason, rock and groundwatertemperature changes due to CAES cavernoperation may be significant in ground-water chemistry changes.

Summary Remarks. Actual hydraulic datafor rocks of low permeability are notabundant in the literature. Most exist-ing data are for clastic and permeablechemical sedimentary rocks. Also dataconcerning porosity and permeability asaffected by cyclic temperature changesare very inadequate.

It is also evident that the over-riding permeability factors in rocksgenerally considered favorable for CAESare the dimensions and frequency of dis-continuities of all types. As indicatedin Figs. 2 and 3, the literature suggestsgeneral reductions of permeability (dueto a combination of frequency and closing

1

4• til

f'\^-EMPIRICAL FUNCTION

(AFTER BURGESS)

50 •

=? 100 -

150 •

-O BEST FIT EMPIRICAL FUNCTION AFTER BURGESS• KTH FIG. 3 FROM GUSTAFSSON

— • SNOWK-MT 1- 6 >•*"<<>— •-* CARLSSON i OLSSEN FIG. 3

° CARLSSON t OLSSEN FI6. 10* CARLSSON i OLSSEN FIG. 7

SGU KRAKEMALA Kl

FIG. 2. FIELD DETER/1INED ROCK MASS PERMEABILITY ASA FUNCTION OF DEPTH FOR IGNEOUS AND METAMORPHICROCK MASSES (AFTER STILLE, ET. AL.12)

1000

100

10

-LIHESTONE(OHLE, 1951)

WESTERLY

GRANITE•WATER:AARGON: 15'ARGON: 10•ARGON: 5

(AFTER BRACE,

ET.AL.15)

FIG. 3.

0 100 200 300 400

EFFECTIVE CONFINING PRESSURE (IIPA)LABORATORY DETERMINATION OF PERMEABILITYAS A FUNCTION OF CONFINING PRESSURE

368

GEOTHERMAL GRADIENT AMD HEAT FLOW

The possible major sources of heat inthe upper few km of the earth's crustinclude: (1) the outward flow of heat fromthe central core of the earth; (2) thepresence of cooling magmas; (3) the dis-integration of radioactive elements; and,(4) subcrustal convection currents(Levorsen11). The discussion of thesesources and the concept of global heatflow is most relevant to the field of platetectonics and is not particularly signifi-cant in the CAES situation. What is mostsignificant here is simply that rocktemperature does increase with depth, andthat the temperatures likely to beencountered at the relatively shallowdepths of a CAES cavern (600 to 700 m)are dependent on such factors as atmospher-ic temperature changes and groundwatercirculation, and also on material thermalconductivity and local geologic structure(Levorsen1l). Elevated temperature at depthmay be significant in the initial miningphases, where it could pose problems incooling and ventilation. However, it isdoubtful, considering available published -jdata and again, depending greatly on ""specific site and cavern depth, that the _,geothermal gradient of an area could pose gserious difficulties in the development Qand operation of mined CAES caverns.

Published averages for geothermalgradient include 0.04°C/m (Leverson11) and0.03°C/m (Jackson6), or temperatureincreases in one km of approximately 40°Cand 30°C, respectively. Such temperatureincreases probably would not be significant,considering that CAES operation may causecavern temperatures to increase to as muchas several hundred °C. A fair amount ofsuch data exists for specific localitiesand for larger regions; however, the natureof the factors are site specific and couldbe obtained for a particular area, underusual circumstances, using presentlyavailable technology.

IN SITU STRESS STATE

The in situ stress state in a rockmass is defined as that state of stresswhich exists prior to disturbance of therock by excavation,. This natural state ofstress, in conjunction with the strengthand structural characteristics of the rock,are important in determining the geometricalshape and dimensions of an undergroundexcavation. Generally, the in situvertical stress is taken to be that in-

duced by the weight of the overburden,but may be perturbed by regional or localtectonic features. As indicated in Fig.4, the average vertical stress to a depthof 3 km is of the order of 0.025 MPa/m(after Haimson11*) to 0.027 MPa/m (afterbrown and Hoek 1 5), which corresponds toan average bulk density of 2,550 to2,755 kg/m . The approximate limitsgiven in Fig. 4, as deduced from thecompilation of worldwide published databy Brown and Hoek 1 5, are indicative ofvariations in the overburden density fordifferent rock types and of the influenceof tectonic features.

VERTICAL IN SITU STRESS, O"z

10 20 30 HO 50 60 70

500 -

1000 -

1500

2000

2500

3000

\ \c \ ^

<T2 = 0.025 Z - ^(AFTER HAIMSON14)

1 1 1

APPROX. LIMITS -^ - (AFTER WORLD-WIDE

^ DATA COMPILATIONBY BROWN S H0EK15)

crz = 0.027 Z -' ' • • < ' (AFTER BROWN &

S '•-.. H0EK15)

? < ? \ \ •••...

. • IGNEOUS 7 U.S.A, 8 CANADA '•••. \ > .D HETAM0RPH1C > (AFTER COMPILATION - - s \ ^° SEDIMENTARY S BY BROWN 8 HOEKlS) --..\

0 * • .

1 1 1 1 1 1

FI3. i). VERTICAL IN SITU STRESS AS A FUNCTION OF DEPTH.

The ratio of the in situ horizontalstress to the in situ vertical stress isknown as the coefficient of lateral earthstress. Fig. 5 illustrates that thiscoefficient varies from less than one togreater than three at depths of a fewhundred meters. For depths of severalkilometers, the coefficient ranges fromabout one-third to one. As indicated bythe limiting curves due to Haimson11*, asobtained from hydrofracturing data, thetwo orthogonal stresses in the horizontalplane are not necessarily equal. Thecurves by Brown and Hoek'15 are upper andlower "average" limits as deduced from aworldwide compilation of published data.We note that the in situ principal stressesmay not be aligned with the vertical and

369

horizontal directions due to the influenceof tectonic features.

COEFFICIENT OF LATERAL EARTH STRESS. K,, - O " H 0 R / c r z

, 0 1 2 ^ 3

METANORPHIC > (AFTER COHPILATSEDIMENTARY S BY BROWN 8 HOEKi ' i -^ i I3000

FIS. 5. COEFFICIENT OF LATERAL EARTH STRESS AS A FUNCTION OF DEPTH

THERMAL/MECHANICAL ROCK PROPERTIES

On the basis of many articles in thepublished literature, a brief compilationof thermal/mechanical rock properties isgiven in Table 2. The properties arelisted 'in terms of ranges of values, andare indicative of the relative strengthand thermal characteristics of the variousgeneric rock types, without differentiationfor competency. In general, the data forjointed rock under conditions, of elevatedtemperature and confinement stress areinadequate for cavern stability evaluations!.

M K K OF TKHmCHUICN. HOCK NOPRTIES

mx WE

IOKVSinmivt

xatmnnwtvt

stimmungwicu.

saimnuiCUItIC

KMMMCnu«uICTMOHIC• M T H M 1 U

NUmn

m z

• IK

Mti

M I

SB

TOBltfmoon»t)

1S.0

} .<

M.S

1.5

lD.1

n.i22.1

1.1

U. I

g.c

11.0

WCWFIPCICOIVCS*

sireSTWWH«f»>

1n

f . j

JSJ

17.2

M l

a.«wo

22.7

«5

t.«

«K

K.I

rocaisOF

usncm

I M

1.5

IT?

••

>4119

7.J

IDS

i n

» • «

•oissw-sMTI0

n.no.os

Q.«

0.0>

<f.V

<»-W

0.*g.ua.*o.»a.uo.«

WMLanucT-IHTT

i.n?

0.76

2.71

1.17

<.so

l.»J.H5.7>

l.«7

J.(B

!.tl

0.75 .

now.

ao-V/s1.751

0.X1

i . m

O.(K

1.S71

»,523

a.ml.mi

n.«

' "O J M

COEF. OFLISMT1CHW.3TWSindO-'/1*)

».n«.s

u.o

$i0

. . f i , ;

15,0

«

5.0

Ji.a

1.5

SKCIFICKAT

not7»

ffi

« t

1IM79C

1050

re

i n n

•57

The survey of thermal fatigue, spall-ing, and cracking literature for rock hasnot yielded abundant applicable informat-ion, and much of that available (partic-ularly regarding thermal fatigue) is ofa qualitative nature. However, enoughstudy has been conducted on the threeprocesses to indicate that they wouldmerit consideration in the design of hardrock CAES caverns.

Rock thermal fatigue is essentiallya type of weathering process. The effectof cyclic heating of a rock, with or with-out the effects of cyclic moisture con-ditions, generally causes some degree ofrock disintegration, particularly leachingof compounds from the rock and subsequentweight loss. Tne significance of thisphenomena is evident when the cyclictemperature and moisture conditions of aCAES cavern are considered. Likewise,the processes of thermal spalling andthermal cracking are also significant, andtheir disintegration effects (along withthose resulting from fatigue) may occursimultaneously. Thermal spalling and/orcracking of a rock surface may occur whenheating induces thermoelastic stresseswhich exceed the fracture strength of therock. Spalls are relatively thin and areusually curved pieces of rock broken offfrom the rock surface during or afterheating, and may be several meters inlength or microscopic in size. Thermalcracking results in the production offractures in the rock, usually microscopic,as indicated in Fig. 6.

The causes and degree of suscept-ability to fatigue, spalling, and crackingare many. They includte rock type (com-position and fabric), maximum temperaturechange, rate of temperature change, pres-ence of fluids, composition of fluids, andfrequency of cyclic temperature and mois-ture changes. Thermal fatigue effects ina CAES cavern could conceivably causesevere leaching of compounds which couldalter groundwater chemistry, make cavernwater unsuitable for effective equipmentoperation, and cause elastic and strengthproperties of the rock mass to change.Major spalling effects could also producegross cavern instability, while micro-spalling could produce significant quan-tities of particulate matter in the cavernair which could have destructive effects onturbine operation. Thermal cracking couldresult in irreversible changes in rockelastic modulus, fracture strength,porosity, permeability, fracture toughness,

370

and thermal expansion and diffusivity(Friedman16).

160

100100 300 500 700

TEMPERATURE (°C)

• SLOWLY CYCLED, THIN SECTION

O QUENCHED, THIN SECTION

Q SLOWLY CYCLED,, POLISHED

SURFACES

FIG 6. MICROFRACTURE INDEX AS AFUNCTION OF TEMPERATURELEVEL (AFTER FRIEDMAN16).

More study is needed on thermaleffects in rocks, particularly with regardto the contribution of each factor listedabove to total rock mass instability.Some general conclusions drawn from presentliterature are that thermal effects combin-ed with moisture effects generally aremore degenerative than thermal effectsalone, and that abundant microfissurizationpromotes more disintegration (Aires-Barros,et al17) as does higher porosity andhigher permeability.

In addition, carbonates, such aslimestones and dolomites may be much moresusceptible to thermal and moisture effectsthan igneous rocks (Mailhe, et al"*).Igneous rocks with abundant mica mineralsand calcium plagioclase feldspar may bemore susceptible than other igneous rocks(Aires-Barros, et al 1 7). At present,however, the literature on the cyclicthermal and moisture effects of intactrock are only marginally adequate and thaton jointed rock totally inadequate. There-fore, more general laboratory testing ofsuch effects, and also testing of specific

rock types from possible CAES host rockmass, must be undertaken.

STABILITY CRITERIA FOR CAES CAVERNS

DEFINITION OF INSTABILITY

The notion of instability of CAEScaverns may be categorized as follows:

(1) Global Rock Instability.Identified by massive roof falls,wall slabbing, and floor heave,leading to the loss of structuralintegrity of the cavern and/orits entrance.

(2) Local Rock Instability.Identified by localized thermo-mechanical spalling and thermo-chemical disintegration of therock over the cavern periphery,leading to participate transportduring compressed air withdrawalto the turbine system.

(3) Global Air Leakage Instability.Identified by unacceptable airleakage from the cavern duringcompressed air injection andstorage (due to greatly enhancedhydraulic conductivity as aresult of induced fracturing orjoint dilation).

In practice, we may define the timeperiods of instability concern for a systemof CAES caverns as: (1) Excavation, (2)Operation, and (3) Decommissioning. Thenotion of global rock instability appliesto the excavation and decommissioningperiods, while all three instabilityconcerns apply to the operational period.Clearly, the concept of stability criteriainvolves the specification of limits on thethermal/rock mechanics/hydrological behaviorof the rock mass, wherein instability isprevented when the limits are not exceeded.

EXCAVATION STABILITY

We shall assume that the geotechnicaland thermal/mechanical properties of arock mass for a potential CAES site can beproperly characterized and quantitativelydefined. For a choice of cavern depth,shape, dimensions, and spacing, we maycompute, by say the finite-element method,the state of stress in the rock mass duringa simulated excavation. The appropriatecriterion for evaluation of the globalstability of the excavation is the Mohr-Coulomb failure condition. This criterion,

371

which mathematically relates the tensileand compressive strengths of a rock to astate of applied stress, permits evaluationof the potential for incipient rock failure.Specifically, if the state of stress aroundthe excavation violates the Mohr-Coulombcriterion, failure of the rock mass isindicated. The criterion is applicable toboth the intact rock and the joints, withappropriate characterization of thestrength properties in each case, and toa failed rock mass in the sense ofresidual strength. By altering the caverngeometry and dimensions, and the sequenceof excavation, the stability of the cavernmay be effectively optimized for a givenrock mass and state of in situ stress.

An actual measure of instabilitymust be quantified in terms of loss ofcavern serviceability. Regions of failurewithin the rock mass around the peripheryof the cavern, as indicated by finite-element modeling, do not necessarily implyglobal instability if the regions arelocalized and reasonably disconnected, orcan be "hardened" by use of artificialsupport. Thus, a certain degree of"contained" rock failure may be acceptablein the sense that the future serviceabilityof the cavern is not impaired in adetrimental fashion.

OPERATIONAL STABILITY

During the operation of a CAEScavern, the rock is subjected to cyclicvariations in applied pressure andtemperature. These conditions induceadditional stress perturbations in thesurrounding mass, and the global insta-bility of the cavern must be evaluated byuse of the Mohr-Coulomb condition of rockfailure with temperature-dependent pro-perties. However, in this situation,the strength of the rock will probably beprogressively reduced with the number ofloading cycles.

The hydraulic conductivity of therock, which is a function of stress andtemperature, will be perturbed by theinitial excavation, and subsequentlyperturbed by the cyclic pressure andtemperature loadings. Failure of theintact rock and/or joints will also giverise to conductivity perturbations. Fromthe viewpoint of global air instability,the criterion will be related to the lossof air from the cavern in an economic oroperational sense.

The local rock instability of thecavern periphery is related to the spalllngand microfracturing characteristics of therock under cyclic pressure/temperatureloading and air/water intera. ion. Thelimit of acceptable rock disintegrationmust be established from the viewpoint ofallowable particulate transport to theturbine system during compressed airwithdrawal.

DECOMMISSIONING STABILITY

After cessation of the operationalphase of a CAES cavern, considerationmust be given to the eventual collapse ofthe cavity, leading to possible surfacesubsidence. It would appear that an appro-priate evaluation of "long-term" stabilitywould involve consideration of a creeprupture criterion in conjunction with thestress state around the cavern.

DESIGN CRITERIA FOR CAES CAVERNS

Due to operational and/or economicconsiderations, the choice of a CAEScavern site may be somewhat restricted.Apart from some lateral variation inselection in a given area, the depth androck quality may be the only significantvariables of choice. An important real-ization is that in practice it may benecessary to encompass the design of aparticular CAES cavern system within thethermal/rock mechanics/hydrologicalcapabilities of a given underground rockmass. As a consequence, "hardening" ofthe caverns by artificial support meansmay be required in order to obtain atechnically and economically feasible CAESsystem. In fact, the initial start-up ofthe operational phase may require a specialsequence of compressed air cycling in orderto promote hardening or shakedown of therock by psuedo-plastic deformation withinthe limits of global stability.

The design criteria for a cavern mustbe established from the viewpoint of (1)the type of CAES system, (2) the desiredair volume and pressure, and (3) thethermal/rock mechanics/hydrologicalconstraints appropriate to lue rock mass.The constraints must be utilized in thedetermination of optimum cavern shape,dimensions and spacing, and excavationsequence.

372

ACKNOWLEDGMENTS

The authors are indebted to Mr. HenryWaldman for the computer processing of therock properties data, and to Mr. Joe L.Ratigan and Dr. Arlo F. Fossum for theirconstructive suggestions during the courseof the study.

LIST OF REFERENCES

(1) Milne, I. A., Giramonti, A. J., andLessard, R. D.: "Compressed Air Storagein Hard Rock for Use in Power Application",Rockstore 77, Stockholm (1977), Vol. 2,pp. 199-205.

(2) Willett, D. C. & Lawrence, J. D.:"The Design of Reservoir Caverns forUnderground Pumped Storage", Rockstore 77,Stockholm (1977), Vol. 2, pp. 145-148.

(3) "Preliminary Feasibility Evaluationof Compressed Air Storage Power Systems",Final Technical Report (R76-952161-5) toERDA & NSF by United Technologies ResearchCenter (1976); Vols. I & II.

(4) Mailhe, P., Comes, G., Perami, R.:"Geological and Geotechnical Process forthe Siting of a Hydropneumatic PumpedStorage Plant in Brittany (France)",Rockstore 77, Stockholm (1977, Vol. 2,pp. 495-500.

(5) Wittke, W., Pierau, B., & Schetelig,K.: "Planning of a Compressed-Air Pumped-Storage Scheme at Vianden/Luxembourg",Rockstore 77, Stockholm (1977), Vol. 2,pp. 149-158.

(6) Jackson, K. C.: Textbook of Lithology,McGraw-Hill, Inc. (1970).

(7) Travis, R. B.: "Classification ofRocks", Quart. Colo. Sch. of Mines, V. 50(1955).

(8) Hamblin, W. K. and Howard, J. D.:Exercises in Physical Geology, BurgessPub. Co. (1975).

(9) McKenzie, G. C , Pettyjohn, W. A.,and Utgard, R. 0.: Investigations inEnvironmental Geoscience. Burgess Pub. Co.(1975).

(10) Walia, M. and McCreath, D. R.s"Siting Potential for Compressed Air andUnderground Pumped Hydro Energy StorageFacilities in the United States",Rockstore 77, (1977), V. 1, pp. 117-123.

(11) Levorsen, A. I.: Geology ofPetroleum, W. H. Freeman and Co. (1967).

(12) Stille, H., Burgess, A., andLindblom, U. E.: "Groundwater MovementsAround a Repository", KBS Tech. Rept. No.54:01 (1977).

(13) Brace, W. F., Walsh, J. B., andFrangos, W. T.: Permeability of Graniteunder High Pressure", J. Geophys. Res.,V. 73 (1968), pp. 2225-2236.

(14) Haimson, B. C : "The Hydro fracturingStress Measuring Method and Recent FieldResults", Int. J. Rock Mech. Min. Sci.,Vol. 15 (1978), pp. 167-168.

(15) Brown, E. T. and Hoek, E.: "Trendsin Relationships between Measured In-SituStresses and Depth", Int. J. Rock Mech.Min. Sci., Vol. 15 (1978), pp. 211-215.

(16) Friedman, M.: "Thermal Cracks inUnconfined Sioux Quartzite", Preprint -Proc. 19th Symp. on Rock Mechanics(1978), pp. 423-430.

(17) Aires-Barros, L., Graca, R. C , andBelez, A.: "Dry and Wet Laboratory Testsand Thermal Fatigue of Rocks", Eng. Geol.,V. 9 (1975), pp. 249-265.

373

PROJECT SUMMARY

Project T i t l e : Long-Term S t a b i l i t y of Compressed A i r Energy Storage Caverns

Pr incipal Invest igator : R. L. ThornsJoseph D. Mart inez, Coprincipal Invest igator

Organization: I n s t i t u t e for Environmental StudiesAtkinson H a l l , Louisiana State Universi tyBaton Rouge, LA 70803(504) 388-8521

Project Goals: To conduct a s ta te -o f - t he -a r t survey on storage caverns in s a l tdomes r e l a t i v e to compressed a i r energy storage (CAES) andto formulate prel iminary lonq-term s t a b i l i t y c r i t e r i a f o r s a l tdome CAES caverns.

Project Status: Caverns in s a l t domes o f fe r perhaps the most promising type ofgeostorage space fo r compressed a i r energy. Storage of hydrocarbonshas been pract iced i n s a l t domes of the U. S. Gulf Coast regionf o r approximately twenty-seven years; however, s ta te -o f - t he -a r ttechniques r e l a t i v e to geostorage have been considered la rge lyproprietary until only recently.

Many early storage caverns were not designed specifically forstorage purposes, but were a result of primary solution mining(brining) operations. L i t t le effort was made to monitor storagecaverns except when serious problems obviously had alreadydeveloped, e.g., collapse into caverns that had penetratedthrough the dome surface by uncontrolled dissolution.

Generally storage of hydrocarbons, particularly in l iquid form,in salt dome caverns has been highly successful. However,l i t t l e is known about the long-term effects of cyclic loadson salt rocks.

Compressed air energy storage (CAES) wil l involve continuousdaily cycling of pressure, temperature, and humidity. Possibledeleterious effects due to cyclic loadings must be controlledso that the CAES caverns remain stable and functional overtime periods of approximately thir ty- f ive years.

Long-term stabi l i ty cri teria are presented for CAES caverns insalt domes. The cr i ter ia , which are relatively general innature because of the unknown effects of cyclic loadings, arebased on a review of relevant technical l iterature and informationfrom persons knowledgeable about storage in salt dome caverns.

Finally, a methodology is presented for development and imple-mentation of quantitative, long-term criteria applicable to CAEScaverns in site specific salt domes.

Contract Number: Special Agreement B-54804-A-L

Contract Period: Oct. 1977 - Aug. 1978

Funding Level: $17,000

Funding Source: Battelle Pacific Northwest Laboratories

375

PRELIMINARY LONG-TERM STABILITY CRITERIA FORCOMPRESSED AIR ENERGY STORAGE CAVERNS IN SALT DOMES

Robert L. ThornsInstitute for Environmental Studies

Louisiana State University, Baton Rouge, LA 70803

ABSTRACT

Caverns in salt domes offer perhaps the most attractive type of geostorage for com-pressed air energy. Storage of hydrocarbons in salt dome caverns has been practiced inthe U.S. Gulf Coast region for approximately twenty-seven years, however this kind ofstorage has involved mainly liquids and relatively slowly varying pressure loadings.Compressed air energy storage (CAES) in salt caverns will involve continuous daily cy-cling of pressure, temperature, and humidity. Possible deleterious effects due to cy-clic lo. dings must be controlled such that the CAES caverns remain stable and functionalover time periods of approximately thirty-five years. Long-term stability criteria arediscussed relative to CAES caverns in salt domes. The criteria are based on a review ofthe technical literature and interviews with persons knowledgeable about storage in saltdomes. The methodology for development and implementati-on of quantitative criteria spe-cific to any particular potential CAES dome also is presented.

INTRODUCTION

This, study is directed to the long-term stability of compressed air energystorage (CAES) caverns in salt domes orsalt anticlines. The Gulf Coast Basin,as depicted in Fig. 1., is the only con-firmed salt dome region in the U.S.A.;however, the salt anticlines of the Para-dox Basin also would possess many geo-logic features similar to domes.1

Ml•>!>•« ItIM JMHM 111 I H I I H I , IfTI

FIG I SALT DEPOSITS IN THE UNITEO STATES

Projections of a desirable useful"life" for CAES facilities include a timeperiod of around 35 years of cyclic oper-ation. Thus the air storage caverns, whichare an essential and integral componentof a CAES plant, should be designed andoperated so as to perform satisfactorilyover the intended life of the overall fa-cility. It follows that the long-

tenn "stability" of air storage cavernsmust be considered as a primary concernin projecting the satisfactory operationof CAES facilities.2

As used in this report, "stability"of a storage cavern implies the extent towhich an acceptable amount of cavernstorage volume can be utilized with rou-tine maintenance for a specified time in-terval , e.g., 35 years. In this context,cavern stability is relative to bothplanned utilization and time interval ofoperation.

Although the storage of liquid hydro-carbons has been practiced in United States(U.S.) Gulf Coast salt domes for aroundtwenty-seven years, the cyclic pressureand temperature variations which are anintegral part of CAES operations intro-duce new considerations relative to thelong-term stability of the associatedstorage caverns. Fig. 2 is a schematic ofa "leached-out" storage cavern in a saltdome along with symbols for significantspatial dimensions. Fig. 3 depicts a typ-ical weekly cycle for pressure and temper-ature within a CAES air reservoir cavern.

Many hydrocarbon storage caverns nowin operation in salt domes are a secon-dary benefit of an original brine solu-tion-mining operation and were not de-signed with stability as a primary con-cern. More recently however, energy util-ization concerns have gained such promi-nence that optimizing cavern designs in

376 "383

FIG. 2. SCHEMATIC OFCAES CAVERN IN SALT DOME

• 0

TO

• 0

90

40

90

10

10

• \J-A A'

\

T"H"1"''

A

/ \/ \PCI *

(MM I

/ » /

/

/ /

'—V

WON TOE S ' WED THUR Ft I SAT SUN

FIG. 3. TYPICAL CAES CAVERN CYCLE

salt .domes for storage purposes has be-come a primary, rather than a secondary,consideration. Thus a great deal of in-terest now exists in further developingthe technology currently associated withdesign of salt dome storage caverns, andextending this technology to implementrelatively new storage concepts, such ascompressed air energy storage.3

GENERAL STABILITY CRITERIA

General criteria for long-term sta-bility of CAES caverns in salt structuresmust account for a number of factors. Theprincipal factors are listed in Fig. 4.Criteria related to these factors followin a summary format.

The site-specific utilization historyof potential air storage salt structuresshould be researched to insure man-madeeffects will not threaten the integrity ofthe containing salt. For example, theremust be an absence of man-made hydrofrac-turing connections or solution channels

• Site Specific Utilization Historyof Domes

• Geology of Salt Domes IncludingCaprock and Neighboring Formations

• Material Properties of Salt andAdjoining Materials

• Configurations of Dome andCAES Caverns

• Loading Cycles for CAES Operations

Fig. 4. Principal Factors AffectingLong-Term Stability of CAESCaverns in Salt Domes

through the salt that could open underCAES operations; and a good record ofability to maintain well casings throughcaprock (if any) into the salt. Any pre-vious brining and/or sulphur mining opera-tions in the caprock should be checked toinsure that associated effects such assurface subsidence have essentially ceasedand will not otherwise significantlyaffect a CAES facility.

The geology of potential air-reser-voir salt domes, including the caprock andneighboring formations, should be studiedand reported in the early stages of plan-ning for CAES facilities. For example,the presence of liquids or gases in thesalt stock itself may not disqualify a po-tential storage dome. However, the pres-ence of a highly permeable anhydrite sandat the caprock-salt contact could permitcirculation of brines with H2S gas thatcould cause rapid corrosion of well cas-ings, unless preventive measures weretaken by using appropriate cement aroundthe casings. Leaks in casings could leadto an abrupt depressurization of CAEScaverns with resulting damage to the sur-rounding salt ("wall slabbing" and "rooffalls"). Furthermore, the geology of thesalt, caprock, and adjoining formationsis directly related to the mechanicalproperties of the associated materials.

Salt domes display anomalous zones inwhich weaker salt, gas pockets, and hydro-carbons are present in the salt stock.Megascopic features associated with thesezones would directly affect CAES caverns.The geology of salt domes in the U.S. GulfCoast generally varies in character, andshould not be assumed to be the same evenfor domes separated only by a few miles.

377

The state of tectonic stress withinthe salt stock of U.S. Gulf Coast domesis usually assumed to be hydrostatic incharacter. However, this assumptionshould be checked; and, it is suspect inthe Paradox Basin salt anticlines. Inboth cases, the study of tectonic stres-ses within the salt stock deserves care-ful attention, because creep closure ratesof caverns in evaporitic rock are stronglyaffected by the state of initial geostaticstress.

An ideal CAES dome would include thefollowing geological features: (1) mass-ive anhydrite caprock (if any) free ofvugs and lost circulation zones, (2) solidcaprock-salt contact, (3) relatively homo-geneous salt of uniform character and freeof liquid and/or gas.

The material properties of the saltand adjoining geologic units, e.g., cap-rock, should be established so that sta-bility analyses will have a basis for pre-cluding cavern failure under environmentsassociated with cyclic CAES operations.Currently there does not exist an adequatematerial properties (mechanical, thermal,and chemical) data base for predicting thelong-term response of rock salt subjectedto cyclic loadings of pressure, tempera-ture, and humidity typical of a CAESfacility as depicted in Fig. 3. Somecreep data exists for laboratory specimensof rock salt subjected to elevated temper-atures and varying load paths. Also, anumber of associated material "laws" havebeen proposed for numerical modellingapplications. However, for CAES opera-tions, the important long-term effects ofcreep-rupture and low-frequency fatigueof rock salt in a cyclic environment arelargely unexplored phenomena.

Deterioration of salt around cavernsalso can be anticipated to depend uponsite-specific material responses to arapid drop of cavern pressure. For ex-ample, a pressure drop from 2,219 psi(15,300 kPa) to 114 psi (786 kPa) over158 hours apparently caused a significantroof fall in an experimental gas storagecavern in a German salt dome.1*

The topic of material properties ofrock salt and the relationship to cavernstability is relatively complicated andsomewhat lengthy. Details on this topicare beyond the scope of this review re-port. A later section outlines a pro-

gram methodology for developing data spe-cific to potential CAES operations.

The configuration of a potentialCAES dome must be capable of accommodat-ing a system of caverns suitable for acompressed air reservoir. Depths of CAEScaverns apparently can ranqe from approx-imately 2000 to 5000 ft .(609.6-1524.0 rO .Minimum dimensions for salt "wall" thick-ness between caverns and "roof" thicknessover caverns (as depicted in Fig. 2.)will require analyses utilizing site-specific material properties responseassociated with CAES loading environments.Examples of minimum wall thickness be-tween storage caverns include 100 ft(30.5m) for pseudostatic hydrocarbonstorage5 and 7?1.8 ft (220m) for the twoCAES caverns at Huntorf.6 Examples ofcavern roof thickness include 200 ft(61.Om) for hydrocarbon storage and ap-proximately 328 ft (100m) at Huntorf.These recommended and existing dimensionsapparently are based mainly on a greatamount of first hand experience in hydro-carbon storage and brining operations inthe U.S. Gulf Coast and in West Germanyrespectively. In addition, the WestGermany design incorporated data frommodel testing of rock salt.

Cavern shape, size, spacing, andnumber obviously can all vary dependingupon the requirements of any particularCAES facility operation. A systems opti-mization analysis should be used as abasis for establishing these parameters.Intercavern air flow effects, e.g., fric-tion and thermal losses, along with totalair volume flow requirements would neces-sarily be coupled with cavern system sta-bility requirements in such an analysis.

The general CAES criteria for domeconfiguration requires that an adequatevolume of salt must exist at a workableair reservoir depth so as to provide asurrounding envelope of salt sufficientto prevent the formation of connectionsfrom the reservoir to the dome surface.Such connections could be of a long-termprogressive character, and would be dueto a combination of creep-rupture, mech-anical and thermal fatigue, air penetra-tion, and hydrofracturing effects.

The loading cycles for CAES operationsmust range within bounds obtained from anengineering analysis and a verified database. A verified data base necessarilywill include field experiments within a

378

potential CAES salt structure with ob-served test results correlating with pre-dicted results from an appropriate analy-sis. The analysis should be based on acombination of laboratory and/or in-situtests and numerical modelling. Sincecavern pressure is the loading parametermost easily controlled under actual oper-ating conditions, it should receive maxi-mum attention in tests and numerical mod-elling simulating prototype CAES loadingenvironments. However, at the same timethe dependent and coupled loading param-eters of temperature and humidity alsomust be properly represented. It shouldbe noted that most of the factors affect-ing long-term cavern stability under CAESoperations could be strongly coupled.

The planned cavern pressure cycle atthe Huntorf CAES facility includes charg-ing for eight hours to 70 atm, then dis-charging for two hours to 50 atm. An-other CAES study includes an example cyclewith 24 hr pressure variations of approx-imately 13 atm. The daily cycle is em-bedded in a weekly cycle with approximatemaximum and minimum values of 78 and 46atm respectively.

In addition to effects of CAES load-ing cycles on the rock salt around air-reservoir caverns, effects on well cas-ings and "casing seats" into the saltalso must be considered. Such effectsare probably better understood (thaneffects on salt) by the storage industryoperating in Gulf Coast salt domes. How-ever, the progressive deterioration ofcasing-seat grouts should be exploredunder cyclic pressure and thermal load-ings. The Huntorf CAES plant utilizedan inner "free" tube to transmit air, thusuncoupling the outer well casing from ther-mal or pressure "shock" loading effects.

In summary of this section, generalcriteria for long-term stability of CAEScaverns in salt domes have been reported;however, a data base specific to thecyclic loading character of CAES opera-tions must be developed prior to specify-ing quantitative criteria with any degreeof confidence. As noted previously, amethodology is proposed later within thisreport for establishing site-specific andquantitative criteria.

STATE OF THE ART SURVEY;PRINCIPAL FINDINGS

A principal part of this report inclu-ded the findings from a state-of-the art

survey relative to long-term stability ofcaverns in salt domes. Attention was di-rected to information sources on aspectsof stability which were particularly rele-vant to potential CAES operations in saltdomes. Thus a literature survey was under-taken, and also discussions were held withpersons experienced in the geostorage in-dustry involving salt domes. The princi-pal findings are reported in this sectionin summary format.

Field data associated with storage ofnatural gas in salt dome caverns currentlyis most directly applicable to long-termstability studies of CAES caverns. Reviewof th<= geostorage related literature anddiscussions with operators of a variety ofstorage operations (hydrocarbons and/orbrine; in salt domes have revealed signifi-cant amounts of data. However, the datatypically is fragmented and not directlyapplicable to stability criteria for CAESoperations. The most similar field opera-tions to CAES exists currently in the nat-ural gas storage industry. Experiences instorage of natural gas in salt dome cav-erns should be carefully evaluated infuture planning of CAES operations.

CAES storage raises more concerns withstability problems than liquid hydrocarbonstorage, solution mining, or natural gasstorage currently operational in salt domecaverns. Long term cyclic loadings ofpressure and temperature could have moredeleterious effects on cavern stabilitythan current psuedo steady-state storageoperations in salt dome caverns. However,the extent of the possible increase indeleterious effects is currently unknown.Roof falls reportedly have occurred inconjunction with rapid pressure drops insalt dome natural gas storage caverns,which implies rates of pressure dropsassociated with daily CAES cycles couldaffect cavern stability. Surface contactby brine, as contrasted to gas, is be-lieved by some investigators to increasethe plasticity characteristics of rocksalt. If this is true, spalling of thesurfaces of CAES caverns will be moresevere than for brine filled cavernsunder similar pressure and temperatureloading cycles.

A number of workers with field exper-ience in the domai storage and miningindustry are dubious about the signifi-cance of conventional rock mechanics lab-oratory testing programs relative to pre-dictions of salt dome cavern behavior.

379

They share a strong conviction that onlyfield tests and related experiences underactual site-specific conditions can berelied upon at the present time to yieldreliable results for predicting cavernbehavior.

Test caverns have been employed inWest Germany to obtain field data forperformance of natural gas storage cav-erns. Apparently no such scaled down cav-erns with tests have been utilized thusfar in the United States (U.S.). Someborehole test procedures have been pro-posed both in the U.S. and abroad as anintermediate method for obtaining data onin-situ rock salt performance that couldbe related to cavern stability.

Laboratory testing programs fordetermining rock salt behavior have beenconducted mainly with compression tri-axial or uniaxial tests. However, moreinfrequently used triaxial extension testswere found to be of primary value in atleast one major study of creep closure ofsalt dome caverns.7 Triaxial extensiontests permit a more comprehensive evalua-tion of creep-rupture, which is a signifi-cant effect in long-term stability ofopenings in rock salt.

Crystal sizes (around 0.25 in (0.64mm) "diameter") in polycrystaline rocksalt are relatively large, thus scaleeffects are probable in laboratory testswith relatively small specimens. Sometest data imply dependence on size of cyl-indrical rock salt specimens falls offsharply for 4 in. diameter (and above)cylinders with standard length to diameterratios of 2 or slightly larger.

Numerical modelling of rock salt be-havior with the finite element method hasbeen in use for some time. Numerical mod-elling can be employed as a powerful aidin developing and verifying stability cri-teria during the laboratory and in situtesting phases as well as during the pilotcavern (if used) and field operationalphases of prototype CAES air-reservoirs.

Empirical methods combined with fieldexperience have been used successfully incavern stability studies." The CAES cav-erns at Huntorf, West Germany, were de-signed primarily on this basis.

Stability criteria ultimately must beverified by field data from pilot and/oroperational prototype CAES caverns. Rel-

atively sophisticated monitoring equipmentnow exists for such purposes, e.g., alaser ranging device that can monitordistances within a gas-filled salt cavernunder pressure with an independent sensi-tivity of ± 10 cm.9 Other sensitive mon-itoring equipment also may be useful forrecording cavern performance, e.g., micro-seismic monitoring systems and tiltmeters.

The study of fundamental rock saltmechanics for many kinds of storage indomes is still in a relatively early stageThe more complicated loading conditionsassociated with CAES caverns will entail aconsiderable amount of careful investiga-tion before full confidence can be placedin associated specific and detailed quanti-tative long-term stability criteria.

Concerns with compensated (constantpressure), as contrasted to noncompensated(constant volume), CAES caverns would in-""elude possible (1) erosion of salt in theair-reservoir "walls' and connecting open-ings due to liquid flow, and (2) abruptloss of cavern pressure due to a champagne-effect blowout of the compensating water(brine) "leg". However, compensated cav-erns would possess an apparent advantageof low amplitude cyclic pressure loadingof surrounding rock salt.

PRINCIPAL RECOMMENDATIONS

A relatively brief set of summaryprincipal recommendations are presentedin this section. They are based on thefindings from the state-of-the-art surveyand additional study relative to require-ments for long-term stability of CAEScaverns in salt domes.

Long-term stability criteria forCAES caverns should be concerned witheffecting acceptable limits on time de-pendent creep closure, creep rupture(spalling or slabbing), and superimposeddeleterious cyclic loading effect. How-ever, short-term effects that couldthreaten the overall integrity of the air-reservoir also must be limited, e.g.,hydrofracturing, well casing failure, andcavern roof and wall "falls" due toabrupt depressurization.

Specific long-term stability cri-teria for CAES caverns should be devel-oped in a closely coordinated and bal-anced program utilizing complementary:(1) laboratory and/or bench-scale test-ing of site-specific salt. (2) numerical

380

modelling, and (3) field testing. Latersections of this report include schemat-ics illustrating this concept in detail.

Correlation studies should be per-formed on results from laboratory and/orbench-scale tests, borehole and/or modelcavern tests, and proto-type caverntests. Consistent findings relative tolong-term cavern stability then could beused to develop an optimum testing pro-gram for determining necessary site-specific input parameters for stabilitycriteria. Such a program would be ex-tremely valuable in planning additionalCAES cavern systems and operations insalt domes.

Figure 5 is a summary list of topicswhich should be studied to obtain addi-tional information relative to long-termstability of CAES caverns in salt domes'."

Long-Term Creep, with Creep Rupture,of Rock Salt.

• Effects of Pressure and TemperatureLoading Rates.

Low Frequency Fatigue, with CoupledCyclic Pressure, Temperature, andWetting Conditions.

Progressive Air Penetration of SaltFabric.

• Cavern Monitoring Methods.

• Correlation of Laboratory, Analytical,and Field Results.

Fig. 5. Topics for Additional Study OnLong-Term Stability of CAESCaverns in Salt Domes.

DEVELOPMENT OF SITE-SPECIFICSTABILITY CRITERIA

The development of a complete set ofquantitative site-specific stability cri-teria requires: (1) the gathering of site-specific technical data; (2) an appropriateanalysis which will yield predictive quan-titative results; and ultimately, (3) theverification of criteria by agreement be-tween predicted and monitored performance.

In this section a three-stage formatis proposed for developing site-specificcriteria. Figure 6 illustrates the three

stages which incorporate the three listedrequirements for development of long-termstability criteria for CAES caverns insalt domes.

Start Stage (1), Laboratory Testing

(a)J,Formulate Stability Criteria With.

Latest Data Base

<!>)„1

Update Analysis

iclIModel Numerically Current Stage

With Latest Data Base

(d) w

Agreement Between Numerical ModelAnd Physical Behavior? noJ

VisRepeat Steps (a)-(d) For Stage:

(2) Pilot CAES Cavern(3) Prototype CAES Cavern

Criteria Verified

Fig. 6. Development of StabilityCriteria

IMPLEMENTATION OF STABILITYCRITERIA FOR CAES CAVERNS

In this section the implementationcriteria enhancing long-term stability ofCAES caverns is presented in a three-phase format. The three phases generallyfollow a work plan sequence by which aCAES facility could be established andmade operational.

The three phases of stability cri-teria implementation are associated re-spectively with: (1) site selection, (2)cavern system design, and (3) facilityoperation. Figure 7 illustrates schemati-cally the methodology whereby site-specific long-term stability criteriacould be implemented for CAES caverns insalt domes.

381

Start Phase (1): Implementation ofStability Criteria in Site Selection

(a) ,

Propose Salt Dome Site for CAES Caverns

Investigate:

Dome Utilization history

Megascopic features of geomechanicalsystem by geological, geophysical,

I and corehole studies.

(c)

• Apply Long Term CavernStability Criteria:

Relative minimum of extensive early"wild" brining *nd sulphur miningactivities?

Absence of strong evidence of ongoingdissolution of dome at the caprock -salt interface?

Relatively small amounts of gas, brine,and inclusions in salt stock?

Competent, homogeneous rock salt fromapproximately -1000 to -6000 feet, and

I also adequate in volumetric extent? ,

yes no'Return to (la)-

Start Phase (2): Implementation ofStability Criteria in Site Selection

(a)

Propose Design Configuration ForCAES Cavern System

(b) ,

Investigate Via:

Laboratory and bench scale testing ofsite specific salt.

Numerical modelling of proposed designconfiguration. .

Apply Cavern Stability Analysisand Criteria:

Stable stress state in salt around caverns(based on time independent analyses)?

Acceptable rates of creep closure andcreep rupture over life of facility(based on time dependent analyses)?

Acceptable rate of progressive failuredue to cyclic pressure and temperatureloads over the life of facility (based on

[fatigue tests and associated modelling)? ,

7Tyes no% Return to (2a)'

Start Phase (3): Implementation ofStability Criteria for Operating Program

(a)

Propose Operating Program ForCAES Caverns

(*>),

Investigate and Utilize Effectsof Operating Program Via:

Numerical modelling of operatingprogram for stability effects.

Monitor operating effects.

(cLApply Cavern Stability Analysis

and Criteria Application

Preliminary field tests indicate goodagreement between results o f numericalmodelling and test data? If not,reconcile results.

Operating program results in acceptablecavern performance, as indicated byfield monitoring program? Acceptablecavern performance includes measuredacceptable cavern closure, creep closureirates, and surface spa!ling rates.

To 3(d)

382

From 3(c)

yes

(d)

no^Return to (3a)

Continue to Monitor. Update Numerical, Stability Analysis Periodically. .

Fig. 7. Implementation of StabilityCriteria

CONCLUSIONS

CAES technology, relative to long-term stability or air reservoir cavernsin salt domes and salt anticlines., cur-rently is in an early stage of investi-gation. Much work remains to be done,particularly in relating site-specificfield results to results predicted fromengineering analyses based upon test dataand numerical methods.

Adequate numerical programs andtesting methods exist in principle toperform stability analyses of CAES cav-erns. At this time a substantial database should be developed specific to en-vironments anticipated for rock saltsurrounding CAES caverns in domes and/oranticlines. Verification of one or morecompeting methods of analysis also shouldbe completed as expeditiously as possible,with verification based on site-specificfield results.

REFERENCES

1. Johnson, K. S., and Gonzales, S.,Salt Deposits in the United Statesand Regional Geologic CharacteristicsImportant for Storage of RadioactiveWaste. Report Y/OWI/Sub-7414/1,Earth Resources Associates, Inc.,Athens, Georgia, March, 1978.

2. Thorns, R. L., and Martinez, J. D.,Preliminary Long-Term Stability Cri-teria For Compressed Air Energy Stor-age Caverns In Salt Domes, Prepared^ ° ^ Battelle, PNL, Special AgreementB-548O4-A-L, Prime Contract EY-75-C-06-1830, Institute For EnvironmentalStudies, Louisiana State University,DRAFT, August, 1978.

6.

3. Chang, G. C , Loscutoff, W. V., andSchneider, J. R., (Symposium Cochair-men) 1978 Compressed Air Energy Stor-age Symposium, Proc. of Meeting onMay 15-18, 1978, Asilomar ConferenceGrounds, Pacific Grove, California,In Press.

4. Rohr, H. U., Mechanical Behavior of aGas Storage Cavern in Evaporitic Rock,Proc. 4th International Symposium on

1974! J igklSr* 1 1 6 " 1 * 1 0 6 8 0 1 - S ° C "5. State of Louisiana, statewide Order

No. 29-M, "Rules and regulations per-taining to the use of salt dome cavi-ties for storage of liquid and/or gas-eous hydrocarbons, etc.", Departmentof Conservation, Baton Rouge, July,

Mattick, W., Weber, 0., Stys, Z. S ,and Haddenhorst, H. G., Huntorf-TheWorld's First 290 MW Gas Turbine AirStorage Peaking Plant, Proc. AmerPower Conf., Illinois Institute ofTechnology, Chicago, 111., April 1975.

Boresi, A. P., and Deere, D. U., CreepClosure of a Spherical Cavity in anInfinite Medium (with Special Applica-tion to Project Dribble, Tatum SaltDome, Mississippi), for: Holmes andNarver, Inc., May, 1963.

Dreyer, W. E., Results of Recent Stud-ies on the Stability of Crude Oil andGas Storage in Salt Caverns, Proc. 4thInternational Symp. on Salt, V IINorthern Ohio Geological S o c , 1974,P« 65-92.

Nolte, Wierczeyko, Problems OccurringDuring the Sonar Logging of StorageCaverns, Vth International Symposiumon Salt, May, 1978, Hamburg, NorthernOhio Geol. S o c , Proc. In Press

383

Project Title:

PROJECT SUMMARY

S t a b i l i t y and Desiqn C r i t e r i a Studies f o r Compressed A i rEnergy Storage Reservoirs-Rock Mechanics and GeologyComponents

Pr inc ipa l I nves t i ga to r : Howard J . Pincus

Organ iza t ion : Un i ve rs i t y o f Wisconsin-MilwaukeeP. 0 . Box 413Milwaukee, MI 53201414/963-4017, 4561, 4962

Pro jec t Goals: To evaluate the e f f e c t s o f pressure-temperature c y c l i n g w i t hcompressed a i r on underground rese rvo i r rocks and t h e i rassociated caprocks. We are concerned w i t h changes i nphysical properties and in nicrostructure that can beattributed to ventilation. We seek to contribute todevelopment of cr i ter ia for identifying and evaluatingcandidate sites, to identify geological parameters to bemonitored in f ie ld tests, and to assist in developingmonitoring systems for operating installations.

Project Status: Now starting on third phase. Through September 1, 1978, 137specimens were processed in some way; 61 of these wereventilated. Pressure-tenperature conditions were at lowerend of anticipated operating range. In the phase juststarted, higher temperatures and pressures wi l l be used.

Contract Number: Jan. 1, 1977 - Sept. 30, 1978 - Battelle PNL AwardB-37774A-KOct. 1, 1978 - Sept. 30, 1979 - Battelle SubcontractB-21286A-K

Contract Period: (1) Jan. 1, 1977 - Sept. 30, 1977(2) Oct. 1, 1977 - Sept. 30, 1978(3) Oct. 1, 1977 - Sept. 30, 1979

Funding Level : (1) $25,000(2) $73,800(3) $62,000

Funding Source: Battelle Pacific Northwest Laboratories

385

FABRIC ANALYSIS OF ROCK SUBJECTED TO CYCLING WITH HEATED, COMPRESSED AIR

Howard J. PincusDepartments of Geological Sciences and

Civil EngineeringUniversity of Wisconsin-Milwaukee

Milwaukee, Wisconsin 53201

ABSTRACT

An essential component of current investigations of compressed air energy storagein aquifer-type rocks is the characterization of changes with ventilation in pore-and-grain structure of the rocks, particularly as related to porosity and permeability. Inaddition to conventional microscopic examination, we are using. Fourier optics to expresschanges in spatial frequency and orientation of grains, pores and microcracks. OpticalFourier amplitude transforms depict the distribution of spatial frequencies and orienta-tions in their input-images. We can also express the similarity between pairs of inputimages by cross-correlating their transforms. Each of these Fourier optical procedureshas detected some changes with ventilation in rock fabric. Differences between rocksare often greater than differences in the same rock associated with ventilation.Differences in gross physical characteristics among some rocks are consistent withcontrasts in fabric.

INTRODUCTION

The goals of this project are:

1) To evaluate the effects of pres-sure-temperature eyeling withcompressed air on undergroundreservoir rocks and their asso-ciated caprocks.

2) To contribute to the developmentof criteria for identifying andevaluating candidate sites.

3) To identify geological parametersto be monitored in field tests.

4} To assist in developing monitor-ing systems for operating in-stallations .

We have been concerned chiefly withdetecting and measuring changes in physi-cal properties and in microstructure thatcan be attributed to ventilation.

This paper is concerned with some ofthe methods and results of studies of themicrostructure of some candidate rocks.It will perhaps be useful here to outlineresults obtained so far in other phasesof our ^

1) Values of Young's modulus andcompressive strength of sandstones studiedso far do not show systematic changes

with ventilation.

2) Changes in permeability asso-ciated with ventilation have not beendetected.

3) Changes in porosity and heatcapacity of St. Peter Sandstone, asso-ciated with ventilation, have not beendetected.

4) Microscopic examination in re-flected light has revealed no diagnosticsfor distinguishing with confidence be-tween ventilated and unventilated rocks.

5) Heating of Berea Sandstone andBedford (Salem) Limestone to 260°C. forfive days has resulted in decreases inthe value of Young's modulus. Compressivestrength of the Berea also decreased.

In the ventilation experimentscarried out, typical air temperatures atthe specimen were 110°C., with confiningpressures up to 820 kPa.

The most striking result of the workdescribed and partially documented inthis paper is that initial variations infabric within the same unit, e.g., theSt. Peter Sandstone, appear to exceed thechanges resulting from ventilation underthe experimental conditions used so far.

386

Observation of the St. Peter in iarge ex-posures indicates considerable variabili-ty in lithology and structure.

ROCKS STUDIED

Host of the specimens studied so farare St. Peter Sandstone, collected in anactive quarry in south-central Wisconsin(Fig. 1). Experimental work has also beendone on Platteville Limestone and JordanSandstone, also collected in south-centralWisconsin, and on two members of the U.S.Bureau of Mines-A.R.P.A. Standard rocksuite, Berea Sandstone and Bedford (Salem)Limestone.

Specimen blocks weighing 13-40 kg.have been collected and cored (Figs. 2,3).The time-based eyeling system shown inFig. 4 has been used to ventilate all ofthe specimens for which results are pre-sented in this paper. This equipment hasbeen eyeled at up to almost 1000 cycles infour days with mean air velocity throughsome specimens exceeding IS cm/s.

FABRIC ANALYSIS

We are concerned with the size, shape,arrangment and orientation of grains,pores, and microfractures, with the dis-tribution of cement and other fine-grainedmaterial in the pores, and with changes intime (or ventilation) of any of the fore-going. Microscopic analysis, which isbeing continued, has so far not yieldeduseful diagnostics.

To augment fabric analysis, we havebeen using a Fourier-optical approach tospatial analysis, supported by a recently-developed method of optical correla-tion1'-'3'4

Utilizing coherent light and suitableoptics, the two-dimensional Fourier ampli-tude transform of a transparent image ofa rock surface is produced as a diffrac-tion pattern. The directions of radii topoints in the transform are perpendicularto the directions of the corresponding in-put elements; the radial coordinates ofdots in the pattern vary linearly with thespatial frequency of the corresponding in-put linear elements. Some basic Fouriertransform relations are presented in Fig.5, taken from a pictorial digital atlas-*intended for seismologists.

Photographic images of specially

treated surfaces of rock specimens andtheir optical transforms are shown inFigs. 6 and 7. Note the horizontalelongation of the Fig. 7 transform, cor-responding to the input's vertical"grain".

The scanning of transforms yieldsprofiles that greatly facilitate compari-son. Earlier work ' has shown changesassociated with ventilation that seem toindicate some transport of fines in somerocks when airflow is unidirectional;this is manifested as, for example, largerhigh spatial-frequency content in thetransform of the low-pressure end of thespecimen.

Optical correlation provides anothermeans for quantifying comparisons ofspatial information.

As we have developed and appliedthis method in our work, three genera-tions of optical transforms are produced.First, transforms of the two images to becompared are generated. These two trans-forms are then used as an input, side-by-side, and their joint-transform is genera-ted (Fig. 8, top). The transform of thejoint-transform is then generated (Fig. 8,bottom), and the intensity of its first-order diffraction spots is a measure ofthe degree of similarity of the spatialinformation recorded in the first-genera-tion transforms. These cross-correlationsare normalized by dividing the raw cross-correlation reading by the geometric meanof the autocorrelations of each of thetwo first-generation transforms. Some ofthe results obtained so far are presentedin Table 1, adapted from our last annualreport1.

The top half of the table representsdata for two unidirectionally ventilatedspecimens and the bottom half presentsdata for two specimens, each with oneventilated and one unventilated input.

In the upper half, the correlationis smaller between the high- and low-pressure ends of the red St. Peter thanbetween the high and low ends of the graySt. Peter. In both the field and thelaboratory, the red St. Peter crumblesmore easily than the gray.

In the bottom half of the table, thecorrelation between unventilated andventilated red St. Peter is smaller than

387

between corresponding yellow St. Peterspecimens. In the field, the yellow issomewhat more durable than the red.

Table 1. Optical correlation of venti-lated specimens.

ting result is that the introduction oflots of fines into pore spaces seems tohave a marked effect on the correlationcoefficients (cxe vs. cxd), a result alsoachieved with other inputs not presentedhere.

Rock

St.Peter(gray)BE-I-18

St.Peter(red)BE-I-25

St.Peter(gray andred) BE-I-1S,25

St.Peter(gray andred) BE-I-18,25

NormalizedCross-Correlation

High vs. low 0.97pressure(gray)

High vs. low 0.88pressure(red)

High-gray 0.93vs. high-red

Low-gray 0.83vs. low-red

Table 2.Lath-like

a

b

c

d

Normalizedartificial

a b

.13

CONCLUDING

optical correlations-inputs (Fig. 9).

c d

.13

.34 .37

.47

REMARKS

e

.18

.28

.21

St.Peter(red)BE-I-25

Vent. vs.unvent. (red)

St.Peter(yellow) Vent. vs.BE-I-22 unvent. (yellow)

0.82

0.91

St.Peter(red and Red vs. yellow, 0.85yellow) BE-1-25,22 unvent.

St.Peter(red and Red vs. yellow, 0.84yellow) BE-I-25,22 vent.

Further, correlation between red andyellow unventilated specimens is aboutthe same as between red and yellow venti-lated specimens; both of these correla-tions indicate less spatial similaritythan is indicated by the correlationbetween yellow unventilated and ventilated.Variations within the same unit, say theSt. Peter, may exceed changes resultingfrom ventilation.

OPTICAL CORRELATION OFARTIFICIAL INPUTS

Interpretation of the optical corre-lation coefficients obtained so far, re-quires further investigation. We arecurrently determining relative sensitivi-ties to contrasts in grain size, grainshape, grain arrangement, preferredorientation, and sorting.

A sample of the work underway andpartly reported* is shown in Fig. 9.Some of the normalized correlation coeffi-cients are shown in Table 2. One interes-

We are continuing our work withartificial inputs, investigating theeffects of several key fabric variables,singly and in combination. Concurrently,we are continuing analysis of rock speci-mens. Recently we have completed con-struction of ventilation equipment whichwill permit use of air at temperaturesand pressures near the probable upperlimits of an operating system.

The assistance of Dr. Karl Scheiben-graber and Mr. Jack Hopper in preparingthis paper is acknowledged with thanks.

Support for this work has been pro-vided by D.O.E. through B.P.N.L., andthe University of Wisconsin-Milwaukee.

REFERENCES

1) Pincus, H. J. (1978),FY 1978 ProgressReport to BPNL and DOE, 57 pp.

2) Smith, G. C. et al. (1978), FY 1977Progress Report, PNL-2443, UC-94b,Ch. 4, Laboratory Testing.

3) Pincus, H. J. (1978), U.S. Natl. CommRock Mech., Proc, 19th U.S. Symp.,v. 1, 215-220.

4) Pincus, H. J. (1978), Int. Assn. Eng.Geol., Proc, III Cong., Sec. Ill,v. 2, 107-116.

5) Peterson, R. A. and Dobrin, M. B.(1966), "A Pictorial Digital Atlas",United Geophysical Corp., Pasadena.

388

\

• - ; / •

Fig. 1 Faulted St.Peter Sandstonein quarry near Verona,Wisconsin.Scale: Haraner in foreground.Rock breaks out in large blockswhich disintegrate quickly. Darkerbeds are red.

Fig. 3 Cored specimens in severalstages of preparation. Host spec-imens are St.Peter 3andston«,5i*flu».in diameter.

Fig.2 Specinen blocks weighing 15-kO kg. Scale is 15 en. long.Two of the blocks shown have beendrilled for NX(54*a) cores.

Fig. 4 Time-based cycling system.Specimen chamber is in front ofthree circular pressure gauges,right-center. Air is heated in tallcolumn at extreme right. At left,three pyroneters(rectangular).

389

TIME DOMAINOR FREQUENCY DOMAIN

FREQUENCY DOMAINOR TIME DOMAIN

- 1 1 1 1 I 1 1 1

WINDOWFUNCTION If

- • -

J_L

J__L

b) nnnnfjinnnn

o

nnifinn

c)

Fig. 5 Fourier transforms oftruncAted functions. The windowfunctions correspond to aperturesof illumination. In a),b),and c)in the left-hand column, the inf-ir.ite function times the windowfunction yields the truncatedfunction iwaediately below. Trans-

form equivalents are in the right-hand column ; the * denotes conv-olution, by which the transform ofthe truncated function is obtainedfrom the functions above. Thesquare wave in b),left side, corr-esponds to a profile across whiteand black lines of equal width.

390

Fig. 6 Top. Light gray Berea Sand-stone. Specimen pores are filledwith whits paint.Short bar: O.kmn.Bottom. Optical transform of above.(O.D.A.2372).Scale bar:2.5 lines/mm

Fig. 7 Top. Yellow and pink St.Peter Sandstone.Same treatment andscale as in Fig. 6. Bottom.Opticaltransform of abovo(O.D.A.23?3).Same scale as Fig.7.

391

Fig* 8 Joint-transform opticalcorrelation.Top. Joint-transform ot two optic-al transforms. Regularity and dis-tinctness of vertical bands aregreater the more similar are thetwo transforms being compared.

Bottom. Optical transform of thejoint-transform above. The bright-ness of the two first-order diff-raction dots, marked by invertedV's, varies with the regularityand distinctness of the verticalbands in the joint-transform.

Fig.9 Top. Artificial inputs sim-ulating lath-like grains. Froma to c preferred orientation de-creases and porosity increases.The large grains in c.d.and aare identical, with fine materialsin the pores increasing fromc to e ,

Bottom. Optical transforms of theinputs above.Mote that the numberof radiating bands increases fromthe transforms of a to c, and thatthe high spatial-frequency contentincreases from the transform of cto that of e .

392

PROJECT SUMMARY

Project Title: Numerical Modeling of Behavior of Caverns in Salt forCompressed Air Energy Storage (CAES)

Principal Investigator:

Organization:

Shosei Serata, Ph.D.

Project

Project

Goals:

Status:

(1)

(2)

(3)

(4)

(5)

(1)

Contract Number:

Contract Period:

Funding Level:

Funding Source:

Serata Geomechanics, Inc.1229 Eighth StreetBerkeley, CA 94710(415) 527-6652

To identify the practical ranges of the variables that areimportant to the geomechanical design of CAES cavities inrock salt.

To develop procedures for the computer simulation of theeffects of these variables on the behavior of CAES cavities.

To perform parametric studies of the behavior of typicalsolution cavities in domed salt formations.

To develop general stability and design criteria for CAEScavities having projected lifetimes of about 50 years.

To identify areas of further research required to arrive atsatisfactory stability and design criteria for CAES cavernsin rock salt.

A literature and data base review has been completed, althoughcurrent literature is constantly checked for new developmentsrelevant to CAES projects.

A limited laboratory investigation of the behavior of rocksalt under cyclic loading conditions is in progress. Anumber of cyclic and static tests have already been completed.

Preliminary parametric studies of cavern depth, cavern airpressure, excess in situ lateral stress, and material prop-erties of rock salt are in progress. Preliminary parametricranges have been established but will be modified as theprogram progresses.

Special Agreement No. B-54809-A-PPrime Contract No. EY-76-C-06-1830

Mar. 7, 1978 - Feb. 28, 1979

$85,000

Battene Pacific Northwest Laboratories

(2)

(3)

393

NUMERICAL MODELING OF BEHAVIOR OF CAVERNS IN SALTFOR COMPRESSED AIR ENERGY STORAGE

Thomas E. Cundey, Senior Math Analystand

Shosei Serata, PresidentSerata Geomechanics, Inc.

1229 Eighth StreetBerkeley, California 94710

ABSTRACT

Finite element techniques are being used to establish generaldesign criteria for caverns in salt formations that could be used forcompressed air energy storage (CAES). The three types of parametersbeing studied are geological (in situ stresses, material properties),geometric (cavern dimensions, cavern spacings), and operational (airpressures, air temperatures). The numerical studies are being supple-mented by a laboratory testing program developed to determine the effectof a cyclic loading environment on the material properties of rock salt.The five phases of the overall program are described, and the work com-pleted to date is discussed. Preliminary ranges of the most significantparameters are presented. Areas in which additional research isnecessary are identified, with particular regard to laboratory testing.

INTRODUCTION

Among the unknowns that affectthe technical and economic feasi-bility of compressed air energystorage (CAES) in rock salt are theresponses of underground openingsto fluctuations in pressure andtemperature that would occur in thedaily charge and discharge opera-tions of CAES plants. The thermo-mechanical stresses generated bysuch fluctuations and by the ele-vated storage temperatures mightcause mechanical damage to thewalls of CAES caverns, therebyaffecting their overall stability.The material properties of rocksalt, including creep behavior,have been studied previously.However, many uncertainties remainconcerning the responses of rocksalt to very long-term loading,cyclic loading, thermal variations,and volumetric changes, and theways in which these responses

might influence the long-termbehavior of CAES caverns in thismaterial.

In order to develop suitablestability and design criteria forCAES caverns in rock salt, amulti-phased program of study hasbeen initiated. The first phase,a state-of-the-art survey andformulation of preliminarystability and design criteria, iscurrently underway at LouisianaState University (LSU). Thesecond phase, which is the subjectof this paper, will use numericalmodeling of the behavior of CAEScaverns in rock salt. The thirdand fourth phases will be alaboratory investigation of thematerial properties of rock saltrelevant to CAES caverns and fieldstudies of CAES cavern behavior.

394 —

PROJECT DESCRIPTION

The five principal researchactivities in this second phase ofthe program and the studies to beundertaken within each activityare described in the followingsub-sections.

IDENTIFICATION OF RANGES OFVARIABLES

There is considerable uncer-tainty about some aspects of themechanical and thermal behavior ofrock salt surrounding CAES caverns.Therefore, for purposes of numeri-cal simulation studies, a practicalrange of values must be establishedfor each variable that has signi-ficant influence on the behaviorof the caverns. The followingsources of data.are being used todevelop ranges for these variables:(1) the published literature;(2) the data base established bythe LSU state-of-the-art survey;(3) operational criteria developedby previous CAES studies; and(4) data and experience accumulatedby Serata Geomechanics, inc. (SGI)in previous studies of the behaviorof salt caverns used for the stor-age of compressed air, natural gas,LPG, and crude oil.

The variables for which para-metric ranges are being investi-gated include the following:cavern geometry, cavern depth,cavern separation (in multiplecavern systems), in situ earthstresses, the rheological proper-ties of rock salt, the permeabili-ty of rock salt, cavern operatingmode (compensated or uncompensated),and cycles of operating tempera-tures and air pressures. Majoremphasis is being placed on uncom-pensated caverns created in saltdomes by solution mining. Thehighly site-specific nature of thegeologic and geometric parametersinvolved in the numerical analysisof abandoned conventionally-mined

underground openings in beddedsalt formations was a major con-sideration in giving priority tosolution cavities in salt domes.

The data available fromexisting sources is insufficientto establish satisfactory generalcriteria for the effects of cyclicloading on the behavior of rocksalt. A limited program of lab-oratory testing is thereforebeing undertaken to investigatethese effects. This work is in-tended to give qualitative andsome quantitative information tosupplement existing knowledge offatigue behavior of earth materi-als . A small quantity of rocksalt has been obtained for thesetests. The basic properties ofthe salt are derived from tri-axial tests and triaxial creeptests on some of the samples.Other samples are subjected todifferent numbers of cyclic re-versals of deviatoric stress underconfining pressures typical ofthose expected in rock salt sur-rounding CAES caverns. Aftercyclic loading, the elastic andcreep properties of the salt aremeasured and compared with theproperties of the salt samplesthat were not cyclically loadedto determine the effects of cyclicloading on salt properties.

COMPUTER SIMULATION TECHNIQUES

The primary computer simula-tion studies are being performedusing SGI's finite element com-puter program (REM). This programhas been developed and used tostudy the time-dependent behaviorof underground openings in con-ventional and solution mines andthe stability of storage cavitiesfor oil, natural gas, LPG, andcompressed air.

To study the effects ofthermo-mechanical stresses and oftemperature changes on material

395

properties, the finite elementprogram HEAT, developed at theUniversity of California, Berkeley,is being used. This programanalyzes steady-state and transientlinear heat conduction in plane andaxisymmetric regions. Output fromthe HEAT program provides insightsinto the magnitude and distributionof rock temperatures which are usedas input to the geomechanicalanalyses to be performed using theREM program.

PARAMETRIC STUDIES

The parametric studies aredesigned to investigate the effectsof variations in the principal geo-logical, geometric, and operationalvariables on cavern behavior andstability. The principal variablesto be studied using the computersimulation models include caverngeometry and depth, material prop-erties of the rock salt, in situstresses, and operating conditions(e.g., magnitudes and rates ofchange of pressure and temperature).

The range of each variableused in the parametric studies isbased on the results of the litera-ture and data-base review. However,if the numerical analyses indicatethat other variables or variableranges should be investigated, theanalysis program will be modifiedappropriately. A limited number ofsimulations will also be conductedto investigate modes of cavernfailure induced by extreme condi-tions such as very high or verylow cavern pressure, elevated airtemperature, extreme cavern dimen-sions, and small cavern separationdistances in multi-cavern networks.

CRITERIA FOR STABILITY AND DESIGN

Drawing upon the results ofthe studies at LSU and the computersimulation studies described above,general criteria for the stabilityand design of CAES caverns are

being developed. The criteriabeing developed include caverngeometry and depth, in situ earthstresses, cavern spacing, materialproperties of rock salt, the per-meability of rock salt, the magni-tudes of operating pressures andtemperatures, and the maximum flowand heat transfer rates.

RECOMMENDATIONS FOR ADDITIONALRESEARCH

Using the results of the cur-rent research and other experi-ences in the design of undergroundcavities, additional geotechnicalresearch necessary to developsatisfactory criteria for thestability and design of CAEScaverns in rock salt will beidentified and recommended. Theserecommendations will include guide-lines for laboratory testing ofrock salt and field studies ofCAES caverns.

RESULTS

LITERATURE REVIEW

A review of the publishedliterature was conducted at theoutset of the research in hopethat some of the environmentalconditions surrounding CAES cavernsin salt had been investigatedpreviously. Because much work hadbeen done by the authors and otherinvestigators in studying thestatic mechanical properties ofrock salt with particular emphasison creep, this area was de-empha-sized in the literature review.Major emphasis was instead placedin three other areas: the effectof cyclic loading on the behaviorof rock salt, the permeability ofrock salt under various stressstates, and the effect of elevatedtemperature on rock salt behavior.

Cyclic Loading. No literaturecould be found on the effect ofcyclic loading on rock salt

396

behavior, and in fact little workhas been done on the behavior ofany rocks under cyclic loading con-ditions. Soil materials undercyclic loading conditions, on theother hand, have been studied ex-tensively in recent years, partic-ularly to advance the art ofdesigning earthquakt;-resistantstructures and for offshore oil-drilling platform foundations. Themost comprehensive laboratory in-vestigation of rock behavior undercyclic loading conditions was done"by Professor B. C. Haimson and hiscolleagues at the University ofWisconsin. Although the differ-ences in the behavior of rock saltand the four hard rock types(marble, limestone, sandstone, andgranite) used in the Wisconsinstudy are appreciated, it is never-theless useful to summarize some ofProfessor Haimson's findings.

In cyclic uniaxial compressiontests on all four types of hardrock, the following observationswere made:

(•1) All rock types showed afatigue strength of 60percent to 80 percent ofthe uniaxial compressivestrength;

(2) Stress-strain curvesshowed large hysteresisin the first cycle,decreasing hysteresis inthe next cycles, nearlyconstant hysteresis loopsin the following cycles,and a reopening of theloops in the final cyclesprior to failure;

(3) Three characteristicstages of cyclic creepwere observed in allfour rock types: a pri-mary stage, in whichstrain increases fromcycle to cycle at a de-celerating rate, a

secondary stage withconstant creep rate, andfinally a stage in whichthe strain rate increasesuntil failure occurs;

(4) The cyclic creep curvewas invariably boundedby the complete stress-strain curve of therespective rock.

Only granite was tested incyclic triaxial compression, andits behavior was similar in everyrespect to that under cyclic uni-axial compression, in addition,an increase of fatigue strengthwith increasing confining pressurewas observed. Other cyclic testswere conducted under uniaxial ten-sion and uniaxial compression-tension, but those loading condi-tions are not relevant to CAEScaverns and will not be discussed.

Attewell and Farmer proposeda hypothesis to explain the resul-tant deformation of a fine-graineddolomitic limestone in cyclic uni-axial compression which appears tobe compatible with Haimson's find-ings and is based on strain-energydependent crack propagation. Abovea threshold stress level at whichcracks are initiated, the deforma-tion from successive sub-failureload cycles is cumulative, withfailure occurring when the strainenergy stored in the rock exceedsa critical level equivalent tofailure under non-cyclic loading.

Permeability. The permeability ofrock salt under various stressstates was studied by Chia-ShingLai at Michigan State Universityin 1971. A special high pressuretriaxial cell was designed and con-structed for the permeabilitytests. Lateral and axial loadingwas controlled by two independenthydraulic systems. Uniform flowof the permeating fluid (kerosene)was applied to the specimens under

397

a constant fluid pressure. Alltests were conducted at room tem-perature .

Numerous tests were conductedusing a wide range of values ofmean stress and octahedral shearingstress, and empirical relationshipsbetween permeability and stresswere developed. The permeabilityof rock salt was found to varybetween 0.0036 and 40.68 millidar-cies for any stress state imposed.The following empirical equationwas derived from the laboratorydata:

log k = 0.62 + 0.00139To"ct "0*000670^

where k is ths permeability (milli-darcies), T'oct i s t^e octahedralshearing stress (psi), and 0"m isthe mean stress (psi). Note thatthe permeability increases withincreasing shear stress but de-creases with increasing meanstress.

Temperature. The effects of tem-perature on the behavior of rocksalt have been studied relativelyextensively in both the laboratoryand the field. Serata has foundthat temperatures up to 300 degreesFarenheit have little effect on theelastic and viscoelastic behaviorof salt, but that the viscoplasticproperties (octahedral shearingstrength and viscosity) are strong-ly affected by elevated tempera-tures. The shearing strength ofsalt was observed to decrease withincreasing temperature, and thesalt became more ductile. Brad-shaw, Lomenick, McCain, and othershave studied the thermal behaviorof salt in both the laboratory andthe field during Project Salt Vault,a study of the feasibility ofnuclear waste storage in salt mines.They also found the thermal effectson creep to be very significant andhave fit their model pillar studiesdata with an empirical function

relating creep rate with tempera-ture, stress, and time:

3.2 x 10-36 fc-0.65

where d is the pillar strain rate(10~6 per hour), T is the tempera-ture (degrees Kelvin), <T is thestress (psi), and t is the time(hours).

The geothermal temperatureincrease with increasing depth .(about 2 degrees Farenheit per100 feet depth) would indicate atemperature of about 150 degreesFarenheit at 4,000 feet, the deep-est depth to which CAES cavitieswould probably extend, and theviscoplastic creep rate is signi-ficantly increased at temperaturesin this range. The pressure dif-ference between the lateral earthstress and the cavity air pressurealso increases with depth. Fur-thermore, the boundary temperaturewill vary cyclically with the airpressure. This will induce a non-uniform temperature distributionaround the cavity which has to bedetermined so that the temperature-dependent material properties ofthe salt near the boundary can beadjusted. A two-dimensional linearheat conduction finite elementprogram developed by ProfessorR. L. Taylor of the University ofCalifornia has been adopted forthis purpose.

LABORATORY TESTING

Cylindrical salt specimenswith diameters of 6.6 centimeters(2.6 inches) and lengths of 13.2centimeters (5.2 inches) werecored from slabs of salt from theDiamond Crystal Salt Mine in theJefferson Island Salt Dome inLouisiana. Some of these samplesare being tested in conventionaltriaxial compression and triaxialcompression creep tests to estab-lish their elastic and time-dependent material properties.

398

Other samples are being loadedcyclically to determine the effectof cyclic loading on the materialproperties as determined in thestatic tests.

Although a number of thecyclic tests have been conducted,the analysis of the test resultshas not yet been completed. Oneseries of tests was conducted ata confining pressure of 17,250kN/m2 (2,500 psi) and axial loadvarying between 17,940 kN/m(2,600 psi) and 53,130 kN/m2

(7,700 psi) at a frequency of 1cycle per hour. This extremelyhigh deviatoric stress, equivalentto about 90 percent of the tri-axial strength at this confiningpressure, is unrealistic to expectin a CAES environment but waschosen to provide evidence offatigue characteristics of rocksalt. A second series of tests,still in progress, employs a con-fining pressure of 17,250 kN/m2

(2,500 psi) and axial load varyingbetween 21,840 kN/m2 (3,165 psi)and 29,190 kN/m2 (4,230 psi) at afrequency of 2 cycles per hour.After a predetermined number ofcycles (which varies from test totest) are completed, the samplesare tested in triaxial compressioncreep tests, and the materialproperties therein determined willbe compared with those of the non-cyclical ly loaded samples.

PRELIMINARY NUMERICAL MODELINGSTUDIES

Parametric ranges for some ofthe principal variables to be in-vestigated have been establishedbased on the results of the liter-ature review, previous CAES studies,and the authors' experiences insalt cavern design, as follows:

Cavern depth = 153 m to 1,529 m(500 ft to 5,000 ft)

Cavern air pressure = 0 kN/n»2 to15,215 kN/m2 (0 atm to 150 atm)

Excess in situ lateral stress =-10,978 kN/m2 to +10,978 kN/m2

(-1,591 psi to +1,591 psi)

Octahedral shear strength of salt =3,795 kN/m2 to 5,175 kN/m2

(550 psi to 750 psi)

Other salt properties = Weak tostrong

In the first studies thebehavior of a horizontal sectionof a single vertical cavern in aninfinite salt mass was investigatedusing an axisymmetric model. Aset of parameter base values wereused:

Cavern depth = 610 m (2,000 ft)

Cavern air pressure = 5,070 kN/m2

(50 atm)

Excess in situ lateral stress =4,391 kN/m2 (636 psi)

Octahedral shear strength of salt =4,140 kN/m2 (600 psi)

Other salt properties = Weak

One parameter was varied throughthe ranges described above whileche others held their base values.Cavern air pressure in these firstanalyses was static rather thancyclic. Final graphs of the out-put data from these runs are notcomplete.

FUTURE WORK

Once the preliminary para-metric studies of cavern depth,cavern air pressure, excess insitu lateral stress, octahedralshear strength of salt, and othersalt properties are completed, thesingle cavern models will be modi-fied so that other effects can beinvestigated. Cyclic cavern air

399

pressure will be studied, and pro-gram HEAT will be used to computetemperature distributions aroundsingle vertical caverns. Thedegree of air permeability will beassessed based on Lai's studies,and material properties of saltnear the cavern boundary will beadjusted to simulate the deterior-ation of the boundary salt. Allthese analyses will employ thegeometry of a horizontal sectionof a single vertical cavern in aninfinite salt mass, as this is themost economical geometry.

The next step after the para-metric studies of the operationaland geological variables is aparametric study of geometricvariables. The first of thesewill be cavern diameter and cavernlength; and again a single verticalcavern in an infinite salt masswill be analyzed axisymmetrically.but the entire cavern will bemodeled rather than just a horizon-tal section. Single-cavernanalyses will be followed bydouble-cavern and other multi-cavern simulations, so that theeffect of separation distancebetween adjacent caverns can beinvestigated. Although therealistic ranges of most of thevariables will have been estab-lished by this time, a number ofextreme cases will also be run togain insight into potential cavernfailure modes.

The results of all the numeri-cal studies will be carefullystudied to determine generalstability and design criteria forCAES caverns. The numerical re-sults will be supplemented by theresults of the literature study.Professor Thorns' state-of-the-artreport, and our experiences insalt cavern design. Recommenda-tions for future work in both thelaboratory and the field will bemade.

Although our numericalstudies are far from completion, itis evident that much work has to bedone to evaluate possible couplingeffects among the key operationalparameters—cyclic air pressureand air temperature. As discussedearlier, no studies have beenreported on the effects of cyclicloading on salt, but there isliterature on rock salt permeabili-ty and temperature effects on saltproperties. It is hoped thatlaboratory investigations of thecombined effects of elevated tem-perature and cyclic loading can becompleted and the numerical modelsre-evaluated after these labora-tory investigations have beencompleted. Furthermore, finalverification of the modelsrequires the collection andanalysis of field data from a CABSdemonstration site.

SELECTED REFERENCES

Attewell, P.B. and I.W. Farmer,1973. Fatigue Behavior of Rock,international Journal of RockMechanics and Mining Sciences,Vol. 10, pp. 1-9.

Bradshaw, R.L., et al., 1968.Properties of Salt important inRadioactive Waste Disposal,Special Paper No. 88 of theGeological Society of America,pp. 643-658.

Haimson, B.C., 1973. MechanicalBehavior of Rock under CyclicLoading, Final Technical Reportto the Bureau of Mines. ContractH0220041.

Lai, Chia-Shing, 1971. FluidFlow through Rock under VariousStress States, Ph.D. Thesis,Michigan state University.

Serata, S., T.E. Cundey and C. Lai,1978. Rock Mechanics Problemsin the Design of Solution Cavi-ties and use of Abandoned Mines

400

for the Storage of CompressedAir Energy, Proceedings of theCAES Technology Symposium,Asilomar, California.

Taylor, R.L., 1975. *HEAT*, AFinite Element Computer Programfor Heat-Conduction Analysis,Report 75-1 for the Civil En-gineering Laboratory, NavalConstruction Battalion Center,Port Huename, California,Purchase Order N62583/75-M-Y695.

401

PROJECT SUMMARY

Project Title: Compressed Air Energy Storage Studies

Principle Investigator: Dr. G. T. Kartsounes

Organization: Argonne National Laboratory9700 South Cass AvenueT-12-3Argonne, IL 60439312/972-7695

Project Goals: The objective of this program is to extend the technology base of thethe compressed air energy storage (CAES) concept for application toelectric utility peak-power generation. During FY 1978, three studieswere conducted. The objectives of the studies were: to develop a gen-eral design optimization procedure for CAES plants having aouifer stor-age reservoirs; to evaluate the performance and capital and'operatingcosts of possible turbomachinery options for CAES plants; and to evaluatethe performance, cost, and manufacturing feasibility of a unique reci-procity expander/compressor which could have a major impact on CAESviability.

Project Status: Due to decreased funding for FY 1979, no further work will be doneon the development of the design optimization procedure for CAESplants having aquifer reservoirs. The evaluation of turbomachineryoptions has been completed, and no further work is anticipated. Apreliminary evaluation of the reciprocating engine has been completed,and further evaluation will continue through FY 1979.

Contract Number: W-31-1O9-ENG-38

Contract Period: FY 1978

Funding Level: $310,000 BO *

Funding Source: Department of Energy, Office of Conservation, Dtvistonof Energy Storage Systems

* Included in this project are the following:

George T. Kartsounes and Choong S. Kim, "Evaluation of Turbomachinery forCompressed Air Energy Storage Plants," page 416;

George T. Kartsounes and James G. Daley, "Evaluation of the Useof Reciproeating Engines in Compressed Air Energy Storage Plants,"page 425.

403

THE DESIGN OPTIMIZATION OF AQUIFER RESERVOIR-BASED COMPRESSED AIR ENERGY STORAGE SYSTEMS

Frederick W. AhrensEnergy and Environmental Systems Division

Argonne National Laboratory9700 South Cass Avenue

Argonne, Illinois 60439

ABSTRACT

The application of a general Compressed Air Energy Storage (CAES) power system designoptimization methodology to the class of CAES plants having aquifer air storage reservoirsis discussed. The resulting procedure incorporates performance and economic models forthe aquifer reservoir, wells, piping, and air compression system. Its use allows identi-fication of designs which minimize the subsystem power generation cost (mills/kWh), whilesatisfying constraints related to the geology, equipment, and utility load curve. Thedesign specification resulting from the optimization procedure includes: land area to bepurchased, well depth, number of wells, well spacing, wellbore diameter, main pipelinediameter, required compressor system power and discharge pressure, and required compres-sion time durations for each day of the week. A capital and operating cost summary forthe optimum design is a final output of the procedure. This paper reviews the models andconstraints incorporated in the optimization procedure. Although the basic framework iswell-developed, some refinements or additions to the modeling may be necessary to improvethe results; these possibilities are discussed. Results of case studies are included inthe paper in order to illustrate the power and potential economic impact of the techniquesdescribed, to demonstrate some of the economic tradeoffs which occur' in the optimal designof aquifer reservoir-based CAES systems, and to show the influence of certain cost para-meters .

INTRODUCTION

A major portion of the Department ofEnergy research program on Compressed AirEnergy Storage (CAES) is devoted to addres-sing air reservoir concerns. In the caseof constant-pressure (hard rock) caverns,it is quite easy to design a reservoirwhich satisfies the planned operatingcycle of the CAES plant, once the turbinesystem air supply requirements have beenspecified. It is then straightforward toinclude the effect of cavern costs in eco-nomic studies of turbomachinery options(e.g., see Ref. 1). The situation is some-what more complicated in the case of con-stant-volume (usually salt cavity) reser-voirs. For these reservoirs, the peakstorage pressure (related to the amount ofcushion air) must be selected, which in-volves finding the economic balance betweenair compression and cavern volume costs.The relative design simplicity of hard rockand salt cavity reservoirs has resulted ina DOE-sponsored research emphasis on thelong-term stability or reliability of thereservoirs undergoing CAES plant operating

conditions.

When aquifers are considered for CAESreservoir application, the number of designparameters to be selected is much largerand, in addition, there are many constraintsimposed by the operating cycle, the inter-action with aboveground machinery, and thegeological characteristics. These factorsgive a great incentive to the careful ex-ploration of aquifer system design options,so that the economic benefit of the plantcan be maximized while simultaneously in-suring its long-term capability to meet theplant operational criteria. The primarygoal of the work reported here has been todevelop an appropriate, comprehensive,means for performing these design studies.The performance, economic and design opti-mization considerations which form thebases of the design procedure are describedand illustrated with example case studyresults in subsequent sections.

404

DESIGN STRATEGY

PHILOSOPHY

Because of the complexity of an aqui-fer reservoir-based CAES system, both interms of subsystem interactions and designconstraints, it was decided to orient theformulation of the design procedure towardthe utilisation of generalized, computer-oriented techniques for solving nonlinear,constrained optimization problems.2 Withthis approach, a basic framework for systemdesign can be developed so that new improve-ments in the technical or economic modelscan be easily incorporated. Implicitly,the choice of basing the design procedureon modern optimization techniques reflectsa recognition of the confusion and inade-quacy which could result from applying themore traditional "parameter study" approachto a problem of this magnitude.

SYSTEM CONSIDERATIONS

In principle, and probably in practice,the design optimization of an entire CAESsystem could be handled as one large prob-lem, especially if the models employed forindividual system components were not toodetailed. However, consideration of ageneral formulation led to the developmentof an advantageous decoupling strategywhich enables separate optimization of par-ticular subsystems without compromisingthe optimal system design. Each of thesesuboptimizations is, of course, simpler toperform than that of the full problem.

In broad terms, a CAES power systemcomprises the following: the air compres-sion train (compressors, intercoolers, af-tercooler); compressed air piping; airstorage reservoir (any type); power genera-tion train (e.g., turbines, combustors, re-cuperator) ; reversible motor/generator andthe utility grid. Although the utilitygrid is not physically part of the CAESplant, this interaction should be consid-ered in designing the plant, since the de-sign (cost) of the plant can influenceutility usage (operating cycle). Conver-sely, the utility load cycle affects theplunt design (i.e., a coupling exists).For the purpose of design optimization theoverall system can be decomposed into threesubsystems (see Fig. 1). The first subsys-tem (subsystem 1) comprises the air com-pression train, the main piping and airdistribution system and the air storagereservoir.

Fig. 1. Typical CAES Power System

Subsystem 2 is the power generation train.The motor/generator and the utility gridare incorporated in the third group (sub-system 3). It is important to note thatthis particular decomposition is general,in the sense that it is not dependent uponthe internal design of any particular sub-system. Furthermore, it minimizes the num-ber of coupling variables. That is, theinteractions of subsystems 1 and 2 withsubsystem 3 are dependent on only one coup-ling "variable" — the utility load cycle.The interactions between subsystems 1 and2 (the ones of principle concern to theplant designer) are dependent on only threecoupling variables — the inlet pressure tothe power generation train (pt.), the spe-cific air mass flow rate (&') and the util-ity load cycle.

The criterion for optimum design ischosen to be the total normalized cost (C)of the system (i.e., cost per unit of elec-tricity generated by the CAES power plant).This total cost is the sum of the individualsubsystem normalized operating costs.* Thecosts have to be minimized subject to vari-ous performance and technical constraints.The implication for CAES plant design isthat, for a given utility load cycle, a sub-optimization of subsystem 1 would providethe minimum subsystem operating cost (c9)and values for the corresponding subsystemdesign variables, as a function of thecoupling variables — p t and &'. Similaroptimization for subsystem 2 would yield

Typically the normalized operating costsinclude fuel costs, maintenance, chargerate on capital investment, etc.

405

C? (the minimum operating cost of subsystem2) and its optimum design, as a function ofthe coupling variables only. Finally, thesum of c!jl and C? can be minimized by inspec-tion to determine the optimum values of thecoupling variables, the minimum plant cost(C*) and the optimal plant design. Theprocess can obviously be expanded (in prin-ciple) to include variations in the utilityload cycle and consideration of the resul-ting economic benefits or penalties to theutility. A noteworthy corollary of theapproach described is that changes in thedesign options considered (e.g., turbinesvs. piston expanders) in one subsystem donot invalidate the optimization results forthe other subsystem. The remainder of thispaper is confined to consideration of thedesign of a particular variety of subsystem1 — one with an aquifer reservoir.

AQUIFER RESERVOIRTECHNICAL CONSIDERATIONS

The design of aquifer reservoirs forCAES requires integration wif.h the charac-teristics of the aboveground machinery,piping, and the utility load cycle. Thedesign also depends greatly upon site-spe-cific geol gical properties like porosity,discovery pressure, permeability andthreshold pressure.3"* Some of these in situproperties enter into the flow performance;others impose design constraints. Compli-cations are introduced by way of distribu-ted flow resistance, formation heterogene-ities and possible two-phase flow of waterand air. Due to the complexities, however,aquifer reservoirs appear to offer a signi-ficant potential for economic optimization.

For an underground porous formationto be suitable for storing compressed air,it should have certain structural features.Suitable aquifers are usually in the ap-proximate! shape of an inverted saucer. Thetop consists of a tight porous caprock,saturated with water. The interfacialproperty of the air-water system in thesetight pores does not permit the flow ofair. Thus, the dome shape will preventany lateral or vertical migration of com-pressed air. The compressed air is con-tained in the pores of the rock betweenthe caprock qnd the bottom layer of waterand/or rock. In aquifers, the adjacentwater moves under an applied pressuregradient and therefore requires carefulmonitoring to ensure zero net movementover a period of time.

For the purpose of analysis, it helpsto make a distinciton between edge-waterand bottom-water reservoirs. Edge-waterreservoirs, shown in Fig. 2, are character-ized by relatively thin formations, a cap-rock of appreciable dip, an underlying im-permeable layer, and water driven to theedge of the field during bubble development.

Fig. 2. Edge-Water CAES AquiferReservoir

In bottom-water reservoirs, depicted inFig. 3, a water-air interface lies in anearly horizontal plane beneath the airbubble. This commonly occurs in thickformations. A characteristic unique tobottom-water reservoirs is the phenomenonof water coning. Because the bottom-waterreservoirs involve more design variablesand constraints, they have received thegreatest attention in the present optimi-zation study.

AQUIFER-RELATED DESIGN CONSTRAINTS

As related to CAES systems, potentialconstraints imposed by the aquifer charac-teristics have been discussed in Refs. 3and 4. These constraints were largelyidentified from experience in natural gasstorage. The DOE-funded work in progresson aquifer reservoir stability may resultin identification of additional ones. Theconstraints presently appearing to be im-portant for inclusion in the CAES designprocedures are as follows.

Air Bubble Sl^e. After growing the airbubble to the desired equilibrium size,further growth or shrinkage due to thedaily variations in pressure is to be nul-lified. This concern is reflected in tworelated constraints. First, the total massof air stored during a weekly cycle shouldequal the total mass removed. Second, for

406

TOP PROJECTION

Aid)- PROJECTED AREAOF AIR BUBBLE

°o°oo o o oooo o o

BOUNDARY OF THEACTIVE REGION ( A o c t )

%m INJECTION/WITHDRAWALWELLS

TE THAT PARTIALUTILIZATION OFRESERVOIR VOLUME

IS ALLOWED

H IS THE MEANHEIGHT OF THEDOTTEDFIGURE

Fig. 3.

IMPERVIOUS;APROCK

WATER LEVEL

Bottom-Water CAES AquiferReservoir

long-term constancy of bubble size, a pres-sure schedule having the average weeklypressure (corresponding to the mean massof air in the bubble during the week) equalto the aquifer discovery pressure should beadopted.

Charging Pressure. In operating CAESplants, no apparent advantage results fromusing high injection pressures; actually,there are economic benefits of injectingair at the minimum possible pressure com-patible with the reservoir dynamics andpower availability. However, during airbubble development, a high pressure willreduce the development time. An upperlimit on charging pressure is imposed toavoid exceeding either the caprockthreshold pressure or the overburden pres-sure.

Aquifer Geometry. It is obvious that thereservoir design and storage capacity mustbe compatible with the site-specific geom-etry of the formation. Contour maps forthe site enable the information needed for

optimization to be determined. This infor-mation includes the total bubble volume asa function of bubble thickness (measured attbe apex of the dome) and the spill point(maximum bubble thickness for which the airwill remain trapped by the caprock). Thecorresponding constraints are that the areaoccupied by the well-field must not exceedthe projected bubble area and the bubblesize must not exceed the spill point volume.

Water Coning. The problem of water coningin bottom-water reservoirs means that, forgiven reservoir conditions and well pene-tration depth, a critical flow rate of airexists above which air cannot be withdrawnfrom the reservoir without simultaneousproduction of water. The critical flowrate is extremely sensitive to in situreservoir heterogeneities. It is knownthat the presence of an impermeable bar-rier like a shale streak below the wellwould drastically inhibit bottom-waterfrom coning into the well. The phenomenonof water coning has been studied exten-sively in the past under the assumption ofsteady state flow, but an order of magni-tude estimation for CAES applications(short discharge time) indicates that anon-steady analysis is required to ade-quately determine the maximum well pene-tration that permits withdrawal of com-pressed air without co—production of water.Little attention has been given to thissituation in the literature. Therefore,the coning height is presently treated asa parameter in the design procedure. Theintention is to calculate the cost andperformance sensitivity of the aquifer sys-tem to this parameter. This will help es-tablish the priority to be assigned to thestudy of transient coning.

Well Spacing. It can be observed5 by con-sidering the dynamics of flow in porousmedia that, for a given charging or dis-charging time, a critical distance existsaround each well beyond which only a negli-gible amount of compressed air storage canoccur. This gives rise to an economic con-straint on maximum well spacing (i.e.,greater spacings would be wasteful of landand bubble volume). The critical spacingcan be calculated from a diffusion timeformula.3'5

AQUIFER FLOW MODELING

Due to the design requirements imposedby the system coupling variables (turbineinlet pressure and mass flow rate) and the

407

need to select a proper pressure ratio forthe air compressor train, a prediction ofthe formation pressure drop is needed inthe design optimization procedure. Basedon an extensive study of the modeling re-quirements for flow in porous radial discgeometry (as applied to CAES applicationsin edge-water aquifers),5 It can be reason-ably expected that a simple quasi-steadymodel will suffice. To allow considerationof bottom-water reservoirs (having a coningconstraint), as well as to insure that po-tential economic benefits of partially-penetrating (i.e., shallower) wells can beexamined by the optimization procedure, amore general, two-dimensional, version ofthe quasi-steady model2 is employed in thepresent study. In order to easily use theformation pressure drop equation (i.e., toemploy a single "typical well" model), theactive part of the actual dome-shaped res-ervoir has to be represented by an equiva-lent cylinder having the same projected areaand a height equal to the ratio of actualstorage volume to projected area. This in-formation is determined from-contour maps.

The amount of mass stored or removedduring a given charge or generation pro-cess is used, together with the void volumeof the active well-field within the bubble,to determine the change in mean formationpressure occurring during that process.Combining this information with the quasi-steady formation pressure drop predictionenables the maximum and minimum wellborepressures occurring during the week to befound. These values, in turn, are neededin assessing the compatibility between thereservoir design and the aboveground equip-ment.

The chief uncertainty in the flowmodeling just described is that it assumesvalues for the effective permeability andporosity of the porous medium are known.These values are influenced not only byheterogeneities in the rock, but also bythe distribution of water throughout theformation following bubble development andsubsequent dryout (to the extent it occurs).Although moisture effects have been con-sidered, 3 more work is required to resolvethe issues. As a reasonable measure, fordesign study purposes, the following sim-plifications are used. First, to accountfor the reduction of storage space becauseof moisture remaining after bubble develop-ment, a modified porosity has to be defined.It is recommended that, until more accurateinformation becomes available, the dry por-osity value, reduced by the connate water

saturation,4 be used in all calculations.Second, since, in a radial geometry, thepressure losses are concentrated around thewellbore, which should be relatively dry,it seems justifiable to use the dry perme-ability values in estimating the pressuredrop in the reservoir.

OPTIMAL DESIGN OF A CAESSUBSYSTEM WITH AQUIFER RESERVOIR

The decomposition concept describedearlier suggests that the aquifer reservoir,compressed air piping, and air compressiontrain should be designed concurrently as a •subsystem. This grouping has minimal in-teraction with the rest of the CAES system.It should be realized that any attempt todesign and optimize only part of this sub-system (namely, the aquifer well-fieldalone) would be less satisfactory and,possibly, misleading. The resulting "solu-tion" would be dependent on assumed valuesof parameters such as piping pressure dropsand would not directly allow the compres-sion costs to impact the reservoir design.

Site-specific reservoir design studiesfor CAES have been discussed in previousliterature. For example, a conceptual de-sign of a complete CAES plant using theBrookville aquifer as the reservoir wasconducted by General Electric.6 The designwas based on more or less state-of-the-artequipment and was used to test some generalconclusions concerning technical and econo-mic feasibility of compressed air storage.Also, Katz and Lady1* have analyzed (andpartially optimized) an aquifer and a reefsystem to illustrate a design philosophyfor CAES plants using underground porousmedia. The general techniques resultingfrom the present project should aid inconducting optimal design studies for CAESsystems in the future.

SUBSYSTEM PERFORMANCE MODELING ANDDESIGN CONSTRAINTS

The performance modeling and designconstraints associated with the aquiferwere discussed in an earlier section.These aspects have received the greatestattention because they are complex andreservoir costs are dominant in subsystem1. Rather simplified compressor train andpiping system performance models are usedin the present subsystem 1 design procedure.However, the incorporation of more detailedmodels would not alter its basic structure.

408

The design considerations are bestillustrated by reviewing a typical step inthe iterative search for the optimum CAESsubsystem design. The typical design stepincludes compression train design, basedon flow rate and pressure drop calculationsfor a charging process, and checking of theavailable turbine inlet pressure, based onpressure drop calculations for a powergeneration process.

First, the compressor train mass flowrate is calculated from the known turbineflow rate and ratio of weekly power genera-tion to storage time. This calculation in-corporates the non-growth constraint forthe air bubble and also assumes (for sim-plicity) that the mass flow rate duringevery charging period is the same. Next,the required compressor train dischargepressure is calculated by adding the pres-sure drops in the wells and compressed airpiping to the maximum wellbore bottom pres-sure, predicted with the aquifer model.The maximum wellbore pressure depends onthe weekly mass charging/discharging cycle,the wellbore diameter, depth of well pene-tration, well spacing, and number of wells.From knowledge of the compressor train dis-charge pressure and flow rate, and specifi-cation of the pressure ratio across thelow-pressure compressor* (either 11:1 or16:1), the total compression power is cal-culated from available data.3'5'7

The pressure difference from well-headto well-bottom reflects friction and grav-ity effects using standard relationships • 3 >il

The piping system friction pressure drop ispatterned after the simplifications employedby Katz.1* It is assumed that the majorityof the pressure drop in the surface pipingsystem occurs in the main pipeline and thatan equivalent pipe length (L) can be de-fined to account for pressure drops in thefeed, cross-feed and branch pipelines. Themost significant design variable of thepiping system is then the diameter of thismain pipeline. Standard relationships areused in these calculations.3 A 2% addi-tional pressure drop in the aftercooler isadded.

After the pressure drop analysis ofthe compression process, some similar pres-sure drop calculations are done for the

The compressor train is modeled as com-prising a low-pressure compressor, abooster compressor, and appropriate inter-coolers and aftercooler.

power generation process occurring at thetime of the week for which the mass stored(bubble pressure) is minimum. This proce-dure enables the determination of the mini-mum pressure available to the power genera-tion train for the design being considered.For a design to be acceptable, this pres-sure must be at least as high as the speci-fied inlet pressure.

In the compressor design stage de-scribed above, the total charging time wasused; it influenced the predicted chargingflow rate and power. It should be notedthat this charging time duration and powerlevel must be checked for compatibility withthe specified utility load cycle. An ideal-ized utility load cycle is shown in Fig. 4,together with the corresponding reservoirair storage cycle.

mmr , nKnaat•r MB PLANT

-.BMCWll MASS STOKED

\A_A • M S JISMD

MOMW TUBBKI K M B M T MUMMY R I M r M r U K W M S » . NOMMT

Fig. 4. Idealized Weekly Utility Loadand Air Storage Cycles

The power generation level and time sche-dule is considered invariant, reflectingthe power demand for which the CAES plantis to be designed. The excess power levelfor 'storage and its daily available timedurations, however, have maximum valuesbut these may not be entirely needed bythe CAES plant which is being designed.Since the compressor power and the timevariation in air storage over the week canboth influence the subsystem costs, thetradeoffs between the two* allowed by thepresent optimization procedure can lead topotential operating cost reduction.

The beginning and ending time for eachcharging process and the compression power(assumed uniform for simplicity) are allconsidered as design variables, subject tothe maximum value constraints imposed bythe utility.

409

SUBSYSTEM ECONOMICS:FUNCTION

THE OBJECTIVE

At every stage in the optimizationprocess, the trial design being consideredhas an associated set of costs. To put thecosts on a common basis, it was decided tominimize the total operating cost (per unitof power generated) attributable to subsys-tem 1. In the terminology of optimizationtheory, this operating cost is the objec-tive function to be minimized. It com-prises the annual carrying charge on capi-tal, subsystem operating and maintenance(O&M) costs, and the cost of compressionenergy (electricity) derived from the baseplant off-peak power. The specific capitalcosts included are:

(1) Main piping and distribution system -dependent on piping design and numberof wells.

(2) Wells - dependent on number, depth,and diameter.

(3) Land - for simplicity, assumed propor-tional to projected area of air bubble.

(4) Compressor train - based on data fromRef. 7.

(5) Bubble development - dependent onequilibrium bubble volume (air com-pression cost) .

The O&M cost is assumed proportional to thecapital charge cost. Further details onthe cost calculations and data are givenin Refs. 2 and 3. The design variables in-fluencing the various cost components aredepicted in Table 1.

Table 1. Subsystem 1 Cost Factors

DesignVariables

Utility LoadCycle

WellboreDiameter

Well Pene-tration

BubbleThickness

Number ofHells

Well Spacing

Main PipelineDiameter

h.V. Compres-sor Pressure

Cost Items

Wells

X

X

X

Land

X

PipinR

X

X

Com-pressor

X

X

X

X

X

X

X

X

BubbleDevelop-ment

X

Air Com-pression

X

X

X

X

X

X

X

X

OPTIMIZATION PROCESS

A detailed discussion of the nonlinearprogramming (optimization) algorithms, ortheir computer code implementations (OPT8

and BIAS9), which were employed in thisstudy, will not be given here. In essence,these generalized procedures interact withcomputer subroutine representations of thesubsystem performance and cost models, andthe constraint definitions, in order tofind that combination of design variableswhich minimizes the objective function andsatisfies all the design constraints.During the course of the search for theoptimum, many (e.g., hundreds) of trialdesigns are considered. The computer codesused work only on the continuous designvariables. Discrete variables (those re-stricted to only a few allowed values, suchas pipe diameters) must be examined "manu-ally" by repeated application of the com-puter code. In the present formulation,the number of wells is approximated as acontinuous variable, because it is typicallya large number (e.g., a few hundred). Thepresent CAES design optimization procedureresults in the specification of the follow-ing independent variables: air bubble size,number of wells, well depth, wellbore di-ameter, well spacing, compression (charging)time duration for each day, compressionratio of the low-pressure (L.P.) compressorand main piping diameter. Much additionalinformation can subsequently be derivedfrom these results (e.g., booster compres-sor pressure ratio, land area to be pur-chased, etc.).

ECONOMIC TPADEOFFS IN DESIGN

The number of design variables (re-lated to the flexibility of the model) re-quires the investigation of many tradeoffsduring optimization. Although some ofthese tradeoffs are perhaps obscured by the"automatic" nature of the optimization, theformulation of the procedure and operationalexperience have led to the identificationof several tradeoffs.

(1) Active Bubble vs. Total Bubble Size.Little incentive exists to sinking wellsnear the outer periphery of the reservoir.For a given bubble thickness at the apexof the dome, as the active land area in-creases, the average well thickness de-creases (see Fig. 3) so that the perfor-mance per well suffers and the number ofwells increases. However, fewer, deeper,wells concentrated near the center of thebubble results in development of a largely

410

Inactive bubble and In greater cost perwell.

(2) Well Penetration. Greater penetrationof wells into the bubble reduces the numberof wells, but increases the cost per well.

(3) Well Spaaing. Closely spaced wellshave less pressure drop (compression cost)but also less storage volume associatedwith each well.

(4) Bubble Thickness. Greater thicknesspermits deeper wells (fewer needed) but re-quires more surrounding land (projectedarea of bubble).

(5) Compression Time. Use of all thecharging time available minimizes the capi-tal cost of the compressor train (lowerflow rate). Use of reduced time (higherflow rate) can alter the shape of the res-ervoir mass storage cycle, reducing themaximum pressure swing. This could reducethe number of wells needed to meet the tur-bine inlet pressure requirements.

EXAMPLE APPLICATIONS

The CAES subsystem 1 design optimiza-tion procedure described in the precedingsections has been successfully implemented.Results of applying it are presented inthis section for illustration purposes, toexamine the potential economic impacts thatcan be achieved with aquifer reservoir sub-system optimization, and to examine theeffects of certain cost parameters.

Originally, it was planned to applythe new procedure to the design of a 600MW plant at Brookville, Illinois, so thatthe optimized design could be compared withthe G.E. design.6 In preparing to do this,however, it was noted that the G.E. designappears to violate the spill point con-straint for the Brookville aquifer site.That is, the storage volume encompassedby the G.E. Bvookville reservoir designexceeds that available above the spillpoint, as determined from contour maps ofthe aq1. ifer layer. In the Brookvillestudy, the actual site-specific properties(porosity, permeability, average aquiferthickness) were used, but the reservoirwas approximated as a constant thicknesscircular disc without water-related con-straints. Application of the proceduredeveloped in the present study, which at-temps to account for geometrical limita-tions more correctly, led to a design withabout 700 wells; 308 wells were recommendedin the G.E. report6 using the less restric-tive aquifer geometry assumption.

The major testing of the capabilitiesof the design optimization procedure hasbeen for the example of a hypothetical 600MW CAES plant using the Media, Illinois,Galesville aquifer as the reservoir. Con-tour maps and "material properties" forthis aquifer were taken from Ref. 4. Thegeometrical information on storage volumeand projected bubble area as a function ofbubble thickness, based on the contour maps,is tabulated in Ref. 3. Other pertinentparameter values used in the study aregiven in Table 2.

Table 2. Galesville Study Parameters

Aquifer DiscoveryPressure

Effective Porosity

Closure (top struc-ture to spill point)

Average HorizontalSand Permeability

Average VerticalSand Permeability

Specific Flow Rate

Turbine SystemInlet Pressure

Utility Cycle:Power GenerationPower GenerationTine

Max. CompressionPower

Max. CompressionTime

Storage Temperature

Base Plant Electri-city Cost

Land Cost

Other Cost and Sub-system Parameters

840 psla

14.3%

110 ft

448 md

354 md

10.4 lbm/kWh

750 psla

600 MM

5 days, 10 hrs/day

590 MW

6 days, 10 hrs/day(excludes Friday)

1 5 0 ^

15 mllls/kWh

$1200/acre

see Ref. 3

As a starting point, a feasible (butnonoptimal) design for the Galesville plantwas developed intuitively, although this isnot essential for the implementation of theoptimization procedure using the OPT8 orBIAS9 algorithms. Table 3 compares theinitial intuitive design with two optimizeddesigns. The first one of these is the re-sult of the formulation described in pre-vious sections. The second design was ob-tained by employing a further simplificationin which the charging time variables wereheld constant. In all these cases, thediscrete design variables were held fixedat the values: main pipeline diameter(48 in.), wellbore diameter (7 in.), L.P.compressor pressure ratio (11:1).

411

Table 3. Sample Galesville Study Results

Aquifer ReservoirSpecifications

Surface area to bebought (acres)

Active well-fieldarea (acres)

Air bubble thickness (ft)

Hell depth (ft)

Well spacing (ft)

Kuaber of wells

Systea Pressures. FlowRates and Powers

Minimum available turbinesystea inlet pressure (pal)

Total storage processtlae (hrs)

Air flow rate duringstorage processes (lbB/sec)

Compressor power required(MW)

Coapreaaor dischargepressure (pal)

System Coats

Land cost ($, millions)

Bubble development coat($, million*)

Hell construction cost($. millions)

Lou pressure compressorcost ($, millions)

Booster compressor cost($, millions)

TOTAL CAPITAL COST($, millions)

REDUCTION IN CAPITAL COST (%)

3ASE LOAD ELECTRICITY COST(mllls/kUh)

TOTAL SUBSYSTEM OPERATINGCOST (aills/kWh)

REDUCTION IN OPERATINGCOST (*)

InitialDesign

5895

2755

105.0

1385

467

700

775.5

59.4

1459

385

850.3

S. 843

6.818

73.379

4.642

4.455

101.59

-

11.45

24.25

OptlalzedDesign

2777

2038

79.9

1369

S30

402

750.0

51.4

1686

449

879.3

4.165

2.580

41.915

4.996

4.897

62.00

39.0

11.54

19.36

20.2

OptimizedFixedStorage

Tlae Design

2944

2944

81.7

1380

533

575

750.0

60.0

1444

384

874.6

4.417

2.779

60.190

4.619

4.511

79.97

21.3

11.53

21.60

10.9

There are many interesting observa-tions to be made from the results in Table3. Both of the optimum designs reduce thenumber of wells, average well depth andbubble size, indicating that the startingpoint was a case of overdesign. This con-clusion can also be drawn from a compari-son of available turbine system inlet pres-sures in the three designs. Thus, a care-fully formulated constrained optimizationproblem has allowed a reduction in "safetyfactors" required in an intuitive designprocess.

The oprimization also underlines thecompromise necessary between compressionpower requirements and capital costs of thereservoir system. Higher compressor powerand cost are tradeoffs for lower land, bub-ble development, and well construction costs.In the first optimum design, the weekly res-ervoir pressure variation is reduced by aneven distribution of air storage over theentire cycle. This is accomplished by re-ducing the weekend storage process dura-tions. On the other hand, the simplifiedoptimization, with fixed storage times,uses a larger active reservoir volume andreduced reservoir formation pressure drop(larger number of wells) to decrease thecyclic pressure fluctuation. A noteworthyfeature of the optimum design is partialutilization of the air bubble. This iscaused by the high cost of constructingadditional wells in the outer region ofthe bubble, where they yield only minimalbenefit due to the tapering of the aquiferformation.

The most important results are the re-ductions achieved in the subsystem 1 costs.The optimization procedure described hereinyielded a 39% lower capital cost and 20%smaller operating cost, compared to theinitial design! Restricting the storageprocesses to fixed values caused these im-provements to be only half as much. Al-though substantial design improvements havebeen made, further cost reductions are ex-pected as the optimization algorithms arefine tuned and the models improved.

Further optimization runs for theGalesville problem have been made, usingdifferent starting point designs, to deter-mine whether the "global" optimum has beenfound. The best of these solutions has anoperating cost of only 17.8 mills/kWh, areduction of 8% from the optimum value givenin Table 3. This design has only 252 wells,an active area of 1294 acres, a bubblethickness of 91.6 ft., and a 53 hr. chargingtime. Interestingly, the fixed chargingtime (60 hr.) version of this solution isvery similar in design and cost.

When a CAES plant using the MediaGalesville aquifer was investigated by Katzand Lady,1* they concluded "... use of 100input/output wells seem reasonable for fulldevelopment (600 MW)." This number, notbased on detailed optimization, is consider-ably less than found in the present study(252). The discrepency may be partiallydue to the imposition of the diffusion time-

412

related constraint on well spacing in theoptimization procedure. Whether that con-straint is conservative or the assumptionsof the previous investigators overestimatesthe reservoir flow capability under CAEScycling conditions remains to be determined.

Examination of the various Galesvilleoptimization runs, and those done for theBrookville site, shows that the optimumwells penetrate nearly to the bottom of thebubble, often being limited by the coningconstraint. Increasing the coning distanceparameter from 1 ft to 5 ft increases theoperating cost by about 1 mill/kWh, indica-ting that the coning problem should bestudied further.

For the Galesville problem, the effectof wellbore and main pipeline diameters onoptimum design operating cost are shown inFigs. 5 and 6, respectively. The main pipesize has little effect.

21.0

costs were varied, thus explaining thelinear relationships in the figures. Al-though this observation may not be of gen-eral validity, it would be comforting toknow that a CAES design would remain opti-mum if the cost of base-plant electricitywere to increase in the future!

19.00

18.75-

18.50-

18.25'

Mill PIPE DIMETER (IK.)

Fig. 6. Effect of Main Pipe Diameterin Galesville Problem

21.0'

2.0 4.0 6.0 8.0 10.0 12.0

HEIUORE DIAMETER (IN.)

Fig. 5. Effect of Wellbore Diameter inGalesville Problem

The effect of certain cost parameterson the optimum Galesville subsystem 1 opera-ting cost has also been investigated (seeFigs. 7-9). It was found that the optimumdesign variables did not change as these

27.0'10 20 JO «

KU.-DMU.IIK COST it/n.)

60

Fig. 7. Effect of Well-Drilling Cost-Galesville.

413

6 5000 10000 15000 20000 25000

LAND COST (S/ACRE)

Fig. 8. Effect of Land Cost - Galesville

20 40 60 80 100

ELECTRICITY COST (MILLS/KKH)

Fig. 9. Effect of Base Plant Electricity- Galesville

CONCLUDING REMARKS

The design procedure described inthis paper appears to be the most completemethod available for designing aquiferreservoir-based CAES plants. Limited com-parisons with published results using moresimplified methods of analysis suggests apossible inadequacy in those methods. Fur-ther work is recommended to resolve theseissues.

The design optimization procedure isgeneral in its structure, but its currentcomputer implementation is somewhat res-tricted (e.g., bottom-water reservoirs,equal compression power for each chargeprocess, etc.). It is also based on asomewhat idealized aquifer model and onparticular judgements on important con-straints. However, extensions and refine-ments can be readily incorporated as re-quired.

Utilization of the design optimizationprocedure can be valuable, when carefullyapplied. It can:

- result in actual capital and operatingcost savings in plant design,

- give insight into the economic trade-offs among design variables, and

- assess the influence of uncertaintiesin cost data.

Furthermore, if combined with a similaroptimization model for the turbine system,a complete CAES plant design optimizationcould be performed.

Some additional information on thework presented is available in Refs. 2 and3. A final report is in preparation whichwill provide full documentation, includinglistings of the computer subroutines embody-ing the optimization-oriented CAES model.

ACKNOWLEDGMENTS

The important contributions of AjaySharma (University of Illinois at ChicagoCircle), Rajesh Ahluwalia (Argonne NationalLaboratory), and Ken Ragsdell (PurdueUniversity) in the development and imple-mentation of the methods and models des-cribed herein, and in the preparation ofprevious manuscripts on the project, aregratefully acknowledged. This work wassupported by the Division of Energy StorageSystems, Office of Conservation, U.S.Department of Energy.

414

REFERENCES

'Kim, C.S., and G.T. Kartsounes, A Para-metric Analysis of Turbomaahinery Optionsfor Compressed Air Energy Storage Plants,Froc. of Compressed Air Energy StorageTechnology Symposium, Pacific Grove, Cal.(May 1978).

2Sharma, A., F.W. Ahrens, K.M. Ragsdell,R.K. Ahluwalia, and H.H. Chiu, Design ofOptimum Compressed Air Energy Storage Sys-tems, to be presented at the 1978 Midwes-tern Energy Conf., Chicago, Nov. 19-21,1978; to be published in ENERGY (journal).

3Ahluwalia, R.K., A. Shamra, and F.W.Ahrens, Design of Optimum Aquifer Reser-voirs for CAES Power Plants, Proc. ofCompressed Air Energy Storage TechnologySymposium, Pacific Grove, Cal. (May 1978).

"Katz, D.L., and E.R. Lady, Compressed AirStorage for Electric Power Generation,Ulrich's Books, Inc. Ann Arbor (1967).

5Ahluwalia, R.K., F.W. Ahrens, A. Sharma,H.H. Chiu, and G.T. Kartsounes, DyanmicAnalysis of Porous Media Reservoirs forCompressed Air Energy Storage, ArgonneNational Laboratory, unpublished infor-mation (1977).

6Bush, J.B., Jr., Principal Investigator,Economic and Technical Feasibility Studyof Compressed Air Storage, ERDA ReportNo. 76-76 (March 1976).

7Davison, W.R., and R.D. Lessard, Studyof Selected Turbomaahinery Components forCompressed Air Energy Storage Systems,prepared by United Technologies ResearchCenter for Argonne National Laboratory,Report ANL/EES-TM-14 (Nov. 1977).

"Gabriele, G.A., and K.M. Ragsdell, OPT;A Nonlinear Programming Code in Fortran-IV-Users Manual, Purdue Research Founda-tion (June 1976).

9Root, R.R., and K.M. Ragsdell, BIAS: Anonlinear Programming Code in Fortran-TV-Users Manual, Purdue Research Foundation(Sept. 1978).

415

EVALUATION OF TL1RBOMACHINERY FOR COMPRESSED AIRENERGY STORAGE PLANTS

George T. Kartsounes and Choong S. KimEnergy and Environmental Systems Division

Argonne National Laboratory9700 South Cass Avenue

Argonne, Illinois 60439

ABSTRACT

This paper presents a study of possible turbomachinery options for compressed airenergy storage plants. The plant is divided into four subsystems: a turbine system,compressor system, motor/generator, and an underground air storage reservoir. The tur-bine system comprises a high-pressure turbine and combustor, a low-pressure turbine andcombustor, and a recuperator. The compressor system comprises a low-pressure compressor,booster compressor, intercoolers, and an aftercooler. A water-compensated mined cavernconstitutes the underground air-storage reservoir. Plant performance is presented interms of five parameters: specific air flow rate, specific heat rate, specific storagevolume, specific compression rate, and overall plant efficiency. The capital and oper-ating costs of the plant as a function of the turbomachinery options are presented. De-sign variables of the turbomachinery are the reservoir pressure and inlet gas tempera-tures to the turbines.

INTRODUCTION

Compressed air energy storage is anear-term technology for the load levelingand peak shaving strategies being consi-dered by electric utilities. Assessmentsof the technical and economic feasibilityof this storage system indicate that it iseconomically competitive with conventionalgas-turbine peaker units. The CAES conceptis based on a split Brayton cycle with anaccompanying underground air storage res-ervoir. During periods of off-peak powerdemand, air is compressed with base-plantpower and stored in the underground reser-voir. For power generation, the air isdischarged through a combustion turbineduring the peak demand period.

Because the storage reservoir is usu-ally the most costly single component in aCAES plant, its volume is a sensitive de-sign parameter. The volume required isaffected by storage pressure and tempera-ture, power level, generation time, reser-voir type, air quantity required by theturbine system, and pressure ranges per-mitted by the turbomachinery (turbines andcompressors). Compressed air can be storedunderground in caverns or in the pore spaceof porous rock formations.

The components of the subsystems of aCAES plant are delineated here for preci-sion of reference in this paper. The tur-bine system consists of a low-pressure gasturbine (LGT) and combustor, a high-pres-sure gas turbine (HGT) and combustor, anda recuperator (see Fig. 1). The LGT is aturbine modified from a conventional gas-turbine peaker unit. For proposed CAESplants, the HGT is a modified steam turbineoperating at gas temperatures of about1000°F. Optimized designs for compressed-air turbines that operate at high tempera-tures have been investigated.1 The com-bustors can be designs modified from con-ventional gas-turbine peaker units. Pre-liminary studies indicate that recuperatorscan be designed that are economically fea-sible for CAES application. These differfrom conventional gas-turbine peaker unitsbecause of the high-pressure air leavingthe reservoir.

The compressor system contains a low-pressure (LC), high-pressure (HC), andbooster compressor (BC), intercoolers, andan aftercooler (see Fig. 1). Intercoolingis required to operate the compressorswithin limits tolerable for standard

417

COOLING AIR

T TMOTOR/GENERATOR

T TGENERATOR

RECUPERATOR

CAVERN (Po.To)

Fig. 1. Schematic Diagram of CAES Plant

materials. An aftercooler is used to coolthe air to avoid possible thermal-stressdamage to the storage reservoir.

The performance of a CAES plant canbe characterized in terms of four specificparameters and an overall plant efficiency:

• Specific air flow is the mass flowrate of air supplied to the turbinesystem per kilowatt power generated.It is the major factor in determiningthe size of the turbines, compressors,and air-storage reservoir.

• Specific heat rate is directly propor-tional to fuel consumption and isequal to the product of specific fuelconsumption and the lower heatingvalue of fuel. It therefore affectsthe operating cost of the turbines.

• Specific storage volume, the volumeof reservoir required per kilowatt ofpower generated, is dependent on thespecific air flow rate and the tem-perature of stored air.

• Specific compression rate is the ener-gy equivalent of the power supplied tothe compressors per kilowatt of powergenerated. This parameter is theamount of off-peak energy required tooperate the compressors.

• Overall plant efficiency is equal tothe total energy output from the tur-bines divided by the sum of the ener-gy input from the fuel and off-peakenergy to the compressor system.

The cost of a CAES plant can be char-acterized in terms of capital cost andoperating cost. Capital cost includes thedirect cost of the air storage facility,the turbomachinery, the balance of plant,and the indirect cost due to a contingencyallowance, engineering and administration,and escalation and interest during con-struction. The operating cost of the plantincludes the capital- charge, cost of fuelto the combustors, off-peak electricity tothe compressors, and operation and mainte-nance costs.

This paper presents a study of pos-sible turbomachinery options for CAESplants with particular emphasis on the tur-bine system. The performance and cost ofthe complete plant resulting from differ-ent turbomachinery options are presented.The turbine system design parameters con-sidered are the reservoir storage pressureand the inlet gas temperatures to the LGTand HGT. The LGT was based on a nominalpressure ratio of 16:1.* A water-compen-sated mined cavern was chosen as the com-pressed air storage reservoir.

Studies have indicated that pressure ra-tio has a minor effect on performance andconventional low-pressure turbines (frompeaker units) having a nominal pressureratio of 10-16:1 can be used. 2' 3

PERFORMANCE EVALUATION

THERMODYNAMIC ANALYSIS

A thermodyanmic analysis was carriedout on each subsystem of a CAES plant, andthe results were combined to evaluate over-all plant performance. Design parametersconsidered in the analysis include: airstorage pressure and inlet gas temperaturesto the high-pressure gas turbine and low-pressure gas turbine. The details of theanalysis are presented in Ref. 4.

Underground Air Storage System. The under-ground air storage reservoir considered isa water-compensated cavern. Therefore, thepressure variation in the cavern during theoperating cycle is negligible. The airtemperature of the storage cavern (To) wasassumed as 120°F (322°K) and four differentair storage pressures (p0) were consideredin the analysis: 30, 50, 70, and 100 atm(3 x 106, 5 x 106, 7 x 106, and 1 x 107 Pa) .

Turbine System. The selection of the tur-bine system (see Fig. 1) evolved from theresults of a previous study.2 The follow-ing values of system parameters were con-sidered:

Turbine efficiencies: nLGT = n H G T = 0.90,

Recuperator effectiveness: e = 0.8

Temperatures: T, = 1000°,1600°,2000°,2400°F(811°,1144°,1366°,1589°K)

T = 1600°,2000°,2400°F3 (1144°,1366°,1589°K),

Pressures: 16 atm (1.6 x 106 Pa).

Subscripts given in the above parameterscorrespond to the components or stationsin Fig. 1. The efficiencies of turbinesand combustors are based on state-of-the-art values of available equipment.1 Recup-erator effectiveness is a function of theheat exchanger specifications. Because thetemperature of the inlet gas to the tur-bines must be kept low enough to avoidthermal damage of the turbine blades andvanes, cooling air is required for higherinlet gas temperatures. The amount ofcooling air required was determined fromdata presented in Ref. 1.

Compressor System. The study was extendedto the compressor system in order to com-plete the analysis of the CAES plant. The

following parameters were assumed to beknown or specified.

Adiabatic efficiency of compressors:

^HC " \C m nBC = °-90;

Temperatures: T . = 77°F, T... = T,_ = T._

= 100°F, T l g » 120°F; and

Pressures: p... = 1 atm, p., = 16 atm.p.,

PERFORMANCE RESULTS

Results of the parametric study arepresented in terms of the five performanceparameters: specific air flow rate, spe-cific storage volume, specific heat rate,specific compression rate, and overallplant efficiency. These values are givenas a function of air storage pressure andinlet gas temperatures to the HGT and LGT.

Specific air flow rate is the flowrate of air coming out of the cavern perunit output of the turbine system. It isdirectly proportional to the turbine andcompressor sizes, and, thus, is an impor-tant factor in determining the cost of theabove-ground facility. A plot of specificair flow rate against air storage pressureat different turbine inlet gas temperatures(Fig. 2) shows that the air flow ratevaries from 6.6-12.0 lb/kWh (3.0-5.4 kg/kWh) for the conditions specified in thisstudy, and it decreases as air storagepressure'increases• In addition, higherturbine inlet gas temperatures result insmaller air flow rate, even though coolingair is required.

Specific storage volume, the requiredstorage cavern volume per unit work output,is directly related to the cost of the un-derground facility for a CAES plant. Thisstorage volume depends on the requiredspecific air flow rate as well as on cav-ern conditions, such as pressure and tem-perature of stored air. Consequently, re-sults for the specific storage volume showa trend similar to that for the specificair flow. Figure 3 shows the effects ofair storage pressure and turbine inlet gastemperatures on the storage volume. It isseen that smaller storage volume resultsfrom higher air storage pressure or higherturbine inlet gas temperatures. Specificstorage volume in this study varies from0.96 ft3/kWh (0.027 m3/kWh) to 5.84 ft3/kWh (0.162 mVkUh).

30 40 50 60 70 80 90 100

STORAGE PRESSURE, Po (aim or l()5 Pol

Fig. 2. Effect of Storage Pressureon Specific Air Flow Rate

*

£

"3H

3§u

STOR

6.c

5.0

4.0

3.0

2.0

1.0

0

r T3=IO0O*F(8ll«K)iT5c|600*(ll44«K)

v/ /V V«00#F(ll44»K)

7\/ TjsTjtaooo'FdJss'K)

Y v k / TjtT5«2400'F(IJ89*K)

\v\ / __V\NS.

— ^-^^s1 1 1 1 1 1

0.175

0.150

0.125

oioo _•

0.075 £

0.050

0.025

030 40 50 SO 70 80 90 100

STORAGE PRESSURE, Po (aim or I 0 5 Po)

Fig. 3. Effect of Storage Pressureon Specific Storage Volume

Specific heat rate is a measure ofpremium-fuel usage for the combustors perunit power output of the system. It variesin this study from 3700 Btu/kWh (3.98 x 106

J/kWh) to 4280 Btu/kWh (4.52 x 106 J/kWh).The effect of storage pressure on the heatrate is given at different turbine inletgas temperatures in Fig. 4: higher stor-age pressure results in lower heat rate.Figure 5 shows that heat rate increases asthe LGT inlet gas temperature increasesand that the HGT inlet gas temperature hasa minor effect on the heat rate.

3,70030 40 50 60 70 80 90 100

STORAGE PRESSURE, Po (aim or 105 Pa)

Fig. 4. Effect of Storage Pressureon Specific Heat Rate

4,200

4,100

T5 CK)

1200 1300 1400 1500

I I I IHGT INLET TEMPEMTURE, Tj=

4.4

4.0 -

3.9

1600 1800 2000 2200 2400

LGT INLET TEMPERATURE, T 5 C F )

Fig. 5. Effect of Turbine Inlet Tempera-tures on Specific Heat Rate

Specific compression rate is .the fuelequivalent of the off-peak electrical ener-gy input to the compressor system per unitpower output of turbine system. In thisstudy, specific compression rate is basedon an off-peak heat rate, including elec-trical and mechanical losses, of 10,400Btu/kWh (1.097 x 10 7 J/kWh). For the con-ditions of this study, the rate varies from5280 Btu/kWh (5.57 x 106 J/kWh) to 7790Btu/kWh (8.22 x 106 J/kWh). Figure 6 showsthat, in general, compression rate increasesslowly with increasing storage pressure andsmaller compression rate is required byhigher turbine inlet gas temperatures.

420

_ 9.000

II «.000

7,000

6,000 =

4,000

T3>KH>0aF((M>K)iTs'!(00aF(ll44>K)

_ 13*15.2400^ 0589'K)'

I I I I I I3 O 4 0 5 O 6 0 7 O S 0 9 0 100

STORAGE PRESSURE, Po (otm or 10$Pa)

Fig. 6. Effect of Storage Pressureon Specific Compression Rate

The overall plant efficiency, the ra-tio of turbine power output to the sum ofthe power input to the compressors and thepower equivalent of fuel energy, variesfrom 0.538-0.581 for the conditions speci-fied in this study. The effects on theoverall plant efficiency are given in Figs.7 and 8.

Figure 7 shows the effects of storagepressures on plant efficiency: (a) for T3- T 5 - 2400°F (1589°K) or T3 - T5 - 2000°F(1366°K), plant efficiency increases withthe storage pressure; (b) for T3 * T5 »1600cF (1144°K), plant efficiency increasesup to 70 atm (7 x 106 Pa) and then decreasesas the storage pressure further increases;and (c) for T 3 - 1000°F (811°K), T 5 - 1600°F(1144°K), plant efficiency decreases mono-tonically with storage pressure.

The effects of turbine inlet gas tem-peratures on plant efficiency are given inFig. 8. It shows that higher plant effi-ciency is obtainable with higher HGT inletgas temperature. It also shows that effi-ciency increases with the LGT inlet gas tem-perature for T3 - 2000°F (1366°K) or 2400°F(1589°K), and it has a minimum at about T5- 2000°F (1366°K) for T3 - 1000°F (811°K)or 1600°F (1144°K).

ECONOMIC ANALYSIS

An economic analysis of the CAES plantwas made to show the effects of the para-meters on capital and operating costs. Theanalysis was based on the performance re-sults described above. In order to pro"*''"

a reasonable basis for the economic analy-sis, the following operating cycle waschosen: 20-hr nominal cavern storage capa-city and 2190-hr/yr generation time.

7 i"*

fa.

6 i• • • *

3£S

0.59

0.58

i

0.57

0.56

0.55

H 0.54 _

I

Fig .

* T3=TS.

.24OO'F(B.»'K)

.

tl600*F(H44»K)

T5=2000»F{I366CK)

_ TjS|000*F<eil»K)i T5x«OO*F(ll44*K)

1 I 1 1 1 1' 30 40 SO 60 70 80 90 100

STORAGE PRESSURE. Po (otm or 10^ Po)

. 7. Effect of Storage Pressure onOverall Plant Efficiency

T5(»K)

0 5 9 1200 1300 1400 1500

1- 0.55

I I I TH6T INLET TEMPERATURE, T3=

2400*F (1989*10-

I6OO*F(II44*K)

IOOO*F(eil*K).

g 053 _ Po*70lt*(7ilO*Po)

I I I1600 1800 2000 2200 2400

LGT INLET TEMPERATURE, T 5 f F )

Fig. 8. Effect of Turbine Inlet Tempera-tures on Overall Plant Efficiency

421

CAPITAL COSTS OPERATING COSTS

Direct capital cost of the CAES plantwas divided into the following: cost ofunderground air storage cavern and water-compensating reservoir, cost of turbomach-inery equipment, and balance of the plant.

The storage cavern cost included thecost of the air and water shafts, cavity,development and mobilization, and comple-tion. The cost of the air and water shaftswas estimated based on the cavern depthwhich was determined by the air storagepressure. The cost of the cavity was esti-mated based on specific storage volume witha 10% capacity margin. Since the storagecavern considered in the analysis is water-compensated, the cost of the water reser-voir was also included. The storage-rela-ted, costs were based on Ref. 5.

Estimation of the turbomachinery costwas based on Ref. 1. In this reference,the selling price is estimated for 10, 20,and 50 units. Based on the evaluation ofregional markets and development potentialfor CAES conducted by Harza EngineeringCompany,6 a 50-unit selling price was usedin this study. The cost of the low-pres-sure gas turbine with a cycle-pressure ra-tio of 16:1 was determined by the inlet gastemperature and the cost of the high-pres-sure gas turbine was determined by both theinlet gas temperature and air storage pres-sure. Costs of the LC and HC with theoverall compression ratio of 1:16 were es-timated from the air flow rate, and thecost of BC was determined by the air flowrate and air storage pressure. A 25% al-lowance was given for the ducting and in-stallation of the turbomachinery equipment.

The remainder of the plant equipment,which includes the clutches, motor/genera-tor, recuperator, combustors, fuel storage,coolers, electrical power system, land, andplant structure was denoted as the balanceof plant. This equipment is relatively in-sensitive to CAES design parameters and afixed cost of $80/kW was used for the bal-ance of plant for all cases of this study.

Total capital cost of the plant wasestimated from the direct capital cost con-sidering the following allowances: 15% forcontingency, 10% for engineering and admin-istration, and 30% for escalation and in-terest during the construction period.

Operating cost of the CAES plant con-sists mainly of capital charge, cost offuel to the combustors, off-peak electri-city to the compressors, and operation andmaintenance. Annual capital charge wasestimated from the total capital cost basedon the fixed capital charge rate of 18% peryear. Estimation of the cost of premiumfuel was made by multiplying the specificheat rate by the cost of No. 6 oil. Costof tne off-peak electricity to the compres-sors was estimated from the specific com-pression rate and the electricity costfrom the base plant. A value of 2 mills/kWh was used as the cost of operating andmaintenance for all cases.

ECONOMIC RESULTS

Results of the economic study aregiven in terms of the two specific costs:capital cost ($/kW) and operating cost(mills/kWh). The values are presented asa function of the storage pressure (po)and the turbine inlet temperatures (T3and T5).

Capital Costs. Capital cost of a CAESplant varied from $285/kW to $406/kW forthe range of design parameters specifiedin the study. The cost of the undergroundstorage cavern was found to be the highestcomponent cost for most cases varying from26-46% of the total capital cost and thecost of the turbomachinery equipment variedfrom 16-31% of the total direct capitalcost. In general, it was found that higherturbine inlet temperatures result in higherturbomachinery cost.

Total capital cost is given in Fig. 9as a function of storage pressure for fourdifferent combinations of inlet gas tem-peratures to the HGT and LGT. Capitalcost sharply decreases with increasingstorage pressure for all the cases up to70 atm (7 x 106 Pa) and either slowly de-creases or increases thereafter. Higherturbine inlet temperatures result in lowercapital cost at low storage pressures, forexample 30 atm (3 x 106 Pa). However, atstorage pressures greater than 70 atm(7 x 106 Pa), higher turbine inlet tem-peratures result in higher capital cost.Among the cases considered in the study,the design parameters that result in thelowest capital cost are those when T3 = T5

= 1600°F (1144°K) and pn = 100 atm (1 x107 Pa).

422

I

COST

(J

IAPI

TAL

400

380

360

340

320

300

260

260

i

1_T5.T5.S

1

'1000*

f

hi

WCF

BUEAAfI'IWV 1

1

P(8II'K

Tj«T5.

IM6*K

'HiiilIII**

1

); Ts'l*(

MOO'F

« ^

i

»«F<

1569

|

1.44'K)

•K)

- -—

130 40 50 60 70 80 90 100

STORAGE PRESSURE, Po (atm or »5po)

Fig. 9. Effect of Turbine Optionson Capital Cost

The dotted curve in Fig. 9 representsthe cost of a plant using a modified steamturbine (%JT " 78Z) for the HGT, whichoperates at I000°F (811°K) inlet gas tem-perature. The solid curve for T 3 * 1000°F(811°F) and T 5 - 1600°F (1144°K) is basedon the assumption that the cost of thisnew, high-efficiency HGT (nngj " 90Z)would be the same as that of the modifiedsteam turbine. The actual cost of thisnew turbine would depend upon the develop-ment cost and the number of units sold.Thus, the actual high-efficiency costcurve should be somewhere between thesolid and dotted curves. The net resultis that the cost differences would be neg-ligible, therefore favoring the use of themodified steam turbine because of provenreliability and equipment availability.

Operating Costs. The operating cost of theCAES plant is given in Fig. 10 as a fun-tion of the design parameters. The costof premium fuel was selected as $2.50/106

Btu and the electricity cost was 15 mills/kWh. In this figure the operating costvaries from 44.8-55.5 mills/kWh.

The capital charge was found to bemuch higher than the cost of fuel or elec-tricity; it amounts to 52-60% of the total

operating cost. Consequently, the opera-ting cost in Fig. 10 shows a similar trendto that of the capital cost. The figureshows that the operating cost decreaseswith increasing air storage pressure for allcases but T3 - T5 - 2400°F (1589°K), whichhas a minimum at about 70 atm (7 x 10 6).It also shows that, among the cases studied,the lowest operating cost results when To »T 5 • 1600°F (1144°K) for p o > 58 atm (5.8 x106 Pa) and T3 • T 5 - 2400°F (1580°K) forp 0 < 58 atm (5.8 x 106 Fa). However, in thepressure range of 50-90 atm, which is themost likely range for CAES with a water-com-pensated reservoir, the difference in opera-ting cost between different turbine systemsIs less than about 3 mills/kWh.

* 92 LL \\T3*l000'F(8ll*K)iT9c|(00*F(ll44*K)

30 40 SO 60 70 80 90 100

STORAGE PRESSURE, Po (otm or I05 Pa)

Fig. 10. Effect of Turbine Optionson Operating Cost

Figure 10 also illustrates that forT3 = 1000°F (811°K) and T5 = 1600°F (1144°K)the difference in operating cost between amodified steam turbine for the HGT and anew, high-efficiency design is negligible;i.e., less than 1 roill/kWh. Thus, the useof a modified steam turbine would be fa-vored because of proven reliability andequipment availability.

Effect of Electricity and Fuel Costs onOperating Costs. Table 1 illustrates theeffect of different electricity and pre-mium fuel costs on the overall operatingcost of a CAES plant. Two plant designsare compared: Plant A where T3 = 1000°F

423

Table 1. Effect of Electricity and Fuel Costs on Operating Costs

ElectricityCost (mills/kWh)

151515202020252525

17tta1 fV^afr

($/106 Btu)

2.503.755.002.503.755.002.503.755.00

Operating Cost (mills/kWh)

Plant A a

47.452.256.951.155.860.054.759.564.2

Plant Bb

45.350.054.748.553.358.051.856.661.3

X Decrease0

4.44.23.95.14.54.35.34.94.5

turbine inlet temperatures: T 3 - 1000°F (811°K> (rw,_ - 78X)

T 5 - 1600°F (1144°K). W 1

Turbine inlet temperatures: T 3 - T 5 - 1600°F (1144°K),C100(Plant A - Plant B)/Plant A.

(811°K) (TTHQJ) - 78%) and T 5 - 1600°F(1144°K), and Plant B where T 3 - T 5 -1600°F. These two plants bracket thehighest and lowest estimated operatingcosts.

From this table it is seen that forfixed electricity cost, the difference be-tween plant designs decreases as the fuelcost increases. This means that if fuelcosts increase faster than electricitycosts, then the type of plant design (i.e.,the selection of turbine system) becomesless significant. This scenario wouldfavor using Plant A because of provenreliability and equipment availability.

For fixed fuel cost, the differencebetween the operating costs of the two de-signs increases as the electricity cost in-creases. This means that if base plantpower increases in cost at a faster ratethan fuel costs, then Plant B would befavored. In this case, the use of a new,high-efficiency HGT would be justified.

CONCLUSIONS

This paper has considered the perfor-mance and cost of possible turbomachineryoptions for CAES power plants. Particularemphasis was directed toward the turbinesystem of the plant. The main design vari-ables were the reservoir storage pressureand the turbine inlet gas temperatures. Awater-compensated mined cavern was selectedas the storage reservoir. The results ofthis study should be applicable to the

other two reservoir types (i.e., aquiferreservoirs and salt caverns), but furtherstudy is recommended to fully evaluate theaffect of reservoir type on CAES plant per-formance and cost.

From the performance analysis, thefollowing trends were observed:

1. Specific air flow rate and storagevolume decrease as p Q, T,, or Tjincreases.

Specific heat rate decreases as p oincreases and increases as Tj in-creases; but is relatively insensi-tive to T,.

Specific compression rate, in general,slightly increases as p o increases;it decreases with increasing T3 or T5.

In general, overall plant efficiencyincreases as T 3 increases; is onlyweakly affected by p or T-.

4.

From the above, it can be concluded thatoptimum performance results from the useof a high storage pressure and high inletgas temperature to both turbines.

The economic analysis, however, illus-trates that minimum cost (capital andoperating) does not necessarily correspondto optimum plant performance. Consideringthe specific operating cost (i.e., mills/kWh), which can be considered the trueindicator of plant cost, at storage pres-

424

aures below about 60 atm, the highest tem-perature turbine system considered in thisstudy (i.e., T 3 = T 5 = 2400°F (1589°K)) re-sults in the lowest cost; whereas, above60 atm, the turbine system with 1 3 = 1 5 =1600°F (1144°K) results in the lowest cost.

A significant result is that for thepressure range of 50-90 atm, which is therange of present interest for water-com-pensated caverns, the operating costs forall of the turbine systems considered inthis study are within about 3 mills/kWh ofeach other; the average cost is about 47mills/kWh. Furthermore, it was observedthat if the cost of premium fuel increasesas a faster rate than the cost of basepower electricity, which seems to be alogical scenario for the future, the costdifference between turbine systems de-creases .

The economic study indicated that fora turbine system with T, = 1000°F (811°K)and T5 = 1600°F (1144°K), the use of a new,high-efficiency, high-pressure turbinecould not be justified and a modifiedsteam turbine could be used with littlecost penalty.

Based on the above factors, theoverall conclusion of this study is thatthe turbine system can be constructedusing available turbines with proven reli-ability without significantly sacrificingcost. The HGT can be a modified steam tur-bine and the LGT can be obtained from apeaker unit which operates at an inlet gastemperature of about 1600°F (or lower),requiring little, if any, cooling air.Interestingly, this is the approach beingused at the Huntorf Plant,7 which is theworld's first CAES plant.

ACKNOWLEDGMENTS

The research activities in compressedair energy storage, which formed the basisof this paper, were funded by the Divisionof Energy Storage Systems, Office ofConservation, U.S. Department of Energy.

NOMENCLATURE

E^ Specific compressic • .• ..e

A1 Specific air flow rate

p Pressure

Q' Specific heat rate

T Temperature

V's

e

noverall

Specific storage volume

Recuperator effectiveness

Efficiency

Overall plant efficiency

Subscripts

BC

ClC2HGT

LGT

HC

LC

0-19

Booster compressor

Combustor 1

Combustpr 2

High-pressure gas turbine

Low-pressure gas turbine

High-pressure compressor

Low-pressure compressor

Correspond to Fig. 1

REFERENCES

Davidson, W.R., and R.D. Lessard, Studyof Selected Turbomachinery Components forCompressed Air Energy Storage Systems,prepared by United Technologies ResearchCenter for Argonne National Laboratory,Report ANL/EES-TM-14 (Nov. 1977).

2Kartsounes, G.T., Evaluation of TurbineSystems for Compressed Air Energy StoragePlants, Argonne National LaboratoryReport ANL/ES-59 (1976).

3Kim, C.S., and G.T. Kartsounes, A Para-metric Study of Turbine Systems for Com-pressed Air Energy Storage Plants,Argonne National Laboratory Report ANL/ES-64 (April 1978).

"•Kim, C.S., and G.T. Kartsounes, A Para-metrio Analysis of Turbomaahinery Optionsfor Compressed Air Energy Storage Plants,Proc. of the 1978 Compressed Air EnergyStorage Technology Symposium (May 1978) .

5Giramonti, A.J., Preliminary FeasibilityEvaluation of Compressed Air StoragePower Plants, United Technologies ResearchCenter, R76-952161-5 (Dec. 1976).

^Underground Pumped Hydro Storage and Com-pressed Air Energy Storage: An Analysisof Regional Mcwhets and Devsloix-mt Poten-tial, prepared by Harza Engineering Co.for Argonne National Laboratory, ArgonneReport ANL-K-77-3485-1 (March 1977).

7Stys, Z.S., Air Storage System EnergyTransfer (ASSET) - Huntorf Experienae,ERDA/EPRI CAES Workshop (Dec. 1975).

425

EVALUATION OF THE USE OF RECIPROCATINGENGINES IN COMPRESSED AIR ENERGY STORAGE PLANTS

George T. Kartsounes and James G. DaleyEnergy and Environmental Systems Division

Argonne National Laboratory9700 South Cass Avenue

Argonne, Illinois 60439

ABSTRACT

The application of reciprocating engines to compressed air energy storage (CAES)plants is presentee! in this paper. The expected advantages compared to plants using tur-bines and compressors are reduced reservoir size and cost, reduced compression energy,and increased overall plant efficiency. The performance of possible engine and plant con-figurations are presented. One configuration uses a reversible, reciprocating expander/compressor engine. Power generation results from engine operation as an internal-combus-tion expander; compression is accomplished using the same engine operating as a recipro-cating compressor. Another possible configuration results when an internal-combustionengine is used as a high-pressure expander and a gas turbine is used as a low-pressureexpander. Compression is accomplished using either separate turbocompressors or opera-ting the high-pressure expander as a reversible-reciprocating compressor in series with" low-pressure turbocompressor. Capital and operating costs of plants using reciproca-ting engines are estimated and compared with that of turbine-based CAES plant designs.It is shown that using reciprocating engines can reduce capital and operating costs byabout 11% and 8£, respectively, compared to a plant using available turbomachinery.

INTRODUCTION

Electric utilities commonly use die-sel/generator sets or gas turbine/compres-sor units for peak power generation. Theuse of a gas turbine offers advantages inhigher power ranges due to its compara-tively smaller size and lower maintenancerequirements; however, many applicationsfavor the better part-load efficiency ofa diesel engine.

Another feature of diesel engine oper-ation that has significance for CAES sys-tems is a greatly reduced air flow require-ment compared to gas turbines. This re-duced need for air is related to the maxi-mum combustion-gas temperature. In eithera gas turbine system or diesel engine, thepower output equals the product of the flowrate of combustion products and the changein specific enthalpy across the engine.This enthalpy change is approximately pro-portional to the maximum absolute tempera-ture. In a gas turbine, the. maximum gastemperature is about 1350"K due to mater-ial limitations, although considerableresearch is in progress to eventually per-mit temperatures as high as 1900°K. Adiesel engine can have an instantaneous

combustion temperature approaching 3000°K.Thus, although the power output of a die-sel engine and gas turbine/compressor maybe equivalent, the higher combustion tem-perature of the diesel causes a largerenthalpy change, hence a lower requiredair flow. This reduction in air flow ratewould benefit a CAES plant through reducedreservoir size and reduced electricalenergy needed for air compression.

A conventional diesel engine cannotbe used directly in a CAES plant becauseit would not be possible to utilize thehigh-pressure reservoir air. However, areciprocating engine similar to a dieselwhich decouples the compression processfrom expansion can be used. Power genera-tion results from engine operation as aninternal-combrstion expander and compres-sion is accomplished using the same engineoperating as a reciprocating compressor.A basic description of the operatingcharacteristics and design features of thereciprocating expander/compressor engineis described in Ref. 1.

427

PLANT DESIGN

Two possible plant configurationsusing reciprocating engines are addressedin this paper. One design uses a rever-sible, two-stage expander/compressor con-cept and the other uses a compound enginecomprising reciprocating expander/compres-sor engines for high-pressure duty and agas turbine for low-pressure expansion.

PLANT USING REVERSIBLE, EXPANDER/COMPRES-SOR ENGINES

A schematic diagram of a CAES plantusing expander/compressor engines is pre-sented in Fig. 1. In this arrangement, asingle high-pressure (HP) engine is con-nected through a shaft to a single low-pressure (LP) engine. Both engines arethen connected to a reversible motor/gen-erator (M/G). In this practice, the HPengine and LP engine may each consist ofseveral separate engines that are mani-folded together. M/G units may be con-nected to each separate engine or a combi-nation of engines.

Fig. 1. Schematic Diagram of a CAES PlantUsing Expander/Compressor Engines

During the compression mode of opera-tion, valves C, D, E and G are closed. TheM/G operates as a motor receiving off-peakpower from a base-load power plant, and theLP and HP engines operate as reciprocatingcompressors. Ambient air is compressed inthe LP compressor and the heat of compres-sion is removed in the intercooler. Theair is further compressed in the HP com-pressor and then cooled in the aftercoolerbefore delivery to the storage reservoir.

Cooling water is circulated throughthe water jacket of each engine removingpart of the heat of compression. This heatremoval reduces compression energy sincethe compression process approaches isother-mal compression. Heated water leaving theengines is then coded in a radiator (e.g.,cooling tower).

In the expansion (i.e., power genera-tion) mode of operation, valves B, F and Hare closed. The M/G operates as a genera-tor, and the LP and HP engines operate asexpanders. Air from the reservoir is pre-heated in the recuperator using exhaust gasfrom the LP expander. Preheated air andinjected fuel are burned and expanded inthe HP expander. The products of combus-tion are further expanded in the LP expan-der and then flow through the recuperator.The cooling water removes part of the heatof combustion. This energy loss is neces-sary to reduce engine metal temperaturesand to improve engine reliability.

Comparing the plant configuration ofFig. 1 with that of a turbine-based CAESplant, the HP expander/compressor enginereplaces the expansion turbine and combus-tor, and the centrifugal booster compres-sors; the I.P expander/compressor engine re-places the low-pressure turbine and combus-tor, and the axial compressor. Furthermore,the intercooler and aftercooler are smallerthan that of a conventional plant becausepart of the heat of compression is removedin the cooling water. However, enginecooling water requires an additional cir-culation pump and radiator.

PLANT USING COMPOUND ENGINES

The plant configuration previouslyconsidered is based on the use of expander/compressor engines for both high- and low-pressure functions. However, a compoundengine design which combines expander/com-pressor engines with conventional turbo-machinery offers improvement over eitherall-turbine or all-reciprocating enginedesigns. This configuration is depictedin Fig. 2.

In this configuration, the high-pres-sure turbine and combustor, and boostercompressors of a conventional CAES plantare replaced with a reversible HP expander/compressor engine. A low-pressure turbineand axial compressor are used for the low-pressure duty cycle of the plant. However,since the exhaust of the HP expander will

428

be sufficiently high (e.g., 1360°K), a low-pressure combustor will be unnecessary.

Fig. 2. Schematic Diagram of a CAES PlantUsing Compound Equipment

During the compression mode of opera-tion, valves C, D and G are closed, bothM/G units operate as motors receiving powerfrom a base plant, coupling B is disconnec-ted, and the HP engine operates a compres-sor. During expansion, valves B, F and Eare closed, both M/G units operate as gen-erators, coupling A is disconnected, andthe HP engine operates as an expander. Inboth modes of operation, cooling water cir-culates through the HP engine.

This plant configuration will haveessentially the same main air flow rate asthat of a plant using reversible HP and LPengines with some additional turbinecooling air possibly being required. Thus,by using compound equipment, the size ofthe reservoir can be significantly reducedleading to significant savings.

However, the possible cost advantageof reversible equipment cannot be entirelyrealized with this configuration. The costadvantages of completely reversible equip-ment could be realized using a HP expander/compressor engine and a reversible LP tur-bine/compressor. The latter equipmentcould be a radial-bladed, centrifugal,single or multi-stage compressor which canbe reversed to function as a radial turbine.

ALTERNATIVE COMPOUND ENGINE PLANT

A possible alternative to the engine/turbine system depicted in Fig. 2 involves

using an additional heat exchanger betweenthe air inlet and the combustion gas outletof the HP reciprocating engine. This heatexchanger would lower the temperature ofthe gas entering the LP turbine and in-crease the temperature of the air enteringthe HP engine. Because of the large tem-perature difference between the fluidstreams, the required effectiveness of thisheat exchanger would be low (estimated tobe about 0.4) and hence it would be inex-pensive. The main advantage of this plantdesign is a reduction in reservoir sizewhich results because turbine cooling aircan be reduced (or eliminated). In addi-tion, simpler, less-expensive piping wouldbe required to transport the combustiongas between the engine and turbine. Thedisadvantage would be potential thermalproblems in the engine inlet valve due tothe higher air temperature.

PERFORMANCE EVALUATION

The proposed reciprocating engineswould function only as a single componentof a CAES plant. Therefore, engine perfor-mance will be given in the context of over-all plant operation as well as in terms ofengine efficiency, temperature, and pres-sure when operating as a power generator orcompressor. Comparisons will be made ofCAES plants having the proposed engines,the Huntorf facility,2 which is the world'sfirst CAES plant, and two other turbine-based designs — one near-term and theother an advanced design.3

MEASURES OF PERFORMANCE

The following parameters were selectedto measure CAES plant performance:

• Specific Air Flow Rate - the amount ofreservoir air required per unit ofgenerated power.

• Specific Heat Rate - the product ofspecific fuel consumption and thelower heating value of the fuel.

• Expansion Power/Compression Power

• Overall Plant Efficiency - plant out-put per total energy input.

THERMODYNAMIC ANALYSIS

A thermodynamic analysis was conductedof the plant schematically illustrated inFig. I. : The analysis considered twotypes of power generation — constant pres-sure and pressure-limited combustion.

429

The analysis was extended to considerplant performance using compound engines.The performance of the turbines and com-pressors was based on Ref. 3.

Simplifying assumptions used in theanalysis include:

(1) thermodynamic equilibrium during com-bustion,

(2) reservoir pressure remains constantat 4.48 MPa (this condition would beapproximated by a water-compensatedreservoir),

(3) reservoir temperature equal to 322°K,

(4) negligible pressure losses in piping,

(5) negligible heat transfer in systemPiping,

(6) recuperator effectiveness of 0.8,

(7) fuel being octane (C8H16) with alower heating value of 44.4 MJ(19,000 Btu/lbm), and kg

(8) turbine efficiency of 90%.

DISCUSSION OF PERFORMANCE

The three turbine-based CAES plantdesigns selected for performance compari-son with the proposed reciprocating engineconcepts are listed in Table 1. TheHuntorf plant is included for the obviousreason that it is the world's only opera-tional CAES plant. The principal reasonthat the performance of the Huntorf plantis lower than the two other systems is thatpresently the plant does not include a re-cuperator. Addition of a recuperator wouldbring all system performance parameterscloser to those of the near-term turbinesystem.

The near-term system can be imple-mented using conventional turbomachinery.In this system, a modified steam turbinecan be used for the 1000°F HP turbine. The1600°F LP turbine can be a modified gasturbine from a peaker unit. This turbinewill not require vane or blade cooling butwill require some auxiliary cooling for thedisk, blade and vane attachments, air seals,and bearing buffers.

The adva'noed-turbine system will re-quire significant air cooling of the HPturbine to operate at a gas inlet tempera-ture of 2400°F; the LP turbine will be un-cooled. The two turbines for this systemdo not exist but represent attainable

extensions to the state-of-the-art of gasturbine technology.

The performance of both reciprocatingengine combustion processes is seen to bevery similar. The chief difference betweenthe two processes is the higher maximumoperating temperature and pressures duringpressure-limited combustion.

It is seen that the reciprocatingengine concepts have significantly lowerspecific air flow rates than the turbine-based designs. This situation is reflec-ted in the favorable values shown for theratio of expansion power to compressionpower and overall plant efficiency. Sev-eral factors enter into the calculated lowspecific air flow value for the engine,with the most important of these being theair/fuel ratio. Actual air/fuel values willbe dictated by engine operation — that is,by the need to limit emissions or to lowertemperatures. Other influences on specificair flow are shown in Figs. 3 and 4. Theimportant influence of heat loss to cylin-der-jacket cooling water is shown in Fig.4. Cylinder heat transfer in the proposedengine will be quite different from conven-tional engines due to the use of pressur-ized reservoir air. Actual values of heattransfer will need to be measured experi-mentally, although improved methods ofanalysis can greatly reduce the uncertaintyin the value of this parameter. The in-fluence of reservoir pressure on air flowis shown in Fig. 3. As can be seen, higherreservoir pressure leads to reduced flowrate, although the change will be only5-7% over the range of expected reservoirpressures.

The plant heat rate is lower for thereciprocating engine concepts than the un-recuperated Huntorf design, but higher thaneither of the two turbine-based designs.This higher heat rate is partially causedby the higher operating temperatures of thereciprocating engines. The advanced, high-temperature turbine system has a higherheat rate than the low-temperature systemfor the same reason; that is, fuel is re-quired to raise combustion temperatures.The incentive for using this turbine de-sign is to improve all other performanceparameters. The heat rate is shown to bestrongly influenced by cylinder heat trans-fer in Fig. 4. High reservoir pressure isseen to provide lower heat rates in Fig. 3.This result is due to the additional energyprovided by the high-pressure air.

430

Table 1. CAES Plant Performance Comparisons

HuntorfPlant

Reservoir Pressure, MPa 4.48C

(psla) (650)

Specific Air Flow Rate, kg/MJ 1.46(lbn/kWh) (11.59)

Specific Heat Rate, J / J 1.61(Btu/kWh) (5500)

Expansion Power/Compression 1.24

Overall Efficiency, Z P ° " e r 41.5

Inlet temperature to HP and LPturbine, respectively.

2000°F in le t gas temperature.

Inle t pressure to HP turbine.

Plant UsingNear-Term

Plant UsingAdvanced

Turbine System Turbine System(1000,1600°F)a I

4.48(650)

1.42(11.2)

1.12(3808)

1.38

54.2

EngineMax.

Max.

[1500,2400°F)a

4.48(650)1.32

(10.4)1.35

(4600)1.65

51

Characteristics:, temp., K

(°F), pressure, MPa

(psia)Air/fuel ratio

Plant UsingConstant-Pressure

CombustionEngines

4.48(650)

.656(5.21)1.46

(4972)3.30

56.8

2136(3885)4.48(650)20

Plant UsingPressure-Limited

CombustionEngines

4.48(650)

.659(5.23)1.46

(4994)3.29

56.6

2552(4134)13.8

(2000)20

Plant UsingCompoundEngines

HP EngineLP Turblneb

4.48(650)

.914(7.26)1.41

(4805)2.37

54.4

1995(3130)4.48(650)22.6

0.74

0.72

CONSTANT-PRESSURE COMBUSTION- A /F«20

HEAT L0SS-25M.HV

0.62 -

Fig. 3.

1.60

4 6 8 10RESERVOIR PRESSURE, MPa

1.30

Influence of Reservoir Pressureon Specific Air Flow Rate andSpecific Heat Rate

0.72

0.68

i~0.64o

_ji.

£ 0 . 6 0

u

| 0 . 5 6

EO

0.52

run

CONSTANT-PRESSURE~ A / F » 2 0

COMBUSTION /_

RESERVOIR PRESSURE»4.48 M P o / /

-

1 1 1

S / ~

/ /

/

-

EXPECTED— VARIABLE —

RANGE—

1 1 1

- 1 . 5

- 1.4

1.3 £e

i.2 i

1.1 £v>

- 1 0

4 8 12 16 20 24 28HEAT LOSS TO JACKET WATER, %LHV

0.9

Fig. 4. Influence of Jacket Water HeatLoss on Specific Air Flow Rateand Specific Heat Rate

Figure 5 is presented to indicateoperational flexibility possible using thepressure-limited combustion process. Itis seen that a constant maximum pressurecan be maintained over a wide range ofreservoir pressure. The penalty is an in-crease in maximum engine temperature andheat rate. This suggests that the enginecould be used with aquifer or salt-cavernstorage to provide constant plant outputwith varying reservoir pressure by con-trolling the fuel injection rate.

Throttling of the air flow would be elimi-nated and the reservoir size would be re-duced.

The use of a compound engine is seento offer performance advantages over theuse of an all-turbine system. As would beexpected, plant performance using the com-pound engine falls between the values usingthe all-turbine or all-reciprocating system.The compound engine operates at a signifi-cantly lower specific air flow rate than

431

any of the turbine systems, which resultsin a higher power ratio and overall plantefficiency.

.3000

I12800

5g2600

£2400

2200

PRESSURE-LIMITED COMBUSTIONMAXIMUM PRESSURE. 13.6 MPoA/Ft 20HEAT LOSS'20% IHV

4 6 6RESERVOIR PRESSURE, MPo

10

Fig. 5. Effect of Reservoir Pressure onMaximum Engine Temperature

PRELIMINARY COST ESTIMATES

COST OF COMPOUND ENGINE

Performance and cost data of a HP tur-bine, LP turbine, and reciprocating enginethat can be combined to function as a tur-bine system or compound engine for a CAESplant are presented in Table 2. The tur-bines were selected to match the flow re-quirements of the reciprocating engine.

The compound engine would consist ofeight Delaval Model RV-20-4 diesel enginesoperating in parallel and feeding the LPturbine. This model is a large, slowspeed (450 rpm) engine designed for sta-tionary power or marine application. Thecost of $46/kW listed in Table 2 includesall auxiliaries and accessories for a stan-dard installation and will, therefore,yield a conservative cost estimate sincemany standard items (such as intercoolersand turbochargers) will not be needed inthis particular application. In addition,some reduction in price should result fromlarge-quantity production. In fact, theestimate may be even more conservativesince the engine eventually chosen for thisapplication may be a lower-priced, compact,high-production engine of less rugged de-sign such as used in diesel locomotives.

The cost of the compressor system tosupply about 375 kg/sec (827 lb/sec) of airat 5.92 MPa (865 psia) to the storage res-ervoir is estimated from Ref. 4 to be $18/kW of power input. Assuming an expansionpower to compression power ratio of 2.37(i.e., the value listed in Table 1), thetotal cost of the expansion/compressionequipment for the compound-engine conceptcan be calculated as $41/kW. However,this estimate does not consider the pos-sible cost reduction using the engines asreversible, HP reciprocating compressors.From Ref. 3, the expansion power to com-pression power ratio for the all-turbinesystem can be calculated as 1.67. Usingthis value, the total cost of the expan-sion/compression equipment for the all-turbine system can be calculated as $38/kW.

Table 2. Performance and Cost of Expanders

Inlet Pressure, MPa (psia)

Inlet or MaximumTemperature, K (°F)

Total Mass Flow (includingcooling air), kg/sec(lb/sec)

Net Power Output, MW

Cost, $/kW

HPTurbine

5.92(865)

1367(2000)

296(653)

112

41

LPTurbine

1.15(167)

1367(2000)

375(827)

200

19

HPReciprocating

Enginea

5.92(865)

2100(3324)

296(653)

238

46

8 - Delaval Model RV-20-4 Diesel Engines

432

COST OF REVERSIBLE-RECIPROCATING ENGINES

As previously discussed, a CAES plantusing reversible-reciprocating engines forpower generation and compression could bedesigned (Fig. 1). It is demonstrated inRef. 1 that four LP expander engines wouldbe needed to match the volume flow of eachHP expander. Therefore, using the Delavalengines, power generation would requirethirty-two LP engines for every eight HPengines. The LP engines would not requirefuel injection and thus should be slightlylower in cost than the HP engines. Assu-ming the same power output as the compoundengine (i.e., 238 MW for HP expansion and200 MW for LP expansion) and equal costfor the HP and LP expander engines, thetotal capital cost of the engines would be$109/kW. Some additional cost would benecessary to make the engines reversibleso that they could operate as reciprocatingcompressors. Thus, the cost of the expan-sion/compression equipment would be greaterthan $109/kW.

This estimated cost is much greaterthan that of the compound-engine or theall-turbine system. The conclusion isthat reversible-reciprocating engines forboth high- and low-pressure duty shouldnot be used.

COST COMPARISON BETWEEN A NEAR-TERMTURBINE SYSTEM AND COMPOUND ENGINE

A comparison of the capital and oper-ating costs of a CAES plant using a near-term turbine system and one using a com-pound engine is presented in Table 3. Inboth plant designs, compression is accom-plished using turbocompressors.

A water-compensated mined cavern wasselected as the air storage reservoir.The operating characteristics of the res-ervoir were 20-hr nominal storage (as de-fined in Ref. 5), 2190-hr generation time,and a storage pressure of 4.48 MPa (650psia). The storage-cavern and surface-reservoir costs were estimated using datafrom Ref. 5. The cost of the balance ofplant and the capital cost of the expan-sion/compression equipment were assumed tobe the same for both plant designs.

The use of the compound engine resultsin a cost reduction of about 25% for thestorage cavern and surface reservoir. Thisreduction is due to the lower air-flow re-quirement of the compound engine. The to-tal capital cost is reduced to about 11%.

The operating cost is equal to thesum of the capital charge, fuel cost, elec-tricity cost, and operating and maintenancecosts. By definition, the reduction in thecapital charge is equal to that of the to-tal capital cost. The fuel cost of thecompound engine is about 26% greater thanthat of the turbine system because of itshigher heat rate. Offsetting this increaseis a reduction of about 42% in the electri-city cost. The operating and maintenancecharge of 3.0 mills/kWh for the compoundengine represents a conservative estimatebased on Ref. 6. The total operating costof a CAES plant using a compound engine isabout 8% less than that of a plant usinga near-term turbine system.

OTHER APPLICATIONSOF RECIPROCATING ENGINES

The performance and cost estimatespresented in this paper are based on large-scale peak-power generation (e.g., greaterthan 200 MW) for electric utilities. How-ever, reciprocating engines are also adap-table to store energy in the form of com-pressed air using dispersed solar or windenergy conversion systems. These systemscould generate 1 kW - 10 MW of peak power.

CONCLUSIONS

The use of reciprocating engines inCAES plants has been evaluated in thispaper. Two plant configurations were con-sidered — one using reversible-expander/compressor engines and a plant using acompound engine concept composed of recip-rocating engines and turbomachinery. Theexpected advantages compared to turbine-based designs are reduced reservoir sizeand cost, reduced compression energy re-quirements, increased overall plant effi-ciency, and reduced capital and operatingcosts.

It is shown that using reversible-expander/compressor engines in place of anear-term turbine system reduces the re-quired air flow by about 53%, decreasescompression energy by about 58%, and in-creases overall plant efficiency by about5%. However, all of the cost advantagesachievable from this improved performanceare more than offset by the increase in thecost of the engines. It was estimated thatthe cost of the engines would be aboutthree times that of a turbine-based system.This result is due to the high cost ofusing reciprocating engines for low-pres-sure expansion. Thus reversible,

433

Table 3. Cost Comparison Between Near-TermTurbine System and Compound Engine

Water-compensated mined cavernStorage pressure = 4.48 MPa (650 psia)20-hr nominal storage; 2190-hr/yr generation time

Capital Cost, $/kWExpansion/Compression Equipment3

Storage CavernSurface ReservoirBalance of PlantIndirect Costs'3

TotalDifference

Operating Cost, mills/kWhC

Capital ChargeFuelElectricityOperating and Maintenance

TotalDifference

Near-TermTurbine System

4189880136

$354/kW

29.59.510.82.0

51.8 mills/kWh

CompoundEngine

4166680121

$314/kW-11.3%

26.212.06.3

• 3.0

47.5 mills/kWh-8.3%

Engines and/or turbines and compressors

15% contingency, 10% engineering and administration,30% escalation and interest during construction.

CCapital charge - 18%/yr; fuel - $2.50/106 Btu;electricity - 15 mills/kWh.

reciprocating, expander/compressor enginesshould not be used for both the high- andlow-pressure duties of a CAES plant.

The compound engine evaluated is com-posed of a reciprocating engine for high-pressure expansion, a gas turbine for low-pressure expansion, and turbocompressorsfor compression. Comparing a plant usinga compound engine concept to one using anear-turbine system, the air flow is re-duced about 35%, compression energy re-quirements are decreased by about 42%, andthe overall plant efficiency remains aboutthe same. However, the capital cost of theexpansion/compression equipment is aboutthe same (or possibly lower) than the tur-bomachinery of the near-term system. Itis estimated that the capital and operatingcosts are reduced about 11 and 8%, respec-tively.

The evaluation presented in this paperindicates that a considerable economic ad-vantage can be realized using a compoundengine concept in a CAES plant. Further

work is planned to evaluate the economicsof the substitution of the high-pressure,reciprocating engine for a gas turbine ina CAES plant. In addition, the use of re-ciprocating engines in dispersed solar orwind energy conversion systems will beevaluated.

ACKNOWLEDGMENTS

The research activities in compressedair energy storage, which formed the basisof this paper, were funded by the Divisionof Energy Storage Systems, U.S. Departmentof Energy.

REFERENCES

'Kartsounes, G.T., and J.G. Daley, The Useof Reciprocating Engines in Compressed AirEnergy Storage Power Plants, Proc. of the1978 Compressed Air Energy Storage Tech-nology Symposium (May 1978).

434

2Stys, Z.S., Air Storage System EnergyTransfer (ASSET) - Huntorf Experience,ERDA/EPRI CAES Workshop (Dec. 1975).

3KIm, C.S., and G.T. Kartsounes, A •Para-metric Analysis of Turbomaahinery Optionsfor Compressed Air Energy Storage Plants,Proc. of the 1978 Compressed Air EnergyStorage Technology Symposium (May 1978).

''Davison, W.R.., and R.D. Lessary, Studyof Selected Turbomaahinery Components forCompressed Air Energy Storage Systems,prepared by United Technologies ResearchCenter for Argonne National Laboratory,Report ANL/EES-TM-14 (Nov. 1977).

5Giramonti, A.J., Preliminary FeasibilityEvaluation of Compressed Air Energy Stor-age Power Plants, United TechnologiesResearch Center, R76-952161-5 (Dec. 1975).

6Raymond, R.J., et al., Cost and Applica-tion of Coal Burning Diesel Power Plants:A Preliminary Assessment, prepared byThermoelectron Corporation for NationalScience Foundation, NSF 75-SP-0917 (Aug.1975).

435

SESSION VIII: COMPRESSED.AIR ENERGY STORAGE

437

PROJECT SUMMARY

Project Title: Compressed Air Energy Storage—Advanced CAES System. Studies

Principal Investigators: W. V. Loscutoff, M. A. McKinnon

Organization: Pacific Northwest LaboratoryPO Box 999Rich!and, WA 99352Telephone: (5095 946-2768

Project Objective: Develop advanced CAES systems that require littleor no firing of the gas turbine with natural gas or oil.

Project Status: Alternate technologies and fuels are being investigatedto determine their potential for reducing the use of orreplacing premium fuels at a CAES plant. Among theAlternative fuel systems beinq investigated are coal firedmagnetohydrodynamics, fluidized bed combustion, nuclearwaste decay heat, and coal gasification. Alternatetechnologies investigated include the use of hybrid systemsusing varying degrees of thermal energy storage andadiabatic systems. The following concepts have been foundto hold the greatest promise: coal gasification, fluidizedbed combustion, and thermal energy storage. PNL isdirecting research at United Technologies Research Centerto study fluidized beds. PNL will concentrate itseffort on the study of utilization of thermal energystorage with CAES. This includes Dlanned activity with autility. The primary responsibility for integration ofcoal gasification with CAES lies with EPRI. PNL willmaintain information exchange with EPRI on this subject.

Contract Number: EY-76-C-06-1830

Contract Period: FY 1978, continuing

Funding Level: $150,000 60

Funding Source: Department of Energy, Division of Energy Storage Systems

439

ADVANCED CAES SYSTEM STUDIES

M. A. McKinnonPacific Northwest Laboratory

PO Box 999Richland, Washington 99352

ABSTRACT

The primary objective of this program during FY-1978 was to evaluate and screen anumber of advanced concepts available for integration with compressed air energy storage(CAES) systems. This paper summarizes the principal results and conclusions reached froma preliminary assessment of four concepts for heating air without using premium fuels.The concepts considered were Nuclear Waste Heat Augmented CAES, Magnetohydrodynamics (MHD)combined with CAES, Coal Gasification used with CAES, and Fluidized Bed Combustion (FBC)used with CAES. In addition to the above studies a comparative economic analysis of CAESin a hard rock cavern was made for three systems, a conventional fired system, a hybridsystem using a single stage of thermal energy storage (TES), and a no fuel double stageTES system. Conclusions reached from the studies are: 1) nuclear waste decay heatutilization in conjunction with a CAES system is technically feasible but the potentialfuel savings will not compensate for environmental concerns or justify the additionalsystem complexities, 2) .MHD has potential for a good match with CAES but should not bestudied further until the MHD problems have been solved, 3) Coal Gasification and FBCrequire more study, 4) Hybrid CAES systems and no fuel CAES systems using TES can becompetitive with conventional CAES if there is a large enough differential betweencompression power costs and fuel costs.

INTRODUCTION

Large scale energy storage systemsare useful in an electric utility grid asa means.for load leveling, i.e., storingenergy during periods of excess capacityfor use during periods of excess demand.There are several alternatives forstoring energy under investigation.Compressed Air Energy Storage (CAES) isone of these alternatives that appearsto be economically and technically viable.There is however a potential long termweakness of conventional CAES concepts.It is the reliance of conventional CAtSon clean petroleum fuels whose futureavailability for power generation isunder question due to increasing demand,decreasing domestic reserves and theuncertainty of potential import embargos.

The objective of the advanced CAESconcept studies at Pacific NorthwestLaboratories is to find and developCAES systems whose dependence on petroleumderived fuels is minimized or entirelyeliminated. During this past year fivesuch systems have been investigated.A comparative economic analysis was madeof CAES systems at a hard rock site with

and without augmentation by thermal energystorage (TES). In addition, PNL examinedhow nuclear waste heat, magnetohydrodynamics,fluidized bed combustion and coal gasifi-cation may be integrated with CAES. Thispaper reports the results of those studies.

INCREMENTAL COST ANALYSIS

The FY 1977 progress report for CAESadvanced concepts concluded that turbinefuel could be saved if Thermal EnergyStorage (TES) was added, but at thecost of increased system size andcomplexity, reduced overall energy utili-zation efficiency, and increased capitalcost. The preliminary economic analysisconcluded that there was currently noapparent economic incentive to include TESin a CAES system. However, because of thepreliminary nature of the analysis andbecause of the minor cost savings ofnon TES systems over that of systems usingTES, additional economic analysis wasrecommended. This section of the reportgives the results of that analysis.

CAPITAL COSTS

This analysis considered the three

440

systems shown schematically in Figures 1,2 and 3. The first figure is that for thereference or base CAES system. The secondand third are for the hybrid and the no-fuel systems, respectively. The operatingand performance variables used in theanalysis are shown in Table 1 and weretaken from reference 1. The no-fuelsystem was selected such that all of thecompression energy would be saved. We hadfelt that the no-fuel system would not lookeconomically competitive; however, as willbe seen from this study, chat is not neces-sarily the case. The no-fuel system canlook good under the proper set of condi-tions.

Fig. 1. Base System High Pressure CAESCycle

Costs generated for the reference CAESsystem were based on estimates developedin a recent Acres American, Inc. feasibilitystudy of CAES for peak shaving in Califor-nia.2 These estimates are also used as abasis for developing the costs of the TESsystems. Where equipment requirementsare similar to those of the base case butof a different size, costs were assumed tobe directly proportional to size. Wherenew equipment was used, other cost sourceswere sought. Becaus? this is an incrementalcost analysis, the primary focus of thecost comparison is on capital equipmentdifferences and on the differences inturbine fuel and compression energy use.More interest was shown in relative coststhan in absolute costs.

Fig. 2. High Pressure CAES Cycle WithRecuperation and Single StageTES/Regeneration

Table 1. Operating and performanceVariables*

System/State Parameter

OPERATING VARIABLES

LPC (Axia l Out letTemperature

HPC (Cent r i fuga l )Out let Temperature

HPT I n l e t Temperature

LPT I n l e t Temperature

HPT Expansion Ratio

PT Expansion Ratio

PERFORMANCE VARIABLES

Compressor Spec i f icMass Flow Rate

Turbine Spec i f ic MassFlow Rate

Coeff icient ofPerformance

Turbine Heat Rate

System Heat Rate

Units

°F

°F°F°F

lb/kwh

lb/kUh

(kWOout(klihiln

Btu/kHh

Btu/kHb

ReferenceCAES System

437

437

1,000

1,500

3.1

15.0

13.5

10.6

1.27

3.771

11.420

•F-Y-iy77 Progress Report Compressed A i r Enerqy St

Hybrid. System

700

4Sil1.000

1,500

5.8

8.0

12.2

10.6

1.15

3.279

11,810

oraqe Advanced

No FuelTES System

896

525502

855

6.1

7.6

11.0

16.0

0.69

14,100

Systems

A capital cost comparison of thethree systems is given in Table 2.These costs assume a 750 MW CAES facilityconstructed over 6.5 years with the plantgoing on line in 1984. All costs wereprice level adjusted to represent the totalinstalled cost in January, 1985 dollars/kW.

441

Table 2. Capital Cost Comparison ($106)

Fig. 3. No-Fuel TES High Pressure CAESSystem

Calculations for the base CAESsystem cost of electricity and theother incremental costs are based on thesame assumptions as were used in the AcresAmerican study. These assumptions arelisted in Table 3. Using these assump-tions, the levelized cost of electricityfor the three CAES systems were calculated.These costs are shown in Table 4.

The sensitivity of the levelized costof electricity to capacity factor, turbinefuel cost, and compression energy cost isshown in Figures 4, 5, and 6. Figures 4,5 and 6 can be confusing unless onerecognizes that the power cost from thebase CAES system (represented by a dashedline at zero in the figures) increases with

Reference Mo FwTCMS 5y»tt» Hybrid ! n l » TES S»it»

Land, Structures tnd Pl intEquipmentTurbo Machinery

Intercoolers

Aftereoolers

Compressor Recuperator

CompressorsCompressor Motors

Turbines

Turbine Exhaust Recuperator

Other Turbo Machinery

Storage Faci l i t ies

Air Storage s Associated Costs

Thenui Storage

Fuel Storage

Electrical and Miscellaneous

Engineering and ConstructionManigeMnt (151 of Direct Costs)

Contingency (15X of Above)

Escalation During Construction

Interest Durinq Construction

July 1984 $

January 1985 S

Plant Capacity

Cost/kN 1985 S

i S.5

1.7

0.8-

25.0

17.3

18.4

8.5

27.7

34.5

-

0.4

_ .1.0.6.

$150.4

$173.0

25.950.4

31.4

«280.7

287.8

750 MW

S384/W

$ 5.5

1.2.

22.0

25.0

19.218.4

8.5

27.7

34.5

2.00.4

. 1 0 . 6

$175.1

26.3

$201.430.2

58.7

36.6

$326.9

335.1

750 HIS447AH

$ 5.2

0.4

-

37.5

31.927.5

-

26.2

51.86.1-

$197.3

$226.934.0

66.1

41.2

$368.2

377.4

750 HI

$S03/kH

Table 3. Reference Case Assumptions forComparative Analysis

Plant Storage Capacity

Plant Capacity

Operation and Maintenance Costs

Fixed

Variable

Operation and Maintenance Cost Escalation Rate

Fuel Cost

Fuel Cost Escalation

Compression Energy Cost

Compression Energy Cost Escalation Rate

Cost of Money

Fixed Charge Rate

Life of Plant

Capital Cost Escalation Rate

Capacity Factor

la)19B5 Price Levels

10 hours

750 MU

$4.85/kW(a)

.32 m1]ls/kWh(a)

5.5I/yr

$4.00/10.6 Btu l a )

7.0%/yr

30mills/kHh la>

7.0J/yr

12*

15.5%

35yrs

51

251

Table 4. Levelized Cost of Electricityfor CAES Systems

CostComponent

Capital

0 S «

Turbine Fuel

Compression Energy

Total

Costs

ReferenceCAES System

27.2

4.5

31.5

19.3

11?.5

(mills/Ml)

HybridSystem

32.7

4.5

27.4

54.7

118.3

No FuelTES System

35.6

4.5

0

91.2

131.3

442

4Q

-10

• REFERENCE CASE

30MILL/KW-H

1657) (986)CAPACITY FACTOR, *

(1,314) 11,643) (1.971) (2,3001

(UTILIZATION RATE, 10° KW-HR/VEAR)

Fig. 4. Capacity Factor (Utilization Rate)Sensitivity

40

-10

• REFERENCE CASE30MILLKW-HM.O/UO6! BTU

NO-FUEL/IES

3.00 3.50 4.00 4.50 5.00

FUll COST, * HO4 81U

5.50

Fig. 5. Fuel Cost Sensitivity

fuel cost and compression energy cost anddecreases with increased capacity factor.These three figures only look at theincremental cost above that of the baseCAES system. As seen from these figures,the incremental cost for this particularhybrid system does not reach zero for anyvalue of the three parameters examined.However, the variance does not becomevery large over the range of values studied.The incremental cost for the no-fuel TESsystem exhibits a much higher degree ofsensitivity. For compression energy costsless than 15 mills/kW-hr in 1985, thisparticular no-fuel system could operateless expensively than the reference CAESsystem. In addition, if a 1985 compression

30

20

10

0

-10

• REFERENCE CASE30MILUKW-H$4.0410*) BTU

/

/ NO-FUEL/TES

HYBRID .

1 1 i

Fig. 6.

20 30 40 50 60

COMPRESSION ENERGY COST, MILLS /KW-HR

Compression Energy CostSensitivity

energy cost of 20 mills/kW-hr were assumed,instead of the 30 mills/kW-hr that wasused to obtain the results for Figures 4,5, and 6 the no-fuel system would be moreeconomical than the reference system forfuel costs greater than $4.75/106 Btu.

This study has shown that a particularno-fuel system can look attractive if thedifferential between compression energyand turbine fuel costs is large enough.However, that no-fuel system had assumed acompressor outlet temperature that isbeyond the limits of current technology.Even so, we are encouraged by the resultsand feel that additional analysis shouldbe conducted on no-fuel systems that useequipment that represents current techno-logy.

UTILIZATION OF NUCLEAR WASTE DECAY HEAT

A preliminary assessment has beenperformed of the utilization of decayheat from canisters containing solidifiedwaste fission products to augment acompressed air energy storage (CAES)system. The primary focus of the studywas to perform a technical assessment ofthe concept and then, if the resultswere sufficiently positive, to followthat study with a detailed economicassessment. The conclusion reachedis that even though a CAES system augmentedby nuclear waste decay heat may be techni-cally feasible, the potential fuel savingswill not be sufficient to compensate forenvironmental concerns or to justify theadditional system complexities.

443

The expansion side of the system usedfor the analysis is shown schematically inFigure 7. It uses exhaust heat recuperationand two stages of expansion. The compres-sion side of this system is the same asshown in Figure 1. The nuclear waste iscontained in a secondary loop and heat isrecovered by means of a heat exchanger.Calculations performed based on heat trans-fer considerations indicated that thenuclear waste would have to be containedin finned canisters in order to obtain therequired heat transfer and to maintain highair temperatures. Current canister techno-logy, even though ill defined, restrictscylinder designs to 12 to 24 inch dia-meter, 10 to 15 feet lengths, maximum skintemperatures between 500°F and 800°F, heatloadings of 3.5 kW/can to 13 kW/can, andmaximum allowable cyclic temperature fluc-tuation of 200°F.

Fig. 7. Expansion Side of a CAES SystemUtilizing Nuclear Waste Decay Heatand Turbine Exhaust Recuperation

The cycle calculations performedindicate that the use of waste decayheat can decrease fuel consumptionby up to 18%. This fuel savingsmust be balanced against the cost ofadding fins to the canisters andproviding waste storage caverns withremote handling equipment. Thissystem will also need air circulatingequipment, heat exchangers and airmonitoring equipment

Location is another factor thatworks against the concept. Ideally,a CAES plant should be situated near aload center, thereby minimizing trans-mission losses. However, locating anuclear waste storage facility near a

populous area would present a number ofdrawbacks. In addition to consideringsafety when selecting a site, the potentialimpact on the environment must be takeninto account. Problems with licensingwill also undoubtedly arise.

The obstacles and costs associatedwith the utilization of nuclear wastedecay heat with CAES appear to be muchgreater than an alternate system, CAESwith TES. The nuclear system will sufferpenalties and costs due to finned canisters,remote handling, siting and licensing.In addition, the nuclear system does notuse the compression heat.

The TES system can store the compressionheat that the nuclear waste system doesnot use. The temperature level of thecompression heat is comparable to theallowable skin temperature of a canisterin a nuclear system. A TES system can bevery simple, have a low capital cost,require minimal maintenance, and haveinsignificant impact on the environment.A nuclear system will not be able to com-pete with a simple TES system.

MAGNETOHYDRODYNAMICS

Magnetohydrodynamics (MHD) was foundto be a prime candidate for CAES applica-tions since approximately twenty percentof its output is used to compress air.Optimum use of MHD and CAES requires thatthey are located at the same facility.This would also allow the heat of compres-sion of the CAES system to be used to heatfeedwater for the MHD bottoming cycle.

CAES incorporated with MHD wouldallow a facility to have base and inter-mediate load capability. The facilitymay be able to supply peak loads also.The system would derive all of the airstorage compression power internally andwould not have to rely on power from theutility grid during off-peak periods. Theultimate goal of the continuously operatedCAES/MHD plant might be to supply theentire spectrum of the demand curve and todo so at efficiencies close to that of base-load power generation.

Presently the MHD system with bestcompatibility with CAES is an open-cycleMHD with a steam bottoming plant asshown schematically in Figure 8. Tocombine the system of Figure 8 with CAES,the compressor must be replaced by a CAESplant which includes an expander turbine

444

FUEL

SEED

_LELECTRICALPOWER-OUT

COMBUSTIONCHAMBER

OWERSTACK

GAS

MHDGENERATOR

AIRPREHEATER"

ELECTRICALPOWER-OUT

THREE-PHASEGENERATOR

STEAMGENERATORS

COMPRESSOI STEAMTURBINE

CONDENSER

A RINLET •=

• • WATER

0FEED PUMP

Fig. 8. Open Cycle MHD with a SteamBottoming Plant

whose outlet pressure is ten atmospheres.During retrieval of air from storage, theten atmosphere expander turbine exhaust'will become the air supply to the combus-tion chamber of the MHD system. Such aCAES/MHD combination would allow an MHDplant with a base output of 970 MW tobecome a plant that can deliver a range ofpower from 628 MW to 1346 MW. Preliminarycalculations indicate the combined CAES/MHDplant can supply mid range power at effi-ciencies close to base load efficiencies.

Before MHD/CAES can be demonstrated,MHD will have to be successfully demon-strated on a commerical scale. Many MHDproblems remain to be solved prior tothat time. The most obvious and hardestMHD problems are associated with hightemperature materials for electrodes.Several other problems which are currentlypreventing MHD from being a near termoption for power generation are: endcurrent losses, electrode arcing, largescale DC-AC power conversion, avoidance ofshocks, and seed separation.

Our preliminary evaluation indicatesthat MHD would be a good companion forCAES, however, we recommend that addi-tional development of MHO/CAES systemsbe delayed until most of the MHD problemsare solved.

FLUIDIZED BED COMBUSTION

During the previous fiscal year apreliminary study of Fluidized Bed Combus-tion (FBC) integrated with CAES wasinitiated. The material reviewed indicatedthat FBCs could provide a clean efficientmeans of burning coal without the need forstack gas scrubbing. FBC was also reportedto have low NO emissions and reduced carryover of volatiTes. The bed's temperaturealso eliminates slagging problems andproduces a soft unsintered ash.

These positive findings promptedadditional study of FBCs. A request forproposals was advertised and an evaluationof the responses has been completed. Acontract with United Technology ResearchCenter to further that study is beingfinalized. The overall contracted effortconsists of four tasks which are:

. Review and assess the state-of-the-art of pressurized and atmosphericFBC as they relate to CAES applications.

. Develop schematic diagrams of possibleFBC/CAES configurations. Perform apreliminary screening analysis toidentify the system or systems whichhave the greatest potential forsuccessful development as near termCAES peaking plants.

. Develop a pr.conceptual design of thebest system as determined from thescreening analysis. Determine avail-ability and estimate cost of systemcomponents.

. Project the potential market availableto a CAES/FBC system.

Meanwhile, additional studies of FBCshave been conducted at PNL to identifypotential problems or limitations imposedby combining FBC with CAES. The majorproblems are associated with either thedesign and cost of a compatible heatexchanger for an atmospheric FBC (AFBC) orthe particulate clean-up problems neededto insure the survivability of the turbinesin a pressurized FBC (PFB) system. Thetechnology to clean the gas at high temper-atures does not exist at the present time.

Most of the available literature onAFBCs addresses their use as boilers in asteam system. As boilers the systems canbe small due to optimum use of the enhancedheat transfer coefficient in the fluidizedbed. The tube wall temperatures are kept

445

low by the cooling effect of the boilingwater on the inside of the boiler's tubes.The superheat region of the boiler can belocated out of the bed.

When an AFBC is used as an air heaterthe tube wall temperatures pose more severedesign problems. The cooling effect ofthe air on the inside of the air heatertubes does not provide as great a coolingeffect as boiling water; therefore, thetube wall temperatures and tube surfacearea can be expected to be higher. Thehigher wall temperature will demand moreexotic materials to withstand the pressuresand corrosion potential. It appears thatIncoloy 800 will be required for the heatexchanger so that wall temperatures of1200°F to 1300°F.can be attained. Theincrease in tube surface area requiresmore tubes, but the density of tubes islimited by its impact on the bed fluidizingand mixing processes. Consequently, toget more tubes in the bed requires a biggerbed.

Material problems and bed size will beincluded in next year's effort to assess theviability of a FBC/CAES.

COAL GASIFICATION

A preliminary study was made to assessthe possibilities of augmenting CAES withcoal gasification. The study identifiedone gasification system that is recommendedfor further study.

Gasification is the process of con-verting solid coal to a clean gaseous fuel.Most processes used air or oyxgen with steamto react with the coal. The product xuelgas may have a heating value range of 100to 1000 Btu/scf depending on the process.The heating value of the product gas hasbeen divided into three major categories.They are: low-Btu gas (LBG) with heatingvalues from 75 to 175 Btu/scf. Inter-mediate-Btu gas (IBG) covers the 250 to400 Btu/scf range, and high-Btu or pipe-line quality gas (HBG) with a 925 to1000 btu/scf range. Air blown reactorsproduce LBG whereas MBG is produced byoxygen blown reactors. HBG requires expen-sive shift reaction and methanation process-ing of MBG. In general, the thermal effi-ciencies of the conversion process decreasewith increasing heating values while theease of storing and transporting the gasincreases.

Gasifier development has produced a

multitude of different processes. Gasifiersare generally classified according totheir coal flow relative to the gas flow.Under this system, three types can beidentified: fixed bed, fluidized bed, andentrained flow gasifiers. Other classifi-cation systems are also used, such aspressure level, source of oxygen (air oroxygen blown), number of stages, or ashremoval method.

It is not the intent of this part ofthe paper to provide a complete descriptionof gasifier processes but only to indicatethat there are many types to choose from.For more detailed information one isreferred.to the references on gasifica-tion J 3" 6)

Power generation systems that usecoal as their primary fuel are high incapital cost with respect to oil andnatural gas systems. Coal gas firedsystems would be even higher in capitalcost. High capital costs and the inabilityfor daily turn off of gasifiers dictatesthat the gasifier facility be operatedcontinuously and preferably at fullcapacity. The continuous operation wouldalso help to defray their high capitalcost.

The application of coal derived fuelgas to CAES has now been narrowed to threegeneral alternatives. The options are asfollows:

. The gaseous fuel produced duringcompression periods would be soldand distributed to industrial usersthat are located nearby.

. Excess fuel produced during compres-sion periods would be stored to beused during generation periods.

. The turbines of the CAES operatecontinuously using all the fuel gasas it is being produced.

The first option has control difficul-ties and is limited to sites near largeindustrial users. The user must be will-ing to use off-peak power. If he is willingto use off-peak power then he could be usedto load level directly without the expenseof a coal gasification/CAES plant.

The second option requires storageof gaseous fuel. The technical prob-lems associated with isolating gas andair storage could make site selection

446

very difficult. Current gasifiers areoperated below 200 psi which would requirelarge storage volumes for the gas or ameans of compressing the gas into smallervolumes. The additional gas storage costadded to the already high cost of thegasifier makes the option less attractivethan the other options.

The last alternate allows continuousoperation of the gasifier without the needto sell or store the product gas. The mainoperational characteristic is continuouscombustion of the product gas as it is beingproduced. The acronym CGC/CAES (ContinuousGasification and Consumption CAES) will beused to denote this alternate. This optionis conceptually different in that the plant'soutput would change by switching compressortrains on or off. The compressors would beon during storage and off during operation.The turbines would be operated continuously.Conventional CAES switches on and off boththe compressor and turbine trains.

The CGC/CAES bypasses the major disad-vantages of other gasification optionsnamely storing or selling fuel gas. It isthis advantage along with the abilityto accommodate a coal gasifier thatoperates continuously and at near fullcapacity that makes CGC/CAES the mostattractive Coal Gas/CAES. The CGC/CAESconcept will be studied in greater de-tail next year. The extent of the studywill depend on the results and directionof the coal gas/CAES work being doneby EPRI.

REFERENCES

1. Kreid, D.K. and M. A. McKinnon, "FY-1977Progress Report Compressed Air EnergyStorage Advanced Concepts," Report No.PNL-2464/UC-946, March 1978.

2. M.J. Hobson, et al., "Feasibility ofCompressed Air Energy Storage, as a PeakShaving Technique in California," AcresAmerican, Inc. for the California EnergyCommission. May 1978.

3. Donald L. Katz, Dales E. Briggs, EdwardR. Lady, John E. Powers, M. Rasin Tek,Brymer Williams, and Walter E. Lobo,Evaluation of Coal Conversion Processesto Provide Clean FueTsT EPRI 206-0-0Report to Electric Power ResearchInstitute by University of MichiganCollege of Engineering, 1974.

Assessment of Low- and Intermediate-BtuGasification of Coal, FE/1216-4, NationalAcademy of Sciences, Washington, D.C.,1977.

Handbook of Gasifiers and Gas TreatmentSystems, Dravo Corp, Pittsburgh, PA,1976.

447

PROJECT SUMMARY

Project Title: Application and Design Studies of Compressed AfrEnergy Storage for Solar Energy Applications

Principal Investigator: Gerard T. Flynn

Organization: M.I.T. Lincoln LaboratoryP- 0. Box 73Lexington, MA 02173617/862-5500 Ext. 7456

Project Goals: The objective is to investigate the applicability ofcompressed air energy storage in combination with thermalenergy storage for use in electric utility power generation.The combination of solar thermal and off-peak energy willbe used in determining cost and performance models. Par-ticular attention will be focused on the cost and performanceof the packed beds for thermal eneroy storage.

Project Status: Work nearly completed on the initial 15 month program.Final report is in preparation.

Contract Number: EX-76-A-01-2295

Contract Period: Oct. 1978 - Jan. 1979

Funding Level: $150,000

Funding Source: U. S. Department of Energy

449

SOLAR THERMAL AUGMENTATION OF CAES*

G. T. FlynnMIT/Lincoln Laboratory

P.O. Box 73, Lexington, Massachusetts 02173

ABSTRACT

This paper describes a storage system which combines an adiabatic compressed airstorage plant with a solar thermal central receiver. A packed pebble bed is used tostore the heat of compression generated during the charging of the compressed airreservoir. This heat is then returned to the air on discharge of the reservoir. Thisstored thermal energy minimizes the solar thermal energy required to raise the airtemperature to 1500°F at the turbine inlet The system described is based on a centralreceiver design proposed by Boeing. The turbomachinery is within the state-of-the-art,but would be built specifically for this application.

Economic screening curves are given which compare conventional and adiabatic CAESfor both coal fired and nuclear base load pumping. This hybrid system is marginallycompetitive with conventional CAES on the basis of levelized fuel costs of $3.55 perMBtu and nuclear off-peak pumping costs of 7.0 mills per kWh.

1.0 INTRODUCTION

The storage of off-peak energy in theform of compressed air could provide aconsiderable saving in the fuel used forpeak power generation. An open-cycle gasturbine peaking plant normally wouldprovide its own compressed air in a Braytoncycle. Approximately two-thirds of thepower from the turbine is used to run thecompressor, and the remainder is availablefor electric power generation. Therefore,the efficiency is typically 30% or lessand the corresponding turbine heat rate is12,500 Btu/kWh.

A CAES (Compressed Air Energy Storage)system which provides compressed air froman air storage reservoir can generatepower with a heat rate as low as 4000Btu/kWh (1). If the compressed air reser-voir is pumped (charged) by nuclear orcoal base load capacity during off-peakhours, CAES could reduce high distillatefuel consumption by a factor of three.

If the air were compressed isentro-pically, i.e., without intercooling andthe heat of compression were stored in aTES (Thermal. Energy Store) , the systemcould be operated adiabatically (2,3,4)and no additional heat would be required

*This work was sponsored by the Divisionof Energy Storage Systems of the U.S.Department of Energy.

before expansion in the turbine. However,an adiabatic system will always have anenergy return ratio (efficiency in thiscase) of less than unity. The energyreturn ratio is defined as the energydelivered to load from storage divided bythe energy used for charging the storage.In the case of a fired CAES, the energyreturn ratio could be as high as 1.4.

If a hybrid scheme were used wherethermal energy was added to an adiabaticsystem before turbine expansion, consider-ably less fuel-derived thermal energywould be used to achieve the same turbineinlet temperature utilized in fired CAESdesign. If the thermal energy is to beprovided by a solar central receiver, theimportance of minimizing the heat rate iseven greater than for the fuel fired casein order to minimize the capital cost ofthe solar central receiver and heliostatfield.

2.0 SYSTEM DESCRIPTION

Figure 1 shows the hybrid solar/CAESsystem. Off-peak energy is stored ascompressed air in an underground cavernat a pressure of 35 BAR (BAR=14.5 p.s.i.a.)In order to maintain a constant pressureduring charging and discharging, thecavern is hydraulically compensated witha water shaft to a surface sited pond.The depth of the cavern from the surface

450 - If $

is determined by the system pressure re-quirement. The depth would be 33.5 feetx storage pressure in BARs. This calcu-lation limits the system pressure to benot greater than the local pore waterpressure, which eliminates any possibilityof air leakage through saturated over-burden. At a pressure of 35 BARs, thedepth would be 1200 feet.

Figure 1 Hybrid Solar/CAES System

A packed pebble bed T£S is used tostore the heat of compression £iom thefirst compressor which provides a com-pression of 16:1. The second boostercompressor has a compression ratio of2.2:1 to provide the cavern storage pres-sure of 35 BAR. On discharge, air fromthe cavern passes through the TES to thecentral receiver. The central receiveris a power tower design in which 35 acresof reflecting mirrors concentrate sunlighton a tower mounted heat exchanger. Thiscentral receiver heat exchanger raisesthe temperature of the air from 820°F(1280°R) to 1500°F (1960°R) beforeexpansion in the turbine. There is nostorage for the solar thermal system anda backup fuel fired combustion chamber ispostulated in the economic analysis toallow for solar unavailability.

The compressor and turbine chains areconnected through clutches to the motorgenerator such that either chain may beappropriately connected on charge ordischarge.

2.1 CAES CHARGING CONSIDERATIONS

The first compressor increases thepressure from atmospheric to 16 BAR,which is a reasonable value for availablecompressors operating without intercooling.With an inlet temperature of 70°? (530°R),the outlet temperature would be 780°F(1240°R). This air must be cooled beforeentering the booster compressor. Ratherthan discarding the thermal energy, it isstored in a TES (Thermal Energy Store).

The TES (5) consists of a largepacked bed of pebbles. The air passingthrough transfers heat to the pebbles.The air velocity is quite slow and at anypoint in the bed, there is only a slightdifference in temperature between the airand the pebbles. Therefore, as the airpasses through the bed and heat is trans-ferred from air to rock, a steep thermo-cline (temperature gradient) is developed.This thermocline moves gradually to thebottom of the bed on charging. On dis-chaiging, air flows in the reversedirection through the bed and the thermo-cline would move back to the top. Becauseof irreversibilities in the hfcat transfer,the thermocline would gradually disperseon successive charge/discharge cycles. Asthis dispersion increases, the outlettemperature of the bed would rise towardsthe end of the charge cycle. To preventthe inlet air temperature from exceedingLhe maximum allowable inlet temperatureto the booster compressor, the air iscooled by passing it through a pipe whichruns through the watershaft to the surface.The effect of this natural heat exchangerhas been calculated assuming a water tem-perature of 68°F (528°R). With a maximumcharging TES outlet temperature of 820°F(1280°R), the maximum booster compressorinlet temperature would be 240°F (700°R).This is a sufficiently low inlet tempera-ture to prevent damage to the boostercompressor. The amount of heat lost tothe water on each cycle is small and isaccounted for in the TES efficiencyrating of 90%.

The way the system has been config-ured, it is not possible to utilize theheat of compression from the boostercompressor. With an adiabatic efficiencyof 75%, the thermal energy is 75.46 Btu/lb or 13% of the turbine heat rate ondischarge. The air in the cavern goesdirectly to the TES on discharge. If itwere not cooled before storage, the lower

451

end of the bed would equilibrate at toohigh a temperature to act as an effectiveintercooler on charging. Storing the airat L-gh temperatues also increases thevolume required and thus would increasethe capital cost. If the pipe from thebooster compressor to the cavern is runthrough the watershaft to form a counter-flow heat exchanger, the air in the cavernwill be at an average temperature of 80°F(54O°R). At this pressure and temperature,the cavern volume required is 3.0 ft3/kWh(discharge).

2.2 CAES DISCHARGE CONSIDERATIONS

On discharge, air leaving the cavernis piped directly to the lower end of theTES. The air temperature is raised from80°F (540°R) to 820°F (1280cR) in passingthrough the packed bed. The dischargetemperature of the packed bed over theentire discharge interval is shown inFigure 2, The average TES dischargetemperature is 78O°F (1240°R). This airis then piped to the solar centralreceiver (or a combustion chamber) wherethe temperature is raised to 1500°F(1960°R) before entering the expansionturbine. The solar thermal heat rate is288 Btu/lb of air. The central receiverchosen as a model for performance and costwas a Boeing design (6,7). The Boeingsystem is designed to operate at 34 BAR ina closed Brayton cycle. The basic designof the system with some scaling as de-scribed below has been taken intact withthe closed cycle turbomachinery replacedby the 32 BAR expansion turbine describedbelow.

TES DISCHARGE TEMPERATURE (*R) vsTIME FOR SUCCESSIVEDISCHARGE PERIODS ON A DAILY CYCLE BASIS

5 HOI

© DISCHARGE PERIOD No 1

© DISCHARGE PERIOD No.5STEADY STATE CONDITION

'or a turbine expansion ratio of 32:1,an isentropic efficiency of 90%, and agenerator efficiency of 98%, the mass flowthrough the turbine is 7.85 lbs/kWh. Thesolar thermal heat rate is 2260 Btu/kWh,which is equivalent to a "thermal effi-ciency" of 151%. This "thermal efficiency"is useful only in scaling the solar thermalportion of the system; that is, the turbinegenerator efficiency in the Boeing designwas 42%. With this hybrid, 3.6 times asmuch energy is produced with the samescale central receiver and heliostac field.Alternately, the cost of the components inthe solar thermal system expressed indollars per kWh are reduced by a factorof 3.6.

3.0 ECONOMICS

3.1 CAPITAL COSTS

The capital costs are based on a 100MWe, 12 hr storage system. The costs forthe solar thermal system have been takendirectly from the Eoeing cost estimatewith appropriate scaling as describedabove. A cost escalation of 8.5% per yearwas used, and all estimates are expressedin 1980 dollars.

The capital costs of the system fallinto three general categories: miningand construction; solar central receiverand heliostats; and rotating machineryand balance of plant.

3.1.1 COST SUMMARY MINING AND CONSTRUCTION

Geology Land Site ImprovementsCavern Excavation @ $1.0/ft

(12 hr)Water ShaftAir ShaftsValves & Misc. FittingsThermal Energy StoragePond (land & sealing)

Mining & Construction Subtotal =

3.1.2 COST SUMMARY SOLAR THERMAL

$/kW

7.00

36.001.3.5.

.80

.40

.0035.002.80

$91.00

DISCHARGE TIME I HOURS)

Figure 2

Land & Site ImprovementsStructures & FacilitiesHeliostatsCentral ReceiverFuel Storage

Solar Thermal Subtotal =

$/kW

0.4120.73

121.9242.6727.63

S213.36

452

3.1.3 COST SUMMARY-ROTATING MACHINERY &BALANCE OF PLANT

$/kW

Compressors, Turbines, Clutches,etc. 160.00

Generators, Transformers,Switchyard 16.14

Balance of Plant 13.25

Rotating Machinery Subtotal

3.1.4 TOTAL SYSTEM COST

$189.39

MINING & CONSTRUCTIONSOLAR THERMALROTATING MACHINERY & BALANCEOF PLANT

CONTINGENCY @ 15%INTEREST DURING CONSTRUCTION

TOTAL COST - 1980

3.2 AVAILABILITY FACTOR

9/K.W

91.00213.36

189.3977.1488.71

$659.60

Due to both planned and forcedoutages, a power generation facility isnot available 100% of the time. However,gas turbine plants have very short outageperiods, i.e., complete overhaul is quickcompared to a nuclear or coal plant.Based on the gas turbine experience, theavailability factor has been set at 85%.

3.3 FIXED ANNUAL CHARGE RATE

The fixed annual charge rate is basedon investor return on capital, life ofthe plant and annual tax rate. For thisanalysis the fixed annual charge rate is18%. The annual capital cost is:

Annual Capital Cost-Total Capital CostAvailability Factor

x Fixed Annual Charge Rate.

3.4 CHARGING ENERGY COST

In order to charge the storage, off-peak energy from a baseload facility isused. Generally, the capital cost of abaseload plant is charged against energydelivered to load and storage is chargedonly the incremental cos?- of fuel, opera-tion and maintenance. For a nuclearplant (8), the incremental cost is 7mills/kWh charging energy and for a coalplant, 15 mills/kWh. Each kWh of dis-charge requires 0.9 kWh of charge. There-fore, the incremental pumping cost fornuclear is 6.3 mills/kWh and 13.5 mills/kWhfor coal baseload.

3.5 FUEL COST

For purposes of estimating deliveredenergy cost, a solar availability factorof 62.5% is used. The remaining 37.5% ofthermal energy would be supplied by oil ornatural gas. On an annual basis, theaverage fuel heat rate would be 0.375 x2290 Btu/kWh or 860 Btu/kWh. With anestimated 1980 fuel cost of $3.55/MBtu, theincremental fuel cost is 3.0 mills/kWh.

3.6 OPERATING AND MAINTENANCE COST

Operating and maintenance cost hasbeen set at 3 mills/kWh which is basedon operating experience with conventionalgas turbines.

3.7 ECONOMIC SCREENING CURVES

In order to decide the relative valueto a utility of any amount of plant ex-pansion, one of the most useful tools forfirst cut evaluation is the economicscreening curve. Alternative generatingoptions may be plotted in terms of costper year per unit capacity versus genera-ting hours per year. If a plant were tobe used for reserve capacity and operatedonly several hundred hours a year, thecapital costs are more significant thanthe fuel costs and the choice is clearlythe simple open cycle gas turbine peakingplant. On the other hand, if the baseloadis expanded, i.e., generation time is sixto eight thousand hours per year, the fueland operating costs would dominate and thechoice would be a nuclear or coal plant.

Passive storage systems do not havefuel costs, but do have incremental opera-ting and maintenance costs as well ascharging energy costs. Conventional fuelfired CAES has both an incrementalcharging cost and a fuel cost and this isindicated on the screening curves by thehigher slope these systems have over thepassive or adiabatic systems.

4.0 SUMMARY AND CONCLUSION

Because of the high capital costs ofthe solar/CAES option, it does not seemeconomically competitive with any of theUPHS (Underground Pumped Hydro Storage)or other CAES systems.

453

The final busbar power cost per yearversus generation hours per year is shownin Figures 3 & 4 for nuclear and coalpumping. Also shown are several otherstorage systems for economic comparison.The busbar energy cost may be derived bydividing the ordinate by the abscissa.With nuclear pumping the hybrid solar/CAESoption is marginally competitive with theopen cycle gas turbine with a generationtime of 3000 hours per year.

POWER GENERATION COST COMPARISON12-32 BAR CAES SINGLE OR NO TESNUCLEAR PUMPING O 7MILLS/kWh

^ 200

i

ft >»ou

iO 100

I3 , .

• u a l A . _ _ - — • / •

C&~"'K

/ ^ ~ ~ —t-——"

IAR SOLA! HVtRID SINGLE TES

R ADrAiATK CA£S SINGLE TCS

•AR AQUIFER NO TCS

4400 ft 2 O I O P T 0 J 9

•AR SOLUTION MINED NO TES

R AD1ABATIC SINGLE TE5 |AIAN0ONCDMINE)

D tO hr WHKNrGMT PUMPING

O 6 hr WEENMGHT PUMPING

2000 4000 4000

GENEIATION TIME HI/YEAH

Figure 3

POWER GENERATION COST COMPARISON12-32 BAR CAES SINGLE OR NO TES

COAL PUMPING 9 IS MILLS /kWhr

32 M l SOLAt H<l

15 t U ADIAIATtC UNCLE f«S

UPHS 4400(1. 2 not if •0.93

2-20 • « • AOUIfCII NO TES

12-20 I A I UXUIION MINED NO ISS

NOONEOMmE)

• 10 kr WEIKNKSHI PUKWNG

O t tr WEEINIOHT PUMPMC

2000 400D «ooo

GENEIATFON TIME Hlt / tEA*

REFERENCES

1. Mattick, W., et. al., "Huntorf - TheWorld's First 290 mW Gas Turbine AirStorage Peaking Plant."

2. Koutz, S. L., "Energy Storage Systemand Method," U.S. Patent No. 3,677,008,July 1972.

3. Stephens, T., "Adiabatic Compressed AirEnergy Storage Systems," Proceedingsof the Workshop on Compressed AirEnergy Storage Systems. ERDA-76-124.

4. Glendenning, I., "Advanced CompressedAir Storage - An Appraisal," Com-pressed Air Energy Storage TechnologySymposium. Asilomar ConferenceGrounds, Pacific Grove, California,15-17 May 1978.

5. Hamilton, N. I., "Packed Beds forThermal Energy Storage in an Under-ground Compressed Air Energy StorageSystem," AS/ISES, Denver, Colorado,August 1978.

6. Gintz, J. R., "Closed Cycle, HighTemperature Central Receiver Conceptfor Solar Electric Power," EPRI ER-183,Project 377-1, February 1976.

7. Gintz, J. R., "Advanced Thermal EnergyStorage Concept Definition Study forSolar Brayton Power Plants," FinalTechnical Report, Volume I, ERDAContract EY-76-C-03-1300, December 1976.

8. Giramonti, A.J., "Preliminary Feasi-bility Evaluation of Compressed AirStorage Power Systems," N.S.F. GrantNo. AER 7400242.

Figure 4

454

APPENDICES

455

THE APPLICATION OF FLYWHEEL ENERGY STORAGE TECHNOLOGY TO SOLAR PHOTOVOLTAICPOWER SYSTEMS

Alan MillnerMassachusetts Institute of Technology

Lincoln LaboratoryLexington, MA 02173

ABSTRACT

INTRODUCTION

Solar photovoltaic (PV) electricpower systems presently being developedinvariably use electric storage batterieswhen on-site energy storage is required.Moreover, studies of future PV powersystems assume continued use of batteriesfor on-site storage, albeit with moreadvanced, efficient and less expensivebattery designs. This preeminence is dueat least in part to the generally heldconviction that no other on-site storagesystem can compete economically withbatteries for PV usage. However,studiesperformed during the past year at MIT/Lincoln Laboratory show that flywheelenergy storage will be technically andeconomically competitive with eitherpresent-day or advanced storage batteriesif the flywheel storage system is pro-perly configured. This conclusion wasreached after comparing battery andflywheel storage in a systems context,whereby their influence on other sub-systems (such as inverters) was deter-mined.

The essence of the proposed approachis the utilization of the flywheel sub-system for more than the energy storagefunction. A PV power system usuallyrequires an inverter to convert thelow-voltage DC output from the solararrays to a (usually) higher voltageAC waveform, and this operation canbe performed by the flywheel unit byuse of a DC drive motor and a permanentmagnet alternator. Also, it is usuallynecessary to provide a good impedancematch between the PV array and the

load in order to maximize the electricpower extracted from the array, andthis function (commonly referred to asmaximum power tracking) can also beprovided by the flywheel. Figure 1shows the system block diagram compar-ison of a battery system and inverter,a conventional flywheel with DC inputand output followed by a DC-to-ACinverter and a combined flywheelstorage and power conditioning system.The simplicity of the last blockdiagram reflects the real cost savingspossible with this implementation.These simplifications hinge on the useof an efficient, low-drag motor-generator, such as one recentlydesigned by MIT/LL for a spacecraftapplication. The system also utilizeslow-drag high-speed bearings, such asa magnetic bearing recently designed,built and tested for the same program.

The costs for a flywheel systemwith the capabilities enumerated abovewere estimated for a PV residential ap-plication and were compared with costsfor a more conventional system contain-ing batteries, inverter and maximumpower tracker. Two different scenarioswere considered: 1) present-day bat-teries and costs and present-day tech-nologies and costs for flywheels withproduction quantities assumed in bothcases, and 2) technologies and costsextrapolated to the 1986 time framefor both the flywheel and battery sys-tems. In both scenarios it was foundthat because of savings resulting

457 —

I BATTERY SYSTEM

2. FLYWHEEL STORAGE PLUS INVERTER

I FLYWH=SL I

iRRAv] » DC M'G ElECTRONICS »-|lNVEBTER | — » • ! 10AD 1

3. FLYWHEEL STORAGE AND CONDITIONING

HYWHEEL

ARRAY M MOTOK-GENERATO8 ElECIRONICS 1 » 1OAO

Figure 1. Solar PV System Comparison

from multiple usage of the flywheelcomponents, such a system would beeconomically competitive with the moreconventional approach utilizing storagebatteries, a separate inverter and amaximum power tracker. The cost com-parisons are given in Fig. 5.

By combining these elements witha power conditioning design developedat MIT/LL and an improved rotor procuredfrom one of the contractors in DOE'sstorage program, a subscale (approxi-mately 1:10) model of a residentialsolar photovoltaic flywheel energystorage unit will be assembled. Inaddition, scaling laws will be derivedfor storage coupled to solar PVsystems in the power range 10 to 100kW peak, and in the storage range 25kWh to 5 Mwh. A detailed design ofa full-sized 10 kW/25 kWh single-residence storage unit will be madeand analyzed for cost and worth tothe user. Similar but less detailedanalyses will be performed for a100 kW/5M¥h load center storage unit.

SYSTEM DESCRIPTION

A solar photovoltaic installation

usually requires a solar cell array, anenergy storage device, a maximum powertracker, and a DC-AC conversion device.This proposal is concerned with thedevelopment of a flywheel system whichwould perform all of these functions.

Such a device would have a DCmotor, an energy storage flywheel anda 60-Hz AC generator. It would besupported on magnetic bearings in avacuum housing which would do doubleduty as safety containment for theflywheel rotor. The motor-generatorwould in fact be the same device withseparate input and output electronics.The electronics are shown schematicallyin Figures 2 and 3 and are very simpleand inexpensive due to the use of thepermanent magnet motor-generatorconcept.

All solar/electric power would gothrough the DC motor to spin up the ro-tor. The motor would be a DC brushless,ironless armature type design whichwould be controlled as a maximum powertracker for the solar array. This isimportant because the varying electri-cal output of a solar array is gener-ally mismatched to the characteristics

458

A- B-

C+

C0NVE6IEB

STABTINGfc COMMUTATOI

Pig. 2: Flywheel Power Conditioning Input Schematic

Fig. 3: Flywheel Power Conditioning Output Schematic

of the storage system and load,causing inefficient operation. Theflywheel rotor could be an advanceddesign of one (or more) of the typespresently being tested for DOE by anumber of organizations. 'The elec-trical output would be accomplishedwith a permanent-magnet brushlessgenerator and a silicon controlledrectifier (SCR) cycloconverter.The generator would be sized toaccept the large surge demands ofmany candidate loads.

The entire rotating unit wouldbe supported on a DC magnetic bearingor hybrid magnetic/ball bearing.This can be powered from windingson the motor, allowing fail-safe,spin-down operation. The resultingrotating unit would have no brushesor physical contact with the rotor,allowing very long life and high

reliability. Also, the higher speedspossible with such an assembly(perhaps 20 thousand rpm) will allowsmaller rotor size and enhance thequality of available AC power. Mech-anical touchdown bearings would befor cold start/stop conditions. Norotating seals would be required.

The vacuum chamber ensures a longenergy storage time (days or weeks) be-fore aerodynamic losses become a prob-lem. With bare-filament rotors, thevacuum system will provide sufficientsafety confinement. If other designsare enployed, the unit can be placedunderground to provide safety confine-ment. The system would be run over a2:1 speed range with the output held at60-Hz, generated independently or syn-chronized to an external line. This co-responds to a 75 percent depth of dis-charge for the energy storage function.

459

An attractive practical appli-cation for this device is to utilizeit with a PV-powered single-familyresidence operating in either a stand-alone mode or coupled to a utility,

with utility power drawn only when theenergy reverse in the flywheel is low.It can be scaled up to larger sizes as-sociated with multiple dwelling unitsor commercial PV applications.

TASX 1 Build one 1/10 scale model lyitetn with a 1 kwh

purchased rotor optimized for PV storage. Refine

scaling laws.

TASK 2 Design a residence-sized system and a 100-kH

load center system and analyze for cost and

worth to the user.

TASK 3 Build one full-scale, residence-sized system and

interface it to a solar array and residential

load.

Fig. 4: PV Flywheel Storage Component Project

Battery System

law High

Storage $/kWh 90 ISO

DC input SAW 50 150

AC output SAW 50 300

Enclosure $/kWh 32 50

Residence Total $3,852 $8,400

(Based on 25 kWh +

6 kHDC + 10 kWAC)

Flywhe

Low

75

85

70

24

$3,705

el System

High

145

125

150

32

$6,875

Fig. 5: Comparison of Prices

FLYWHEEL BASED SYSTEM

DC Motor Electronics 95*

DC Motor 96%

AC Generator 95%

Drag Losses 95%

Gen. Electronics Losses 92%

Max Power

Batteries

Inverters

BATTERY

Tracker

BASED SYSTEM

96%

80%

85%

73.3% TOTAL 65.3%

Fig. 6: Comparison of Efficiencies of Flywheel Energy Storage and ConversionSystem versus That of a Battery Inverter and Max Power Tracker System

460

Motor/generator

Rotor

Bonom bearing

MECHANICAL AND MAGNETIC ENERGY STORAGETECHNOLOGY MEETING

The Mimslyn Motor InnLuray, V i rg in i a

October 24-26, 1978

Tuesday, October 24

8:00-9:00 a.m.9:00-9:30 a.m.

RegistrationIntroduction

9:30 a.m.

9:40 a.m.

9:50 a.m.

10:20 a.m.

10:40 a.m.

11:00 a.m.

11:20 a.m.

Session I: FLYWHEELS9:00 a.m.-12:20 p.m.

Chairman - Thomas M. Barlow, Lawrence LivermoreLaboratory

Electric and Hybrid vehicle Applications (EHV-MEST)T. M. Barlow, Lawrence Livermore Laboratory

The t of Mechanical-Energy-Storage-DeviceAddition on the Performance of E l e c t r i c vehic les

age-ctr!

Robert F. McAlevy, I I I , Robert F. McAlevy & Assoc.

Advanced Flywheel Energy Storage Unit for a HighPower Energy Source for Vehicular Use

Art E. Raynard, Garrett-AiResearch

rage, ec t r i

BREAK

low Cost Flywheel Demonstration

Regenerative Flywheel Energy Storage SystemEdward L. Lustenader, General Electric Company

D . W. r s t , Johns Hopkins University

Materials Program for Fiber Composite FlywheelsJames A. Rinde, Lawrence Livernore Laboratory

463

Session II: FLYWHEELS1:45 p.m. - 3:00 p.m.

Chairman - Robert 0. Woods, Sandia Laboratories

1:45 p.m. Overview of Component DevelopmentRobert O. woods, sandia Laboratories

2:00 p.m. Sandia Composite-Run Flywheel Development£. David Reedy, sandia Laboratories

2:15 p.m. Structural Modeling of a Thick-Rim RotorA. Keith Miller, sandia Laboratories

2:45 p.m. Aerodynamic Heating of High-Speed Flywheels inLow-Density Environments

Mel Baer, sandia Laboratories

Afternoon Free

5:30 p.m. Reception

6:45 p.m. DinnerBiergy—rA Bark Service PerspectiveR. William Hottmeyerr Shenandoan National Park

Session III: FLYWHEELS8:15 p.m. - 10:00 p.m.

Chairman - Robert O. Woods, Sandia Laboratories

8:15 p.m. The Application of Fluid Film Bearings and aPassive Magnetic suspension to Energy storageFlywheels

M. w. Eusepi, Mechanical Technology, Inc.

8:30 p.m. Low-Loss Ball Bearings for Flywheel ApplicationsDavid B. Eisenhaure, Ttie cnarles stark Draper

Laboratories

8:45 p.m. Seal Studies for Advanced Flywheel SystemsI. Anwar, The Franklin institute

9:00 p.m. A Composite Flywheel for vehicle UseFrancis c. Younger, William H. urooecK & Assoc.

9:15 p.m. Progress in Composite Flywheel DevelopmentP. ward Hill, Hercules, Inc.

9:30 p.m. Advance Composite Flywheel for Vehicle ApplicationDonald E. Davis, Rockwell international

9:45 p.m. High-Energy-Density FlywheelD. L. satchweii, Garrett-AiResearch

10:00 p.m. Adjourn

464

Wednesday, October 25

8:00 a.m. Registration

Session IV: SOLAR MECHANICAL9:00 a.m. - 11:45 a.m.

Chairman - Henry H. Dodd, Sandia Laboratories

9:00 a.m. Solar Mechanical Energy Storage ProjectB. c. Caskey, Sandia Laboratories

9:30 a.m. The Band Type Variable Inertia Flywheel andFixed Ratio Power Recirculation Applied to it

David G. Ullman, Union College

10:00 a.m. Cellulosic FlywheelsArthur G. Erdman, Univ. of Minnesota

10:15 a.m. A Concept for Suppression of Nonsynchronous Whirl"in" Flexible FlywheelsJohn M. Vance, Texas ASM University

10:30 a.m. Break

Flywheel Energy Storage Systems

10:45 a.m. Conceptual Design of a Flywheel Energy StorageSystem

Francis C. Younger, William M. Brobeck & Assoc.

11:15 a.m. Residential Flywheel with Turbine SupplyTheodore W. Place, Garrett-AiResearch

12:00 p.m. Lunch

Session V: SUPERCONDUCTING MAGNETIC ENERGY STORAGE1:15 p.m. - 3:15 p.m.

Chairman - John D. Rogers, Jr., Los Alamos ScientificLaboratory

1:15 p.m. Superconductive Diurnal Energy Storage StudiesR. W. Boom, Univ. ot Wisconsin

1:35 p.m. Recent Component Development Studies for Super-conductive Magnetic Energy storage

S. van Sciver, Univ. of Wisconsin

1:55 p.m. Power System Stability Using Superconducting"""Magnetic Energy Storage Dynamic Characteristics

of the BPA SystemLee Cresap, Bonneville Power Corporation

2:10 p.m. Hybrid Computer Study of a SMES Unit forDamping Power System Oscillations

Paul Krause, Purdue University

465

Superconducting Magnetic Energy Storage (Cont'd)

2:25 p.m. Superconducting M&Ljnetic £fcergy StorageJohn D. Rogers, Jr., Los Alamos Scientific Laboratory

3:00 p.m. Superconducting Magnetic Energy Storagefor Power System Stability ApplicaEIons

Carl Chowaniec, Westinghouse Electric Corp.

7:00 p.m. Dinner

Session VI: UNDERGROUND PUMPED HYDROELECTRIC STORAGE8:30 p.m. - 10:00 p.m.

Chairman - George T. Kartsounes, Argonne NationalLaboratory

8:30 p.m. Underground Pumped Hydro Storage: An OverviewShiu-Wing Tam, Argonne National LaboratoryGeorge T. Kartsounes, Argonne National Laboratory

9:10 p.m. Evaluation of One and Two Stage High HeadPump/Turbine for underground Power Stations

John Degnan, Allis-Chalmers Corp.

9:40 p.m. Multistage Turbine-Pump with Controlled Flow RateAlexander Goknman, Univ. of Miami

10:00 p.m. Adjourn

Thursday, October 26

Session VII: COMPRESSED AIR ENERGY. STORAGE9:00 a.m. - 11:40 a.m.

Chairman - Walter v. Loscutoff, Battelle PacificNorthwest Laboratories

9:00 a.m. CAES Program OverviewWalter V. Loscutoff, Battelle Pacific Northwest

Laboratories

9:15 a.m. Fluid Flow and Thermal Analysis for CAES in.PorousRock Reservoirs

L. E. wiles, Battelle Pacific Northwest Laboratories

9:35 a.m. Thermo Mechanical Stress Analysis of Porous RockReservoirs

J. R. Friley, Battelle Pacific Northwest Laboratories

9:55 a.m. Potential Air/Water/Rock Interactions in a PorousMedia CAES Reservoir

J. A. stottlemyre, Battelle Pacific NorthwestLaboratories

10:15 a.m. Break

466

Compressed Air Energy Storage (Cont'd.)

10:30 a.m. Preliminary Design and Stability Criteria forCflES Hard Rock Caverns

P. F. Gnirk, RE/Spec inc.

10:50 a.m. Preliminary Long-Terro Stability Criteria forCAES Caverns in Salt Danes

R. L. Thorns, Louisiana State University

11:10 a.m. Fabric Analysis of Rock Subjected to Cycling withHeated, Compressed Air

H• J. Pincus, Univ. of Wisconsin

11:30 a.m. Numerical Modeling of Behavior of Caverns in Saltfor CAES

S. Serata, Serata Geomechanics Inc.

11:50 a.m. The Design Optimization of Aquifer Reservoir-BasedCAES

Friclerick W. Ahrens, Argonne National Laboratory

12:10 p.m. Lunch

Session VIII: COMPRESSED AIR ENERGY STORAGE1:00 p.m. - 2:50 p.m.

Chairman - Walter V. Loscutoff, Battelle PacificNorthwest Laboratories

1:00 p.m. Advanced CAES Systems StudiesWalter v. Loscutoff, Battelle Pacific Northwest

LaboratoriesM. A. McKinnon, Battelle Pacific Northwest

Laboratories

1:20 p.m. Solar Thermal Augmentation of CAESGerrard T. Fiynn, MIT

1:40 p.m. Evaluation of TurboMachinery for CompressedAir Energy storage Plants

George T. Kartsounes, Argonne NationalLaboratory

Choong S. Kim, Argonne National Laboratory

2:00 p.m. Evaluation of the Use of Reciprocating Enginesin Compressed Air Energy Storage Plants

George T. Kartsounes, Argonne NationalLaboratory

James G. Daly, Argonne National Laboratory

2:40 p.m. Concluding RemarksThomas M. Barlow

2:50 p.m. Adjournment

467

FIRST ANNUALMECHANICAL AND MAGNETIC ENERGY STORAGE

TECHNOLOGY MEETING

The Mimslyn InnLuray, Virginia

October 24-26, 1978Attendees List

ACKERMAN, Sam L.Director, Magnet SystemsGeneral DynamicsP.O. Box 80847San Diego, CA 92138

ADOLFSON, William F.Senior ScientistBooz, Allen & Hamilton, Inc.4330 East-West HighwayBethesda, MD 20014

AHRENS, F.Mechanical EngineerArgonne National Laboratory9700 South Cass AvenueArgonne,IL 60439

ALLEN,BobBattelle Pacific Northwest

Laboratoriesc/o G.C. ChangU.S. Department of Energy600 E Street, NW, Room 416Washington, D.C. 20545

ANWAR, I.Senior Staff EngineerFranklin Institute 20th and ParkwayPhiladelphia, PA 19119

BAER, M.R.Member, Technical StaffSandia LaboratoriesKirtland Air Force BaseAlbuquerque, NM 87185

BAKER, MerlCoordinator of Energy ConservationOak Ridge National LaboratoryBoxXOak Ridge, TN 37830

BARLOW, Thomas M.Program ManagerLawrence Liverrnore Laboratory7000 East AvenueLivermore, CA 94550

BEACH, Raymond F.Project ManagerNASA-Lewis Research Center21000 Brookpark RoadCleveland, OH 44135

BEACHLEY, NormanProfessorUniversity of Wisconsin1513 University AvenueMadison, WI 53706

Beck, CurtBattelle Pacific NorthwestLaboratoriesP.O. Box 999Richmond, WA 99352

BERMAN, IrwinPrincipal EngineerCommonwealth Edison CompanyP.O. Box 767Chicago, IL 60690

BERVIG, D.R.Project EngineerBlack & VeatchP.O. Box 8405Kansas City, MO 64114

BLAY, DominiqueEngineerFrench Atomic Energy Commissionc/o French Embassy1730 Rhode Island Avenue, NWRoom 1217Washington, D.C. 20036

469

BLOMQU1ST, CarlChemical EngineerArgonne National Laboratory9700 South Cass A'. rnueArgonne, IL 60439

BLOOM, Harold L.Project EngineerEnergy Systems Program DepartmentGeneral Electric CompanyBuilding 36, Room 421Scheneetady, NY 12345

BOG ART, LockePlanner, Technological TransferU.S. Department Of Energy600 E Street, NWWashington, D.C. 20545

BOOM, RogerProfessor of Nuclear and Metallic

EngineeringEngineering Experiment StationUniversity of Wisconsin - Madison1500 Johnson DriveMadison, WI 53706

BORTZ, Susan E.ConsultantBradford National Corporation1901 L Street, NVV #301Washington, D.C. 20036

BRAASCH, Richard H.Division SupervisorSandia LaboratoriesDivision 4715Albuquerque, NM 87185

BRANDVOLD, GlenDepartment ManagerSolar Energy ProjectsSandia LaboratoriesDivision 4716Albuquerque, NM 87185

BROBECK, William M.Cha'/man of the BoardWilliam M. Brobeck & Associates1235 Tenth StreetBerkeley, CA 94710

CALLAHAN, Gary D.RE/SPEC, Inc.P.O. Box 725Rapid City, SD 57709

CAMPBELL, JamesProgram Manager, EnergyPropulsion TechnicianU.S. Department of Transportation2100 Second Street, t 'Washington, D.C. 20590

CASKEY, Bill C.Staff MemberSandia LaboratoriesDivision 5716P.O. Box 580Albuquerque, NM 87185

CHANG, George C.Chief, APM BranchDivision of Energy Storage SystemsOffice of Energy TechnologyU.S. Department of Energy600 E Street, NWWashington, D.C. 20545

CHARLWOOD, RobinAcres American, Inc.900 Liberty Bank BuildingBuffalo, NY 14202

CHOWANIEC, CarlWestinghouse Corporation700 Braddock AvenuePittsburgh, PA 15112

CREAGAN, Robert J.Director, Technical AssessmentWestinghouse CorporationGateway CenterPittsburgh, PA 15222

CRESAP, Richard LeeAnalysis EngineerBonneville Power Administration1002 Northeast HolladayPortland, OR 97208

CROTHERS, WilliamLawrence Livermore Laboratory7000 East AvenueLivermore, CA 94550

CUNDY, ThomasSenior Mathematics AnalystSerata Geomechanics1229 8th StreetBerkeley, CA 94709

470

BLOMQUIST, CarlChemical EngineerArgonne National Laboratory9700 South Cass AvenueArgonne, IL 60439

BLOOM, Harold L.Project EngineerEnergy Systems Program DepartmentGeneral Electric CompanyBuilding 36, Room 421Scheneetady, NY 12345

BOG ART, LockePlanner, Technological TransferU.S. Department Of Energy600 E Street, NWWashington, D.C. 20545

BOOM, RogerProfessor of Nuclear and Metallic

EngineeringEngineering Experiment StationUniversity of Wisconsin - Madison1500 Johnson DriveMadison, WI 53706

BORTZ, Susan E.ConsultantBradford National Corporation1901 L Street, NW #301Washington, D.C. 20036

BRAASCH, Richard H.Division SupervisorSandia LaboratoriesDivision 4715Albuquerque, NM 87185

BRANDVOLD, GlenDepartment ManagerSolar Energy ProjectsSandia LaboratoriesDivision 4716Albuquerque, NM 87185

BROBECK, William M.Chairman of the BoardWilliam M. Brobeck & Associates1235 Tenth StreetBerkeley, CA 94710

CALLAHAN, Gary D.RE/SPEC, Inc.P.O. Box 725Rapid City, SD 57709

CAMPBELL, JamesProgram Manager, EnergyPropulsion TechnicianU.S. Department of Transportation2100 Second Street, SWWashington, D.C. 20590

CASKEY, Bill C.Staff MemberSandia LaboratoriesDivision 5716P.O. Box 580Albuquerque, NM 87185

CHANG, George C.Chief, APM BranchDivision of Energy Storage SystemsOffice of Energy TechnologyU.S. Department of Energy600 E Street, NWWashington, D.C. 20545

CHARLWOOD, RobinAcres American, Inc.900 Liberty Bank BuildingBuffalo, NY 14202

CHOWANIEC, CarlWestinghouse Corporation700 Braddock AvenuePittsburgh, PA 15112

CREAGAN, Robert J.Director, Technical AssessmentWestinghouse CorporationGateway CenterPittsburgh, PA 15222

CRESAP, Richard LeeAnalysis EngineerBonneville Power Administration1002 Northeast HolladayPortland, OR 97208

CROTHERS, WilliamLawrence Livermore Laboratory7000 East AvenueLivermore, CA 94550

CUNDY, ThomasSenior Mathematics AnalystSerata Geomechanics1229 8th StreetBerkeley, CA 94709

470

DAVIS, Donald E.Program ManagerRocketdyne DivisionRockwell International6633 Canoga AvenueCanoga Park, CA 91304

DEGNAN, JohnManager, Mechanical DevelopmentsAllis Chalmers Hydro Turbine

DivisionBox 712York, PA X7405

DERBY, RogerProgram ManagerDivision of Energy Storage SystemsOffice of Energy TechnologyU.S. Department of Energy600 E Street, NWWashington, D.C. 20545

DE VINEY, GlenSection EngineerCommonwealth Edison CompanyP.O. Box 767Chicago, IL 60690

DODD, Henry M.Division SupervisorSandia LaboratoriesDivision 5743P.O. Box 5800Albuquerque, NM 87185

EISENHAURE, DavidCharles Stark Draper LaboratoryMail Station 37555 Technology SquareCambridge, MA 02139

EMIGH, C. RobertAssociate Division Leader

Energy TechnologyLas Alamos Scientific LaboratoryP.O. Box 1663Los Alamos, NM 87545

ERDMAN, Arthur G.Associate ProfessorUniversity of MinnesotaMechanical Engineering DepartmentMinneapolis, MN 55455

EUSEPI, Martin W.Program EngineerMechanical Technology, Inc.968 Albany-Shaker RoadLatham, NY 12110

EVANS, Harold E.Branch HeadNASA/GSFCGreenbelt, MD 20811

FARQUHAR, OswaldProfessorUniversity of MassachusettsDepartment of GeologyAmherst, MA 01003

FAUCONNIER, Jean ClaudeEngineerFrench Atomic EnergyCommissionc/o French Embassy1730 Rhode Island Avenue, NWRoom 1217Washington, D.C. 20036

FLYNN, Gerald T.Staff EngineerMassachusetts Institute ofTechnologyLincoln LaboratoryBox 73Lexington, MA 02173

FOSSUM, Arlo F.RE/SPEC, Inc.P.O. Box 725Rapid, City, SD 57709

FRANK, AndrewProfessorUniversity of WisconsinDepartment of Electrical andComputer EngineeringMadison, WI 53706

FRIGO, ArtMechanical EngineerArgonne National Laboratory9700 South Cass AvenueArgonne, IL 60439

FRILEY, John R.Research EngineerBattelle Pacific NorthwestLaboratoriesP.O. Box 999Richland, WA 99352

GAHIMER, JohnProgram ManagerDivision of Energy Storage SystemsOffice of Energy TechnologyU.S. Department of Energy600 E Street, NWWashington, D.C. 20585

471

GIRAMONTI, A.J.Senior EngineerUnited Technologies

Research CenterSilver LaneEast Hartford, CT 06108

GLASER, DickKelsey Hayes2500 Green RoadAnn Arbor, MI 48105

GNIRK, Paul F.RE/SPEC, Inc.P.O. Box 725Rapid City, SD 57709

GOKHMAN, AlexanderAssociate ProfessorDepartment of Mechanical EngineeringUniversity of MiamiCoral Gables, FL 33124

GRASSBERGER, RobertProgram ManagerBDM Corporation2600 Yale, SEAlbuquerque, NM 87106

HAGEN, DavidResearch AssociateUniversity of Minnesota111 Church Street, SEMinneapolis, MN 55455

HAMPSON, ChrisSenior Research ScientistInternational Research and

Technology Corporation7655 Old Springhouse RoadMcLean, VA 22102

HAYES, EdwardVice PresidentKelsey Hayes38481 Hurron River DriveRomulus, MI 48174

HECK, Francis M.Manager, Systems EngineeringWestinghouse Electric CorporationP.O. Box 10864Pittsburgh, PA 15236

HERBERMANN, RichardChief Engineer, Fusion DivisionGrumman Aerospace CorporationMail Stop B-27-35Bethpage, NY 11714

HILL, P. WardSuperintendent, AdvancedTechnologyHercules Inc.P.O. Box 210Cumberland, MD 21502

HOLLIDAY, RobertProgram ManagerDivision of Energy Storage SystemsOffice of Energy TechnologyU.S. Department of Energy600 E Street, NWWashington, D.C. 20545

HOPPIE, Lyle O.EATON-ERCBox 766Southfield, MI 48037

HURLEY, James D.Project Development ManagerGeneral ElectricBuilding 5, Room 425Sehenectady, NY 12345

JOKL, A.L.Supervisor, Physical ScientistMERADCOMAttn: DRDME-EAFort Belvoir, VA 22060

KAPNER, MarkEngineerHittman Associates9190 Redbranch RoadColumbia, MD 21045

KARTSOUNES, GeorgeMechanical EngineerArgonne National Laboratory9700 South Cass AvenueArgonne, IL 60439

KATZ, DonaldUniversity of Michigan2011 Washington AvenueAnn Arbor, MI 48104

472

KOSTAL, KennethSenior Project Engineer AssociateSargent and Lundy55 East MonroeChicago, IL 60525

KRAUSE, PaulElectrical Engineering

DepartmentPurdue UniversityWest Lafayette, IN 47907

KULKARN1, S.V.Lawrence Livermore LaboratoryP.O. Box 808, L-338Livermore, CA 94550

LARRECQ, A.J.PresidentPower Generators, Inc.94 Stokes AvenueTrenton, NJ 08638

LEMMENS, Joseph R.Kinergy Research & DevelopmentP.O. Box 1128Wake Forest, NC 27587

LOSCUTOFF, W.V.Project ManagerBattelle Pacific Northwest

LaboratoriesP.O. Box 999Richland, WA 99352

LUSTENADER, E.L.ManagerGeneral Electric Company1 River RoadSehenectady, NY 12345

MAILLOT, BernardEngineerFrench Atomic Energy

Commissionc/o French Embassy1730 Rhode Island Avenue, NWRoom 1217Washington, D.C. 20036

MANAKER, Arnold M.Mechanical EngineerTennessee Valley Authority1360 Commerce Union Bank

BuildingChattanooga, TN 37401

MARSHALL, H.K.PresidentKinergy Research & DevelopmentP.O. Box 1128Wake Forest, NC 27587

MILLER, A. KeithMember, Technical StaffSandia LaboratoriesDivision 5521Albuquerque, NM 87185

MILNER, AlanStaff MemberMassachusetts Instituteof TechnologyLincoln LaboratoryP.O. Box 73Lexington, MA 02173

MC ALEVY, Roberty F. IllSenior AssociateRobert F. McAlevy III &Associates1204 Bloomfield StreetHoboken, NJ 07030

MC COY, H. E.Oak Ridge National Laboratory132 Balboa CircleOak Ridge, TN 37830

MC DONALD, Alan T.ProfessorSchool of Mechanical EngineeringPurdue UniversityWest Lafayette, IN 47907

MC SPADDEN, William R.Senior Research EngineerBattelle Pacific NorthwestLaboratoryP.O. Box 999Richland, WA 99352

NASH-WEBBER, James L.Facilities Project ManagerMassachusetts Institute of TechnologyEnergy LaboratoryP.O. Box 69Cambridge, MA 02139

NEWHOUSE, Norman L.Design EngineerBrunswick Corporation4300 Industrial AvenueLincoln, NE 68504

473

NICOL, JamesVice PresidentA.D. LittleAcorn ParkCambridge, MA 02140

NORMAN, Thomas A.Program EngineerU.S. Postal Service11711 Parklawn DriveRockville, MD 20852

OMOHUNDRO, Laura L.Vice PresidentKinergy Research & DevelopmentP.O. Box 1128Wake Forest, NC 27587

PARADIS, L.R., Jr.Principal EngineerRaytheon CompanyHartwell RoadBedford, MA 01730

PARSONS, Lawrence F.Civil EngineerDepartment of the InteriorBureau of ReclamationWashington, D.C. 20240

PATRICK, A.J.Manager, Energy DevelopmentAvco Systems Division201 Lowell StreetWilmington, MA 01887

PAX, CharlesU.S. Department of Energy20 Massachusetts AvenueWashington, D.C. 20545

PERRAM, RobertProgram ManagerAerospace Corporation20030 Century BoulevardGermantown, MD 20767

PEZDIRTZ, George F.Director, Division of

Energy Storage SystemsOffice of Energy TechnologyU.S. Department of Energy600 E Street, NWWashington, D.C. 20545

PINCUS, Howard J.ProfessorDepartment of GeologicalSciencesUniversity of Wisconsin -MilwaukeeMilwaukee, WI 53201

PLACE, Theodore W.Project EngineerGarrett AiResearch Corporation2525 West 190th StreetTorrance, CA 90505

POST, DaveUnion Carbide CorporationY-12 PlantOak Ridge, TN 37830

RABENHORST, David W.Flywheels Program ManagerApplied Physics LaboratoryJohn Hopkins UniversityJohns Hopkins RoadLaurel, MD 20810

RAYNARD, Arthur E.Senior Project EngineerGarrett AiResearch Corporation2525 West 190th StreetTorrance, CA 90505

REDDICK, William C.Scientific AdvisorU.S. Department Of Energy20 Massachusetts Avenue, NWWashington, D.C. 20545

REEDY, E. David, Jr.Member, Technical StaffSandia LaboratoriesDivision 5844KAFB EastAlbuquerque, NM 87185

RINDE, J.A.ChemistLawrence Livermore LaboratoryP.O. Box 808 L-338Livermore, CA 94550

ROGERS, John D.Los Alamos Scientific LaboratoryOrg. CTR-9, Mail 464Los Alamos, NM 87545

474

NICOL, JamesVice PresidentA.D. LittleAcorn ParkCambridge, MA 02140

NORMAN, Thomas A.Program EngineerU.S. Postal Service11711 Parklawn DriveRockville, MD 20852

OMOHUNDRO, Laura L.Vice PresidentKinergy Research & DevelopmentP.O. Box 1128Wake Forest, NC 27587

PARADIS, L.R., Jr.Principal EngineerRaytheon CompanyHartwell RoadBedford, MA 01730

PARSONS, Lawrence F.Civil EngineerDepartment of the InteriorBureau of ReclamationWashington, D.C. 20240

PATRICK, A.J.Manager, Energy DevelopmentAvco Systems Division201 Lowell StreetWilmington, MA 01887

PAX, CharlesU.S. Department of Energy20 Massachusetts AvenueWashington, D.C. 20545

PERRAM, RobertProgram ManagerAerospace Corporation20030 Century BoulevardGermantown, MD 20767

PEZDIRTZ, George F.Director, Division of

Energy Storage SystemsOffice of Energy TechnologyU.S. Department of Energy600 E Street, NWWashington, D.C. 20545

PINCUS, Howard J.ProfessorDepartment of GeologicalSciencesUniversity of Wisconsin -MilwaukeeMilwaukee, WI 53201

PLACE, Theodore W.Project EngineerGarrett AiResearch Corporation2525 West 190th StreetTorrance, CA 90505

POST, DaveUnion Carbide CorporationY-12 PlantOak Ridge, TN 37830

RABENHORST, David W.Flywheels Program ManagerApr lied Physics LaboratoryJohn Hopkins UniversityJohns Hopkins RoadLaurel, MD 20810

RAYNARD, Arthur E.Senior Project EngineerGarrett AiResearch Corporation2525 West 190th StreetTorrance, CA 90505

REDDICK, William C.Scientific AdvisorU.S. Department Of Energy20 Massachusetts Avenue, NWWashington, D.C. 20545

REEDY, E. David, Jr.Member, Technical StaffSandi£ LaboratoriesDivision 5844KAFB EastAlbuquerque, NM 87185

RINDE, J.A.ChemistLawrence Livermore LaboratoryP.O. Box 808 L-338Livermore, CA 94550

ROGERS, John D.Los Alamos Scientific LaboratoryOrg. CTR-9, Mail 464Los Alamos, NM 87545

474

ROSE, ShelleySecretaryLawrence Livermore Laboratory7000 East AvenueLivermore, CA 94550

ROSNER, CarlInter-Magnetics General

CorporationP.O. Box 566Guilderland, NV 12084

SAPOWITH, Alan D.Section ChiefAvco Systems Division201 Lowell StreetWilmington, MA 01887

SATCHWELL, David L.Senior Development EngineerGarrett AiResearch Corporation2525 West 190th StreetTorrance, CA 90505

SCHILDKNECHT, HaroldMember, Technical StaffSandia LaboratoriesOrg. 2324KAFB EastAlbuquerque, NM 87185

SCHLIEBEN, Ernest W.PresidentVon Research40 West Lafayette StreetTrenton, NJ 08608

SCHMIDT, Franklin R.ConsultantBradford National Corporation#2 Research PlaceRockville, MD 20850

SCHWARTZ Martin W.EngineerLawrence Livermore LrboratoryP.O. Box 808Livermore, CA 94550

SERATA, ShoseiPresidentSerata Geomechanics1229 8th StreetBerkeley, CA 94709

SMITH, B.B.Union Carbide CorporationY-12 PlantOak Ridge, TN 37830

SPISAK, A.J.Manager, Advanced ProgramWestinghouse Corporation700 Braddock AvenuePittsburgh, PA 15112

STILLER, PaulResearch Project EngineerWestinghouse Corporation*700 Braddock AvenuePittsburgh, PA 15112

STONE, Richard G.Associate Department HeadLawrence Livermore LaboratoryP.O. Box 808-L-123Livermore, CA 94550

STOTTLEMYRE, JimResearch ScientistBattelle Pacific NorthwestLaboratoriesP.O. Box 999Richland, WA 99352

SULLIVAN, Cornelius M.Staff MemberMassachusetts Institute ofTechnologyLincoln LaboratoryBuilding D-250P.O. Box 73Lexington, MA 02173

SURABIAN, GregConsultantBradford National Corporation1901 L Street NW, #301Washington, D.C. 20036

SWISHER, James H.Assistant Director forPhysical Storage SystemsDivision of Energy Storage SystemsOffice of Energy TechnologyU.S. Department of Energy600 E Street, NWWashington, D.C. 20545

TAM, S.W.Assistant Metallurgical EngineerArgonne National Laboratory9700 South Cass AvenueArgonne, IL 60439

475

THOMPSON, JaneConsultantBradford National Corporation#2 Research PlaceRockville, MD 20850

THOMPSON, PhilipProgram ManagerDivision of Energy Storage SystemsOffice of Energy TechnologyU.S. Department of Energy600 E Street, NWWashington, D.C. 20545

THOMS, Robert L.Louisiana State UniversityInstitute for

Environmental StudiesBaton Rouge, LA 70803

ULLMAN, DavidAssistant ProfessorDepartment of Mechanical EngineeringUnion CollegeSchenectady, NY 12308

UPTON, Joseph W.Senior Research ScientistBattelle Pacific Northwest

LaboratoriesBox 999Richland, VVA 99352

VAN SCIVER, S.W.ScientistUniversity of Wisconsin1500 Johnson DriveMadison, Wl 53711

VAN ZANTEN, M.M. scNetherlands Energy

Research FoundationWesterduinweg 3Petten, Netherlands 1755 LE

VANCE, John M.Associate ProfessorTexas A & MMechanical EngineeringCollege Station, TX 77843

VERGARA, Rudolfo D.Research ScientistBattelle Columbus Laboratories505 King AvenueColumbus, OH 43201

W EINSTEIN, Kenneth D.Booz, Allen & Hamilton, Inc.Energy & Environment Division4330 East-West HighwayBethesda, MD 20014

WEISS, Joel A.PhysicistOffice of Solar ProgramsU.S. Department of Energy600 E Street, NWWashington, D.C. 20545

WHITCOMB, MikeEngineerHittman Associates9190 Redbraneh RoadColumbia, MD 21405

WHITMAN, HowardCharles Stark Draper Laboratory555 Technology SquareCambridge, MA 02139

WILES, LarryDevelopment EngineerBattelle Pacific NorthwestLaboratoriesP.O. Box 999Richland, WA 99352

WILKINSON, John P.D.Manager, Solid Mechanical UnitGeneral Electric CompanyP.O. Box 43Schenectady, NY 12301

WILLETT, David C.Vice PresidentAcres American, Inc.900 Liberty Bank BuildingBuffalo, NY 14202

WILLIAMS, John R.Geotechnical EngineerDames & Moore4 Militia DriveLexington, MA 02173

WISSING, Thomas J.Administrator, GovernmentResearch and DevelopmentEaton CorporationP.O. Box 766Southfield, MI 48037

476

KOSTAL, KennethSenior Project Engineer AssociateSargent and Lundy55 East MonroeChicago, IL 60525

KRAUSE, PaulElectrical Engineering

DepartmentPurdue UniversityWest Lafayette, IN 47907

KULKARNI, S.V.Lawrence Livermore LaboratoryP.O. Box 808, L-338Livermore, CA 94550

LARRECQ, A.J.PresidentPower Generators, Inc.94 Stokes AvenueTrenton, NJ 08638

LEMMENS, Joseph R.Kinergy Research 6c DevelopmentP.O. Box 1128Wake Forest, NC 27587

LOSCUTOFF, W.V.Project ManagerBattelle Pacific Northwest

LaboratoriesP.O. Box 999Richland, WA 99352

LUSTENADER, E.L.ManagerGeneral Electric Company1 River RoadSehenectady, NY 12345

MAILLOT, BernardEngineerFrench Atomic Energy

Commissionc/o French Embassy1730 Rhode Island Avenue, NWRoom 1217Washington, D.C. 20036

MANAKER, Arnold M.Mechanical EngineerTennessee Valley Authority1360 Commerce Union Bank

BuildingChattanooga, TN 37401

MARSHALL, H.K.PresidentKinergy Research & DevelopmentP.O. Box 1128Wake Forest, NC 27587

MILLER, A. KeithMember, Technical StaffSandia LaboratoriesDivision 5521Albuquerque, NM 87185

MILNER, AlanStaff MemberMassachusetts Instituteof TechnologyLincoln LaboratoryP.O. Box 73Lexington, MA 02173

MC ALEVY, Roberty F. IllSenior AssociateRobert F. McAlevy III &Associates1204 Bloomfield StreetHoboken, NJ 07030

MC COY, H. E.Oak Ridge National Laboratory132 Balboa CircleOak Ridge, TN 37830

MC DONALD, Alan T.ProfessorSchool of Mechanical EngineeringPurdue UniversityWest Lafayette, IN 47907

MC SPADDEN, William R.Senior Research EngineerBattelle Pacific NorthwestLaboratoryP.O. Box 999Riehland, WA 99352

NASH-WEBBER, James L.Facilities Project ManagerMassachusetts Institute of TechnologyEnergy LaboratoryP.O. Box 69Cambridge, MA 02139

NEWHOUSE, Norman L.Design EngineerBrunswick Corporation4300 Industrial AvenueLincoln, NE 68504

473

ROSE, ShelleySecretaryLawrence Livermore Laboratory7000 East AvenueLivermore, CA 94550

ROSNER, CarlInter-Magnetics General

CorporationP.O. Box 566Guilderland, NY 12084

SAPOWITH, Alan D.Section ChiefAvco Systems Division201 Lowell StreetWilmington, MA 01887

SATCHWELL, David L.Senior Development EngineerGarrett AiResearch Corporation2525 West 190th StreetTorrance, CA 90505

SCHILDKNECHT, HaroldMember, Technical StaffSandia LaboratoriesOrg. 2324KAFB EastAlbuquerque, NM 87185

SCHLIEBEN, Ernest W.PresidentVon Research40 West Lafayette StreetTrenton, NJ 08608

SCHMIDT, Franklin R.ConsultantBradford National Corporation#2 Research PlaceRockville, MD 20850

SCHWARTZ Martin W.EngineerLawrence Livermore LaboratoryP.O. Box 808Livermore, CA 94550

SERATA, ShoseiPresidentSerata Geomechanics1229 8th StreetBerkeley, CA 94709

SMITH, B.B.Union Carbide CorporationY-12 PlantOak Ridge, TN 37830

SPISAK, A.J.Manager, Advanced ProgramWestinghouse Corporation700 Braddock AvenuePittsburgh, PA 15112

STILLER, PaulResearch Project EngineerWestinghouse Corporation1

700 Braddock AvenuePittsburgh, PA 15112

STONE, Richard G.Associate Department HeadLawrence Livermore LaboratoryP.O. Box 808-L-123Livermore, CA 94550

STOTTLEMYRE, JimResearch ScientistBattelle Pacific NorthwestLaboratoriesP.O. Box 999Riehland, WA 99352

SULLIVAN, Cornelius M.Staff MemberMassachusetts Institute ofTechnologyLincoln LaboratoryBuilding D-250P.O. Box 73Lexington, MA 02173

SURABIAN, GregConsultantBradford National Corporation1901 L Street NW, #301Washington, D.C. 20036

SWISHER, James H.Assistant Director forPhysical Storage SystemsDivision of Energy Storage SystemsOffice of Energy TechnologyU.S. Department of Energy600 E Street, NWWashington, D.C. 20545

TAM, S.W.Assistant Meialurgical EngineerArgonne National Laboratory9700 South Cass AvenueArgonne, IL 60439

475

WOODS, R.O.Member, Technical StaffSandia LaboratoriesDivision 4715Albuquerque, NM 87185

YONK, Alan K.Senior GeologistSargent and Lundy55 East Monroe StreetChicago, IL 60603

YOUNGER, Francis C.PresidentWilliam M. Brobeck & Associates1235 10th StreetBerkeley, CA 94710

477

•n. n a n nanm era •