intelligent well technology in underground gas...

4 Oilfield Review

Intelligent Well Technology in Underground Gas Storage

Kenneth BrownPittsburgh, Pennsylvania, USA

Keith W. ChandlerJohn M. HopperLowell ThronsonFalcon Gas Storage Company, Inc.Houston, Texas, USA

James HawkinsMidland, Texas

Taoufik ManaiParis, France

Vladimír OnderkaRWE Transgas NetBrno, Czech Republic

Joachim WallbrechtBEB Transport und Speicher Service GmbHHannover, Germany

Georg ZanglBaden, Austria

For help in preparation of this article, thanks to EdmundKnolle, Falcon Gas Storage Company, Houston; andMonsurat Ottun, Houston.BorView, BlueField, DECIDE!, ECLIPSE, ELANPlus, FMI(Fullbore Formation MicroImager), NODAL, Petrel andPIPESIM are marks of Schlumberger.Excel is a mark of Microsoft Corporation.

Intelligent well technologies are ideal for underground gas-storage facilities.

Formation properties have been determined; storage capacity and deliverability

can be modeled; and analytical tools can track historical production trends.

These technologies provide efficient, cost-effective storage and delivery systems,

helping secure the position of natural gas as a dependable energy resource.

When it comes to applying intelligent welltechnologies to oil and gas production, a primeobjective is maximizing the value of a continuallydiminishing asset. For underground natural gas-storage facilities, the application of these smarttechnologies can be significantly different,primarily because the gas reservoir can bereplenished. Thus, it is the ability to repeatedlyinject natural gas into and withdraw it fromunderground storage at high rates that must beoptimized and intelligently managed.

Formation properties define the optimal levelat which a well flows at high recovery rates. As thestored natural gas is recovered from the reservoir,the pressure decreases and flow rates fall. Cushiongas, the gas that remains in place betweeninjection and withdrawal cycles, ensures thatthere is sufficient pressure to maintain the desiredminimum flow rates on withdrawal. The pressureand volume provided by the cushion gas alsodiminish the likelihood of water influx into the gascap and can prevent gas/water contact movement.Because the most expensive component of anunderground gas-storage (UGS) facility can be thecushion gas, minimizing its volume andunderstanding the reservoir well enough to definethe efficient operating range can reduce theoverall development cost of a storage project, aswell as greatly enhance project profitability.

In hydrocarbon production, intelligent welltechnologies allow reservoir engineers to useinformation such as decline curves, material-balance relationships, inflow-performance-relationship (IPR) curves and reservoirsimulations and models—all in real time oralmost real time.1 A sophisticated system mayautomatically take corrective action or alert theoperator that intervention is warranted. Theultimate goal of intelligent production wells is todeliver more oil and gas with greater efficiency—at a lower cost.

While UGS facilities also benefit from effi -ciencies and lower cost provided by intel ligentwell technologies, they are not operated tomaximize the recovery of hydrocarbon. In fact,gas-storage operations in many parts of the worldare more analogous to a bank than a producingreservoir. Just as currency flows into and out of abank, assets in the form of natural gas flow intoand out of the storage reservoir. When calledupon, sometimes months after injection into astorage reservoir, but increasingly within a fewdays or even hours, the gas is delivered to a buyerwho supplies it to industrial and residentialcustomers. Banks have automated the flow ofcurrency and capital between institutions andusers; similarly, storage facilities are automatingthe flow of natural gas between producers and consumers.

61459schD4R1.qxp:Format Layout V7.2 5/15/08 9:02 AM Page 4

Spring 2008 5

component in the secure delivery of natural gas,these facilities must be managed appropriately.

Maintaining reliable natural gas supplies hasrecently become a geopolitical priority in manyareas of the world. Governmental regulations,such as those in the European Union, have hadincreased influence on how the UGS industryconducts business. Intelligent well technology isbeing adopted as a natural by-product of thesedevelopments because it helps facilitateautomatic storage and delivery of natural gas,improves the operational efficiencies of these

facilities, and assists in optimizing themanagement of assets (gas) in the ground.

The degree of implementation varies between different operations, and not all UGSfacilities operate using these relatively newtechnologies. However, the improved operatingperformance that has been demonstrated isprompting operators to retrofit and upgrademany older storage operations—sometimesyielding unexpected benefits.

UGS operations also differ from traditionalgas production because the wells must be able towithstand high injection pressures, somethingrarely experienced in producing wells, and thewithdrawal rates from UGS can be 5 to 10 timesgreater. UGS wells have long life expectancy;therefore maintaining well integrity and reser voirintegrity are crucial aspects of successfuloperations. Due to the rapidly changing opera -tional modes—injection to withdrawal—theoperator must be reactive and act quickly to avoidwell and reservoir mechanical damage. As a vital

1. Hydrocarbon depletion occurs in a predictable mannerbased on formation properties and completion hardware.The rate of decline of reserves can be plotted to define adecline curve.Material balance is an expression for conservation of mass. The amount of mass leaving a control volume

is equal to the amount of mass entering the volume,minus the amount of mass accumulated in the volume.Through material balance, reservoir pressures measuredover time can be used to estimate the volume of hydro- carbons remaining.Inflow-performance relationship, IPR, is a tool used inproduction engineering to assess gas-well performance

by plotting the well production rate against the flowingbottomhole pressure (BHP). The data required to createthe IPR are obtained by measuring the production ratesunder various drawdown pressures during a multiratetest. The reservoir fluid composition and behavior of thefluid phases under flowing conditions determine theshape of the curve.

> Image courtesy of Falcon Gas Storage Company, Inc.


After a brief review of gas-storage basics, thisarticle examines the different levels of intelligentwell technology being applied to undergroundgas-storage operations in North America andEurope. We present case studies showing howreal-time data are used to identify damage instorage wells and how the implementation of newoptimization and surveillance techniques hasimproved performance. Also included is adescription of a cutting-edge, automated opera -tion that integrates three levels of intelligent well technology.2

An Underground Gas-Storage PrimerTraditionally, natural gas has been considered aseasonal fuel because of the higher demand forheating during winter months. Beginning in the1940s, the US natural gas industry recognizedthat long-distance pipeline capacity was notsufficient to supply natural gas to largepopulation centers during peak-demand periods.To balance the gas-demand cycle, a gas-storagenetwork was developed to inject gas intounderground storage facilities when demand waslow and to release gas during periods of highdemand. This buffering of demand is referred toas peak-shaving.

Underground gas storage, however, has beenavailable almost as long as long-distancepipelines. In 1915, natural gas was firstsuccessfully stored underground in WellandCounty, Ontario, Canada. Several wells in apartially depleted gas field were reconditioned,and gas was injected into the reservoir during thesummer and withdrawn the following winter.

In 1916, Iroquois Gas Company placed theZoar field, south of Buffalo, New York, USA, intooperation as a storage site, and it is still inoperation today. In 1919, the Central KentuckyNatural Gas Company injected gas into thedepleted Menifee gas field in Kentucky, USA. By1930, nine storage sites in six different stateswere in operation with a total capacity of about18 Bcf [510 million m3]. Before 1950, essentiallyall underground gas storage consisted of reusedpartially or fully depleted gas reservoirs.

Today, the two primary types of undergroundgas-storage locations are caverns and porousreservoirs. Leached salt caverns and abandonedmines account for a small portion of the totalstorage capacity, while depleted oil and gasreservoirs and saline aquifers are by far the mostcommon UGS medium (left). Salt-cavern storage,better suited for high-rate delivery and injection,is primarily used for peak-day delivery purposes.3

Typically, 20 to 30% of the gas must remain inplace to maintain the structural stability of thecavern. Saline aquifers can provide high-ratedelivery, but cushion-gas requirements aresignificant, ranging from 50 to 80% of the totalstorage capacity. By far the most common type ofstorage, depleted hydrocarbon reservoirs areused for seasonal delivery or buffering highdemand. Typically, 30 to 50% of the storagecapacity must be maintained as cushion gas.4

In recent years, UGS withdrawal practiceshave changed in the USA because of theincreased use of natural gas for electricitygeneration. Drawdown during summer months ishigher than in the past because natural gas isbeing used to generate electricity for airconditioning and cooling requirements (left). Inmany ways, this has altered the scope of gasstorage in the USA. UGS facilities that arelocated in proximity to gas-fired power plants areused to moderate the supply for seasonal, as wellas hourly, variations. On a daily basis, gas instorage can be tapped during high-demandperiods and stored during low-demand periods.Commercial pipelines may be incapable ofsupplying sufficient quantities of gas during thepeak periods—or putting away gas duringperiods of low demand—but the UGS facility canmake up the shortfall in either case.

6 Oilfield Review

> Underground gas storage by type. UGS facilities can take several forms,but depleted hydrocarbon reservoirs and saline aquifers make up 96% ofthe global supply. The choice of storage type can be driven by availability:aquifers and salt caverns make up 34% of Western Europe’s storage capacitycompared with just 14% in the USA where there is greater access todepleted fields. (Adapted from Wallbrecht, reference 6.)

Depleted oil and gas fields81.6%

Aquifers14.5%

Salt caverns 3.9% Abandoned mines0.02%

Global Working-Gas Volume Distribution by Storage Types

> The cyclicity of natural gas usage. As a source of residential heating in the USA (blue), natural gas from storage peaks during winter months. Whenused for generating electricity to provide cooling (red), usage peaks duringsummer months. Commercial usage, driven by temperature (black), tracksresidential demand. Also note the rise in natural gas usage for electricitygeneration in successive years. [Adapted from http://www.eia.doe.gov/oil_gas/natural_gas/info_glance/natural_gas.html (accessed February 29, 2008.)]

Gas

usag

e, Tc

f

1.0

0.8

0.6

0.4

0.2

0

Gas

usag

e, b

illio

n m

3

0

4

8

12

16

20

24

28

Year2004 2005 2006 2007 2008

Residential

CommercialElectric power


Spring 2008 7

UGS is not just a North Americanphenomenon—storage facilities are currentlyopera ting in 33 countries—although the USA hasby far the greatest number. As an energy source,natural gas utilization in residential andcommercial sectors of Western Europe hasexceeded 44%, highlighting the importance ofmaintaining a secure, uninterrupted supply. InFrance, storage facilities have at times suppliedmore than half the residential gas needed tomeet temperature-driven demand.5

Western Europe has recently experienced arise in gas trading between holding companiesand market suppliers. The use of gas-storagefacilities is often driven by short-term buying andselling, rather than traditional peak-shaving.Profitability for both buyers and suppliers isdetermined by the ability of the UGS facility tostore and deliver gas on demand in a cost-effective manner.

In 1997, there were 580 UGS sites worldwide,of which 448 were in depleted reservoirs. In 2006,of the estimated 606 UGS sites, the number indepleted reservoirs had grown to 495.6 In 1996,there were 92 UGS operations in Europe,excluding Russia. By 2006, the total number hadgrown to 127—a 38% increase. The working-gasvolume in storage facilities in the same area grewfrom 60.6 million m3 [2.14 Bcf] to 110.5 million m3

[3.9 Bcf], an 82% increase.Although the USA has had a slight decrease in

the number of UGS sites between 1995 and 2004,its total storage capacity has experienced amarginal increase through improved fieldutilization and retrofitting of existing facilities(top right). Many of these older UGS operationswere developed before the introduction of thereservoir modeling tools available today. Advancesin sensor technology and surface equipment arebeing applied to these older facilities, makingthem “smarter” and more versatile.7

How Smart Is an Intelligent Well?The relative intelligence of gas-storageoperations can be grouped into three levels.Level I, automated data flow, is reactive: receivedata, analyze data and respond. Level II,surveillance and optimization, is reflective butfocuses on action: analyze data, compare andvalidate models, manage models and determinenecessary courses of action. Level III can bedescribed as the digital oil field: integrateprocesses, optimize, automate and operateremotely, where it is applicable, in a proactivemanner (bottom right).

2. For more on intelligent field applications in producingwells: Dyer S, El-Khazindar Y, Reyes A, Huber M, Raw Iand Reed D: “Intelligent Completions—A Hands-Off Management Style,” Oilfield Review 19, no. 4 (Winter 2007/2008): 4–17.

3. For more on underground gas storage: Bary A,Crotogino F, Prevedel B, Berger H, Brown, K, Frantz J,Sawyer W, Henzell M, Mohmeyer K-U, Ren NK, Stiles Kand Xiong H: “Storing Natural Gas Underground,” OilfieldReview 14, no. 2 (Summer 2002): 2–17.

4. For more on current trends in UGS: https://www.ferc.gov/EventCalendar/Files/20041020081349-final-gs-report.pdf(accessed January 23, 2008).

5. Chabrelie MF, Dussaud M, Bourjas D and Hugout B:“Underground Gas Storage: Technological Innovationsfor Increased Efficiency,” http://217.206.197.194:8190/wec-geis/publications/default/tech_papers/17th_congress/2_2_09.asp (accessed December 12, 2007).

6. Wallbrecht J: “Underground Gas Storage,” InternationalGas Union, Report of Working Committee 2, Basic UGSActivities, presented at the 23rd World Gas Conference,Amsterdam, June 5–9, 2006, http://www.igu.org/html/wgc2006/WOC2database/index.htm (accessed March 27, 2008).

7. For more on recent sensor and instrumentation develop-ments in intelligent wells: Bouleau C, Gehin H, Gutierrez F,Landgren K, Miller G, Peterson R, Sperandio U,Traboulay I and Bravo da Silva L: “The Big Picture: Integrated Asset Management,” Oilfield Review 19, no. 4(Winter 2007/2008): 34–48.

> A growing supply. The working-gas volume has grown steadily in the past 35 years with most of theincrease occurring outside North and South America (blue), especially in Eastern Europe and theMiddle East (black), which includes Russia. Current projections indicate that capacity is insufficient tomeet the long-term demand and increased growth is required. (Adapted from Wallbrecht, reference 6.)

350

300

250

200

150

100

50

01970 1975 1980 1985 1990 1995 2000 2005

Year

Wor

king

-gas

vol

ume,

109 m

3

TotalEastern Europeand Middle EastAmericasWestern Europe

Middle East 2%

North and SouthAmerica 35%

Asia 2%

Western Europe 19%

Eastern Europe 42%

Working-Gas Volume Distribution by Region in 2004

> Levels of intelligence. Three levels can be identified in theimplementation of intelligent well technology. Each level brings addedcomplexity and builds upon the others. The most comprehensive is thedigital oil field with optimization and opportunities for automation.

Reservoir

Wells

Gatheringsystem

Facilities

Level I, Reactive

Dynamic data: SCADA systemStatic data: well, reservoir

and technology

Level II, Active

Level III, Proactive

Digital oil field

Surveillance and optimization


Level-I intelligence begins by developingconfidence in the data. Supervisory control anddata acquisition (SCADA) systems can be foundin most UGS operations. These computerizednetworks remotely acquire well data such as flowrate, pressure and flowing volume, and controltransmission of gas throughout a pipeline system.With millions of data points thus acquired, it isimpossible to manually validate all theinformation. Automating quality control of thedata flow is a necessity.

Software for traditional oil and gas produc -tion is often used for UGS applications to identifyperformance problems as well as monitorindividual wells, evaluate completions andcharacterize the reservoir. Trend analysis andtype-curve matching are frequently used in theseprograms. However, most petrophysical programsare poorly equipped to handle the huge volume ofSCADA data coming from UGS operations. Also,they cannot effectively deal with noisy dataresulting from sensor errors and spuriousresponses (above).8 Since proper use of theseapplications often depends on the ability to

identify the onset of linear trends over time orclearly identify subtle features in various type-curves, the data must be cleansed and reduced sothat proper identification of such features can beaccomplished (next page). Therefore, intelligentdata reduction is applied before importing thedata into these programs.

The data provide insight for evaluating therelative health of individual wells, as well as thatof the producing field. The repeated cyclingability of gas-storage wells—periods of injectionfollowed by periods of production—is afundamental difference between producingreservoirs and storage reservoirs. Occasionally,the storage wells remain static for varyinglengths of time and the collected data can betreated as a conventional short-time builduptest. Changes that occur from cycle to cycle canbe indicative of problems developing inindividual wells, in the reservoir or in the surfaceequipment. By analyzing these data, thepresence of damage can be recognized, andremediation plans implemented.

Early uses of electronic flow measurement(EFM) data in UGS fields clearly demonstratedtheir value in monitoring well performance,conducting routine surveillance and identifyingoperational problems.9 In 2002, as part of a GasTechnology Institute (GTI) sponsored study,Schlumberger engineers used EFM data todevelop a reasonably accurate, cost-effectivemeans of detecting wellbore damage in gas-storage wells.10 The impetus for this work was thefairly common practice of performing surfacebackpressure tests in UGS wells to evaluatedamage on a very infrequent basis—testingevery 1 to 3 years was typical. The infrequentnature of such testing made it impossible todetermine incremental damage or suddenchanges over reasonable time frames, such asduring an injection or withdrawal cycle.Determining changes in damage in near realtime is important, since damage might beoccurring during injection or withdrawal, orduring the changeover from one to the other. Thiswork made it possible to estimate damage levels

8 Oilfield Review

8. Holland J, Oberwinkler C, Huber M and Zangl G: "Utilizingthe Value of Continuously Measured Data," paper SPE90404, presented at the SPE Annual Technical Confer-ence and Exhibition, Houston, September 26–29, 2004.

9. Brown KG and Meikle DE: “The Value of Wellhead Electronic Flow Measurement in Gas Storage Fields,”paper SPE 31000, presented at the SPE Eastern RegionalConference and Exhibition, Morgantown, West Virginia,USA, September 17–21, 1995.

10. Brown KG and Sawyer WK: “Novel Surveillance HelpsOperators Track Damage,” paper SPE 75713, presented at the SPE Gas Technology Symposium, Calgary, April 30–May 2, 2002.

The z-factor, or ideal-gas deviation factor, is the departureof a gas behavior from that of the ideal gas law.The gas formation volume factor represents the fractionalchange of volume per unit change in pressure. The coefficient of isothermal compressibility is a measure ofthe relative volume change of a fluid or solid in responseto a pressure (or mean stress) change. The gas formationvolume factor is used to convert a volume of gas at reser-voir conditions to a volume of gas at standard (surface)conditions, since the volume of any gas depends on itspressure and temperature.

11. Spivey JP, Brown KG, Sawyer WK and Gilmore RG:“Identifying the Timing and Sources of Damage in GasStorage Wells Using Smart Storage Technology,” paperSPE 97070, presented at the SPE Annual Technical Conference and Exhibition, Dallas, October 9–12, 2005.

12. The pseudocritical temperature and the pseudocriticalpressure are the pressure and temperature conditions of a multicomponent mixture at which liquid and vaporcannot be distinguished (because the properties areidentical at this combination of pressure and temperature).

> Cleaning and reducing noisy data. Data sampled at high frequency may exhibit considerable noise; sensor errors also may occur. Even though most of these data are usable, the spikes and noise in the choke size (top left, blue), bottomhole pressure (top left, red), wellhead pressure (bottom left, green) andwellhead temperature (bottom left, black) make them difficult to use with modeling programs. After cleansing and reduction, the data (right) are usable inpetrophysical software. (Adapted from Holland et al, reference 8.)

Time Time

Wellhead pressureWellhead temperature

Bottomhole pressureChoke size

Data Sampled at High Frequency Data Cleansed and Reduced


Spring 2008 9

in UGS wells on a more frequent basis than waspreviously possible.

In a more recent GTI study, Schlumbergerengineers developed a method to utilize EFM datato continuously track and identify wellboredamage over time.11 Developments in intelligentwell technology, including sensor improvementsand real-time streaming data, have beencombined with the experience gained from earlierEFM studies to develop damage-identificationtechniques for use in UGS facilities, such as oneoperated by Columbia Gas TransmissionCorporation in the northeast region of the USA.The reservoir consisted of a consolidatedsandstone, which formed a strati graphic trap withan average thickness of 10 ft [3 m], averageporosity of 10% and average permeability of150 mD. Of the 20 wells in the field, five wereidentified as key wells for the purpose of the study.

A SCADA system collected high-frequencypressure and flow-rate EFM data from thewellheads at 15-second intervals. The operatorcollected monthly records, with as many as115,000 data points per well per day, and suppliedthem to Schlumberger engineers. A softwareroutine parsed the field-wide data into individualfiles for each well, reduced the datasets to a moremanageable volume and automated the process ofmaking the information useable.

Because each well generated 3 million datapoints over the course of the study, a routine wasdeveloped just to handle the raw data. It per -formed three primary functions: a gas-propertiescorrelation, a bottomhole-pressure calculation anddata processing. The gas-properties correla tionmodule calculates pseudo critical temperature andpressure, z-factor, coefficient of isothermalcompressibility, gas formation volume factor and

gas density, viscosity and pseudo pressure.12 It alsoformats the data for export to Excel workbooks.

The high-frequency EFM data are aggregatedover 10-minute intervals during flow. Thesoftware routine computes average flow rate andaverage pressure, along with the standarddeviation of these quantities. It flags outlyingdata points as invalid if there is a flow reversal orif there is a mix of zero and nonzero EFM-rate measurements.

During shut-in periods, a variable-widthwindow is applied to the data to give approxi -mately the same number of points for each cycle.The software routine fits the data to establishweight factors. To qualify the data as a valid shut-in period for buildup or falloff analysis, a series offilters is applied based on the length of the priorinjection or production period, the length of theshut-in period and the length of time necessary

> Cleaner, more coherent data. High-frequency electronic flow measurement (EFM) data (left) obtained from SCADA systems display considerable scatter,making them difficult to use. It is almost impossible to identify the linear portions of the data as well as determine subtle features that are critical for type-curve matching. The object of the radial flow plot (top left) is to determine the slope of the linear portion of the data. By eliminating outlying data pointsand averaging the data over 10-minute intervals (top right), the scatter is reduced, and the slope of the line is more obvious. Wellbore-storage effects,hidden in the raw data, can now be seen on the far right of the curve. Similar improvement in the log-log plot of the buildup and derivative curves (bottom)helps to make sense of the data. The noise and scatter seen in the original data (bottom left) would make curve fitting and identification of the linearportion difficult. With the aggregated and cleansed data (bottom right), the linear section can be determined and used for permeability estimation. Thesubtle changes in the shape of the right half of the derivative plot, which is used for determining reservoir boundary conditions, are discernable. Prior tocleansing and reduction, these data would be extremely difficult to interpret. (Adapted from Brown and Meikle, reference 9.)

0.1

0.01

0.001

0.0001

0.01 1 10

High-Frequency EFM Log-Log Plot

Equivalent time, h0.1

Δpr

essu

re/Δ

rate Buildup

Derivative

0.1

0.01

0.001

0.0001

0.01 1 10

Cleansed and 10-Minute Averaged Log-Log Plot

Equivalent time, h0.1

Δpr

essu

re/Δ

rate

Time function

123

122

12110 100

High-Frequency EFM Radial Flow Plot

Pseu

dopr

essu

re

Cleansed and 10-Minute Averaged Radial Flow Plot123

122

12110 100

Time function

Pseu

dopr

essu

re


for sufficient data to be acquired for the analysisto be valid. Additional controls wereimplemented to avoid limitations when exportingdata to an Excel worksheet.

At this point in the data-handling process, auser must intervene to select the plots and adjustthe scales as applicable. Apparent mechanicalskin damage factor, S, can be plotted as afunction of time or rate. If the mechanical skindamage factor, s, and non-Darcy flow coefficient,D, are constant with time, then the apparentmechanical skin damage factor, S, will be a linearfunction of rate and can be used to estimate sand D.13 If s or D changes with time, the data willnot be a linear function of rate. Abrupt changeswill cause clusters of data points, and gradualchanges will result in the data drifting away fromthe baseline model (above).

One study found that 60% of the reservoirsevaluated had wells in which non-Darcy flow wasidentified as a damage mechanism.14 If thedamage is assumed to be related to mechanical

skin damage factor alone, erroneous conclusionsmight lead to inappropriate or ineffectiveintervention. Mechanical skin damage is oftenimproved by pumping acid into the perforationsor by hydraulically fracturing the formation.These types of remediation would not effectivelytreat non-Darcy flow effects, and performingthem could be a waste of time and money.

A change in non-Darcy flow effects duringinjection was observed during the Columbia GasTransmission study period. The onset of damagewas isolated to a particular week in 2004. Severalwells in the field exhibited this increased non-Darcy flow effect, which was identified by anabrupt change of the slope on the apparent skindamage versus rate plot, while other wells in thefield did not experience this change. The analysissuggests that the cause of this perceived damageis related to an increase in turbulent flow duringinjection, rather than mechanical damage toindividual wells. Thus, no remediation was

warranted. Had the problem been detectedduring an annual test conducted on an individualwell, rather than by continuous monitoring of allthe wells in the field, it is possible that theresults could have led the operator to the wrongconclusion and needless expenditures.

Existing data sources and new data miningtechniques were used to perform the analyses,allowing the operator to determine the source ofperceived damage and make the proper decisionfor dealing with it. In this case study, the wells inquestion were able to remain in operation and noremediation was necessary.15

Practical Improvement Leads to Level IIA systemic approach to the processes ofinjection, storage and delivery for UGS facilitiesis likely to provide the greatest benefit tooperators. Individual-well analysis, reservoirmodeling, surface hardware and system ineffi -ciencies need to be fully evaluated, but it is not enough to focus on one or two aspects of

10 Oilfield Review

> Monitoring well performance. If mechanical skin damage factor, s, and non-Darcy flow coefficient, D, remain constant, apparent mechanical skin damagefactor, S, plotted versus flow rate should be a straight line (left, black). If s or D changes, the plot changes depending on the type of damage. From EFM data(top right), a step change in s occurs between the two withdrawal cycles (arrows). Plotting flow rate (bottom right) versus injection S (buildup) andwithdrawal S (falloff) reveals changes occurring during injection but not during withdrawal. Had s and D remained constant, the data would have fallenalong the reference line (red). In this example, the change in slope indicates an increase in D. (Adapted from Spivey et al, reference 11.)

Withdrawal Injection Withdrawal

0

Mec

hani

cal s

kin

dam

age

fact

or, s –1

–2

–3

–4

–5

Date4/1/041/16/0411/1/03 6/16/04 8/31/04 11/15/04 1/30/05 4/16/05

Shift in both slope andintercept: s and D changed

Shift in slope:D changedShift in intercept:s changed

Appa

rent

mec

hani

cal s

kin

dam

age

fact

or, S

Flow rate

FalloffBuildupBest fit

Appa

rent

mec

hani

cal s

kin

dam

age

fact

or, S

0–1,000–2,000–3,000–4,000–5,000 1,000 2,000 3,000 4,000 5,000

3

2

1

0

–1

–2

–3

–4

Flow rate, Mcf/d

S = –2.67D = 1.124 D/MMcf

61459schD4R1.qxp:Format Layout V7.2 6/25/08 6:23 PM Page 10

Spring 2008 11

the operation, such as EFM data, reservoircharacterization or surface hardware. Theoptimization process requires that all thecomponents be considered together to develop amodel that represents the total system.Optimization, modeling and surveillance are keycomponents in developing a Level-II intelligentUGS facility.

Most testing experts would agree that thefirst step in optimization is to characterizeindividual wells or completions. The conditionand quality of the reservoir in the vicinity of thewellbore and the level of damage affect the flowefficiency of the completion. Once these aredetermined, a predictive model can be developedto provide the expected performance undervarious operating conditions. Multirate welltesting is the standard method for fullcharacterization of well damage and flowperformance, and, with regard to developing arealistic model, it is indispensable.

Multirate pressure-transient testing quantifiesmechanical skin damage factor, s, and non-Darcyflow effects, D, and establishes a baseline forfuture comparison. Once the individual wellshave been tested, the next step is to characterizethe production properties of the system with amultirate test across the entire field. Usingoperational data, such as flow rates, tempera -tures and pressures from wellheads, treatmentfacilities and metering stations, an operator candesign an effective field-wide multirate test. If allthe wells in the field can flow simultaneously,this type of test provides a field-wide delivera -bility curve. The field-wide flow rate should behigh enough to identify the first bottleneck in the system.

Bottlenecks can be characterized as systeminefficiencies that affect overall performance atsome specific operating condition. Wellbore size,tubulars, wellhead equipment, gathering linesand treatment facilities impact the system andmay act as bottlenecks (right). Once thesebottlenecks are identified, economics determinewhether it is worthwhile to remove the cause ofthe bottleneck. For example, if a multirate testindicates that larger tubing will eliminate abottle neck but the wellbore size limits thetubulars that can be installed, there is little thatcan be done to fix the problem. Adding wells orreplacing existing wells might be the onlysolution, albeit an expensive one.

If a total-field flow test cannot be conducted,a deliverability curve can be constructed fromtests performed on individual wells. However,

surface-facility effects must then be measuredand included. Nonetheless, the total-fieldmethod will yield the most accurate results.

Finally, an inventory model for the storagefield can be developed that describes the total gasinventory as a function of reservoir pressure. Thisis accomplished by maintaining a constant flowrate from the storage field for a period of timesufficient to reach pseudosteady state, and usingthe pressure-decline rate observed duringpseudosteady-state flow to calculate the reser voirpore volume. In other words, this technique candetermine the size of the underground storage“tank” available for storage and delivery. Once thisreservoir pore volume is quantified, the operatorcan estimate the total gas contained in thereservoir for a given average reservoir pressure.

With existing technology, the averagereservoir pressure can be estimated from abuild up test any time there is a sufficiently longshut-in period during normal operations. Since

the field-inventory model quantifies therelationship between average reservoir pressureand total inventory, the operator can performmore frequent inventory audits than werepossible using data acquired only during routinespring and fall shut-in periods. Calculation ofaverage reservoir pressure from flowing data,perhaps available with currently developingtechnology, may allow real-time inventoryauditing in UGS fields rather than waiting forlong- or short-term buildup data. This isespecially important in cases where frequentcycling of gas is occurring and there is noroutine shut-in.

Once the pore volume has been establishedby collecting both individual well and field-widedata from extended tests, two independentestimates of pore volume can be derived andcompared. This information is integrated withthe deliverability-testing data to create the totalsystem model.

13. Mechanical skin damage factor, s, is a dimensionlessnumber calculated to determine the production efficiencyof a well by comparing actual conditions with theoreticalor ideal conditions. A positive skin value indicates thatsome damage or influences are impairing well produc- tivity. A negative skin value indicates enhancedproductivity, typically resulting from stimulation.The non-Darcy flow coefficient, D, is calculated fromfluid flow that deviates from Darcy's law. Darcy’s lawassumes laminar flow in the formation, and if the flow isturbulent rather than laminar, it is referred to as non-Darcy flow. Typically observed near high-rate gas wells,turbulent flow occurs when the flow converging to thewellbore reaches velocities exceeding the Reynoldsnumber for laminar or Darcy flow. Since most of the

> Identifying bottlenecks with multirate flow testing. NODAL production systemanalysis indicates that the tubing size is a bottleneck in this test. Tubing insidediameter (ID) is varied from the existing size of 3.092 in. (green) to 8.0 in. (blue).The analysis confirms that increasing the tubing ID from 3.092 in. to 5 in.(red) would provide a 54% increase in flow capacity, from 11 MMcf/d[311,000 m3/d] to 17 MMcf/d [481,000 m3/d]. Above 5-in. ID, only a nominal flowincrease would be added. (Adapted from Brown and Sawyer, reference 14.)

Pres

sure

, psi

500

400

300

200

100

05,000 10,000 15,000 20,0000 30,00025,000

Gas rate, Mcf/d

Inflow at sandface A B CD

E

Case 1 (A) - 3.092Case 2 (B) - 4.000Case 3 (C) - 5.000Case 4 (D) - 6.000Case 5 (E) - 8.000

Outflow, tubing ID, in.

turbulent flow in producing formations occurs near the wellbore, the effect of non-Darcy flow can be represented as a rate-dependent skin effect, D. Apparent mechanical skin damage factor, S, is similar tos but may be a result of mechanical damage or non-Darcy flow effects such as a rate-dependent restriction.

14. Brown KG and Sawyer WK: “Practical Methods toImprove Storage Operations—A Case Study,” paper SPE 57460, presented at the SPE Eastern Regional Conference and Exhibition, Charleston, West Virginia,October 20–22, 1999.

15. For more on recent sensor and data-handling develop-ments: Bouleau et al, reference 7.


Various programs, such as the PIPESIM andNODAL production system analysis software, canbe used to construct a nodal analysis model(below). Model creation begins with identifyingthe physical components of the system andintegrating the reservoir properties obtained

from well logs and well tests. Characteristics ofthe gathering systems, pro cessing equipmentand surface equipment are added to the model,which is then calibrated for system variablessuch as pipe roughness and gathering-linelengths. The model is adjusted to match the

pressure data obtained during the well testing,focusing on bottlenecks.

Trend analysis compares actual results fromongoing, routine operations with the model-basedresults established during the character izationphase. More advanced predictive tools, such asthe DECIDE! data mining based productionoptimization software, have been utilized toautomate the process of comparing model-derived data with operational data.

Starting from Scratch—or Not The potentially long lifetime of a UGS facility—the Zoar field, for example, has been in operationfor more than 90 years—may demand wellconstruction practices that differ fromnonstorage wells. Storage wells must be able totolerate high injection rates, high productionpressures and frequent cycling. Reuse of existingdownhole and surface equipment may bepossible, but more common is a mixture ofexisting wellbores and newly drilled wells. Theoriginal producing operation may, however,dictate well placement and facility location. Thetransition from production to storage shouldfocus on optimization and thoroughunderstanding of the reservoir.

Ideally, optimization begins during theidentification of prospective fields to be used forgas storage. The first step in choosing acandidate is to understand the reservoir.Characteristics for UGS prospects are goodporosity and permeability and an effectivetrapping mechanism. If the choice of reservoir isnot an option, such as an existing gas-storagefacility in need of upgrading or improving, newtechnologies can still be employed to enhancethe value of an UGS operation.

An example of optimizing an existing gas-storage field with intelligent well technologies isthe Falcon Gas Storage Company, Hill-Lake fieldoperation in Texas (left). Formerly an oilproducer, this field was discovered in the 1920sand by the 1950s had reached the end of itsproductive life. It was converted to a gas-storagefacility in the 1960s. When Falcon took over theoperations in 2001, there were 21 wells in thefield, 10 of which were active in the gas-storageoperation. No development had taken place sincethe 1950s, and 2.5 Bcf [71 million m3] of gascould not be accounted for by previous operators,attributed to “meter error.”

The original interpretation provided toFalcon was fairly simplistic. From limited wellcontrol and old electric logs obtained in thevicinity of the injection site, the structure was

12 Oilfield Review

> PIPESIM production system analysis model. The PIPESIM program can be used to create a visualmodel of the mechanical components of the facilities. Reservoir and production information can beintegrated into the model. Field-wide multirate tests are used to calibrate the model. In this example,the wells are producing. Reversing the direction of flow provides an injection model.

Compressorstation

NodeProducing well

Additionalgroup of wells

Additionalgroup of wells

> Falcon Gas Storage Company Hill-Lake underground gas-storageoperation. Located in Eastland County, Texas, the facility has the capacity todeliver 515 MMcf/d [15 million m3/d] and inject 300 MMcf/d [8.5 million m3/d].(Photograph courtesy of Falcon Gas Storage Company.)


Spring 2008 13

mapped as a fluvial delta (right). Between 2003and 2006, Falcon added 16 wells to the field, with well placement aided by FMI FullboreFormation MicroImager analysis. The imagesfrom the FMI data suggested a newinterpretation, that the reservoir was a braidedchannel sand, not a delta.

In 2006, Schlumberger geologists used Petrelseismic-to-simulation software to develop a basicgeological model (bottom right). The existingwells were incorporated into the model, and anadditional 17 wells, drilled by Falcon in 2006 and2007, were included in the analysis.

These new wells, drilled as step-outs from theoriginal injection site, followed trends indicatedby the interpretation of the FMI images. Theadditional wells led to some interestingdiscoveries, such as a previously unknown sandlobe to the southwest of the main injection wells.As an isolated sand, it should have been depletedby earlier production but, unexpectedly, thepressure was similar to that in the rest of theHill-Lake reservoir, proving that they were incommunication. Based on information derivedfrom the Petrel geological model, this structureis believed to have contained the 2.5 Bcf ofmissing gas.

Not only was unaccounted gas discovered, butas gas was injected into the field, the originallydepleted reservoir was recharged. New wellspenetrating down-dip sand sections encounteredoil left behind during the initial productionphase that could now be recovered because ofthe increased reservoir pressure.

A by-product of the recovery of stored gas isadditional oil production in the form of naturalgas liquids (NGLs). When gas is injected andrecovered, it is enriched by hydrocarbon liquidsthat were left behind after the original oilproduction ceased. These liquids are stripped outof the recycled gas using a cryogenic gasprocessing plant, and then recovered and sold,adding to the profitability of the UGS operation.The Petrel reservoir model identifies candidatelocations for future field development where thesand properties are most conducive to theproduction of liquids during gas withdrawal.

In 2007, because of the insight provided by thepreliminary model, Falcon initiated a detailedPetrel geological and reservoir model, incorpo r -ating a total of 72 wells. The ELANPlus advancedmultimineral log analysis program and FMIanalysis were used to interpret 29 of the wells.Core data from five wells provided calibration forthe model. With advanced petrophysical analysis,the initial simplistic interpretation of a deltaicdeposition evolved into a more realistic model ofthe reservoir.

> Changing interpretations. The Hill-Lake field was originally mapped as a fluvial delta (left). Twostructural highs were identified on the isopach map (top left). New well locations were drilledaccording to the original structural map, which was based on wells drilled before 1960. An updatedinterpretation, using Petrel seismic-to-simulation software (right), included 17 new wells and identifieda previously unknown southern lobe (top right). The original sand body’s geometry could be moreaccurately visualized, and the structure was characterized as a braided channel sand (bottom right).Because the FMI interpretation provided flow direction, well placement was improved. The southernlobe contained 2.5 Bcf of stored natural gas that the original operators assumed had been lostbecause of metering errors.

Fluvial Formation

Braided Channel Sand

Wells productivefrom Hill-Lake sandappeared connected

Mitcham Est 1

HLGSU 11

AJ August 1A

Storm 1

Mitcham 1A

HLGSU 1

Mitcham 3 Mitcham 1F

August 3

HLGSU 3HLGSU 2

HLGSU 4Mitcham Est 3

HLGSU 7

HLGSU 6HLGSU 5

HLGSU 12

HLGSU 9HLGSU 8

Williams 1

Mitcham 2

50 40 30 20 10 0

40302010

0

Mitcham 2

Williams 1

August 3

Mitcham Est 3

HLGSU 1

Main lobe

Sand Facies Thickness Maps

Southern lobe(connected)

40302010

0

10

2030

40

> Modeling a reservoir. Geologists used Petrel software to model the Hill-Lake reservoir. Once the geological model is fully developed, the reservoir properties can be used to developsimulations for volumetrics and field performance.

Data inventory andloading (logs, markers,

surfaces )

Petrophysics andborehole geology

(ELANPlus, BorViewsoftware)

Interwellcorrelation

Rock typeidentification

(facies properties)

Framework modeling(faults, surfaces, grids,

zonation layering)

Property modeling(facies, net to gross, porosity,

permeability, saturation)

PetrelGeological

Modeling

PetrelReservoirModeling

Reservoirsimulation model

(history and forecasting)

Volumetrics

Softwaremodules

Fluid PVT ECLIPSEsoftware

Petrel Model Construction Workflow


As additional wells were drilled, greaterinsight into the reservoir geometry was gained, and the depositional environment wasrecharacterized as an ancient riverbed withanastomosed channel sands.16 Optimal wellplacement depends on understanding the reser -voir structure. Visualization of the subsurfacetopography provided by the Petrel model wascrucial in identifying the correct depositionalenvironment and ruling out the two previousinterpretations (below). Thorough under -standing of the reservoir has added new storagecapacity to the field, defined new explorationareas to recover oil left behind after the initialproduction ceased, and helped optimize futurefield development. Production and capacityimprovements in the USA, such as those foundwith the Hill-Lake operation, help explain theUGS capacity increase, despite a decrease in thenumber of sites.

Falcon’s use of new applications and tech -nology is not limited to subsurface modeling andoptimization. Because this facility had been inoperation since the 1960s, production facilitiesneeded an upgrade. A SCADA system was

installed, providing instantaneous informationabout temperature, pressure and flow rates. Theflow of gas can be managed from the wellhead, atindividual compressor sites or at the central fieldcontrol room. Although the SCADA system is notused to remotely control the facility at present, ithas the capability to do so.

Falcon’s Hill-Lake facility is now a state-of-the-art, multicycle UGS operation with thecapability of being used in a variety of ways,including storage, high-rate delivery, peak-shaving, “park and loan” and market trading.17

The maximum capacity is now 15 Bcf[425 million m3], representing 11 Bcf[311 million m3] of working gas and 4 Bcf[113 million m3] of cushion gas. The field candeliver 515 MMcf/d [15 million m3/d] and inject300 MMcf/d [8.5 million m3/d]. Injecting insummer and supplying in winter have beenreplaced by a flexible operation capable of on-demand delivery and storage as required bycustomers, while recovering oil and NGLs thatwere left behind during initial production.

Falcon’s success with Hill-Lake resulted inthe recent retrofitting of another gas-storage

field in north Texas, its Worsham-Steed facility.Utilizing an abandoned oil and gas field originallyconverted to gas storage by another operator, thisretrofit is a multicycle UGS operation employingsimilar intelligent well technologies. This fieldproduces oil and NGLs along with providing24 Bcf [680 million m3] of working-gas capacity.

Level III—Automated Reservoir SurveillanceWhether UGS facilities are buffering demandcycles or acting as gas repositories, the ability toautomate the process is an attractive reason forimplementing intelligent well technologies. Anoperator in Europe, working with Schlumbergerreservoir geologists and engineers, designed andimplemented an automated reservoir surveil -lance operation using an integrated platformbuilt around DECIDE! software. An operator canoptimize and perform predictive modeling forhighly complex systems using the artificialintelligence and software simulation of the PC-based DECIDE! software (next page, top).18

This software provides a way to bring togetherpeople, technology, processes and information ina secure, global system—reducing cost, loweringrisk and enhancing operations. The DECIDE!program has two major components—a data huband an engineer’s desktop. Responsibility forretrieving, storing and distributing data, as wellas automation of tasks, lies primarily with thedata hub. The engineer’s desktop uses state-of-the-art data mining techniques to performanalytical petroleum-engineering calculations,giving the operator a powerful tool for managingthe asset.

RWE Transgas Net, an independent naturalgas operator in the Czech Republic, has installedDECIDE! software to manage and optimize all ofits depleted reservoir and aquifer gas-storagefacilities. Implementation began in 2004, andwas completed in 2007 (next page, bottom).

14 Oilfield Review

16. Anastomosed channel sands are found in multichannelrivers that have relatively low gradients, deep and narrow channels, and stable banks characterized ashaving been deposited in slow-moving rivers. Braidedchannel sands are found in high-energy environments,characterized by excessive deposition of sand bars orgravel bars or both, so that water flows in many branch-ing and reuniting channels.

17. Park and loan refers to storing gas for later use (parking)and taking gas (loaning) to avoid purchasing gas at highspot-market prices. Designed as a balancing service,customers save money by using the service at timeswhen they are out of balance on the pipeline. Customersalso can take advantage of short-term price swings.

18. For more details on applying ECLIPSE softwareand DECIDE! software: Barber A, Shippen ME, Barua S, Cruz Velázquez J, Garrido Hernández AM,Klumpen HE, Moitra SK, Morales FL, Raphael S, Sauvé B,Sagli JR and Weber M: “Optimizing Production fromReservoir to Process Plant,” Oilfield Review 19, no. 4(Winter 2007/2008): 18–29.

> The final answer. A clearer picture developed as more wells were added to the field. With 72 wells,and FMI data from 29 of those wells, the final Petrel geological model was created. The structure wascharacterized as an anastomosed channel sand (bottom left). Compared with braided channel sands,anastomosed channel sands are deposited from lower energy water flow. The more tortuous path ofthe riverbed required future wells to follow a more curved path than would have been indicated hadthe reservoir been either a delta or a braided channel sand. The FMI tool provided the direction to drillstep-outs and develop the field. Knowledge of the reservoir also indicated the best areas to drill torecover oil left behind by previous production. New structure was also identified, locating untappedreservoir potential and adding storage capacity for the facility.

Anastomosed Channel Sand

Computer-postulated lobes(potential future storage capacity)

Computer-postulated lobes(potential future storage capacity)

Main lobe(original unit)

Cooper A-3 lobe(separate lobe)

Northeast lobe(limited connectivity)

30

504030

2010

0

010

20

2010

0

302010

0

HLGSU 1

Mitcham Est 3

August 3

Mitcham 2

Williams 1

Southern lobe(connected)

5040302010 0

61459schD4R1.qxp:Format Layout V7.2 6/25/08 6:24 PM Page 14

Spring 2008 15

RWE Transgas Net, working with Schlumbergerengineers, began the process of implementingthe DECIDE! program by first developing anintegration platform. A SCADA system wasinstalled to provide continuous high-frequencymeasurements (on the order of seconds), whichare grouped into 15-minute increments and

streamed in real time from individual wells,gathering systems and facilities. At this Level-Istep in the process, the software system checksthat a connection to a datastream has beenestablished and generates a notifi cation if thereis a failure. When a valid connection isconfirmed, the high-frequency data are

imported, filtered, quality checked andaggregated over longer time intervals to reducethe size of the dataset. The software filterssensor errors and transmission errors prior todata aggregation and generates statisticalreports to allow the engineer to evaluate thereliability of the information. Artificial intelli -

> DECIDE! program workflow. SCADA data are streamed into the data buffer where they are quality checked, cleansed and reduced using a neuralnetwork (NN) proxy model. Data are fed to various software modules for automated surveillance, report generation and preparation. Proxy models processthe information and use trend analysis and simulation-matching to look for optimization opportunities and to detect developing system problems. Reportsare available in almost real time. History-matching is available to determine the ongoing health of the operation. The majority of these processes arecarried on in the background with little operator intervention required.

Engineer’s desktopSCADA systemShort-term

control module

Surveillance and reporting modules

Manuallycaptured data

Workflow design and automation

Network modeling Simulation

NN proxy

Material balance proxy

Data buffer Data hub

Proxy models

Database

Seconds Minutes to hours One hour to one day One month to one year

Legacy data

> Level-III intelligence. Level III combines all the components of intelligent well operations together. By integrating processes such as trend analysis,modeling and simulations, the UGS facility can be optimally managed with high levels of automation.

Data hub

Databasecomponents

Core of Expert System

• Monitoring and data QC• Event detection and performance monitoring• Automated data transfer and model execution

Material balance Simulation Function of influence Proxy model Well simulator Network simulator

Operational workflow

Design workflow

Model Integration and Control


gence has been developed to automate theseroutine tasks as well as increase the speed ofdelivery. With the newly acquired data, keyperformance indicators are available to evaluatethe ongoing operation.

At Level II, the cleansed data are fed tosoftware modules to validate proper systemperformance. The DECIDE! program can integrateexternal applications that allow data exchange,including ECLIPSE reservoir simulation software,PIPESIM production system analysis software andvarious modules available in the DECIDE!software. The process automatically conductshistory-matching for trend analysis, providesindividual well status and determines productionand capacity constraints. Current and futuredelivery requests for injection and withdrawal—inputs into a dispatcher’s module—are passed to a DECIDE! software module. This module then provides all the necessary calculations and predictions to ensure that the reservoir has sufficient capacity to meet the dis-patcher’s requests.19

The advanced programs used for the reservoir-surveillance models require computer-intensivecalculations. Running optimization iterationscannot provide satisfactory results in therequired time frame using the large volume ofdata, even after the data have been aggregatedand cleansed. Proxy models, although not asaccurate, are substituted for full-scale

simulations and can provide results in seconds or minutes.20

Proxy models, in the form of trained neuralnetworks (NNs) optimized to require a reducednumber of input parameters, use artificialintelligence to mimic large-scale simulators. TheNN learns to behave like the simulator and, oncetrained, it can perform a set of calculations in afraction of a second for a given set of inputparameters. The NN drastically reduces thecomputation time necessary, allowing real-timehistory-matching with the optimized originaloutputs of full-scale numerical models.

An example of the use of a NN is a productionforecast and optimization simulation. If smallchanges of the input parameters are involved,such as tubing-head pressure, a forecast can becalculated immediately, rather than waiting for atime-intensive full simulation to be performed.Multiple iterations can also be run quickly todetermine the best course of action. Additionally,NNs are used to evaluate uncertainties in theinput data provided in the Level-I data-acquisition phase. This speeds up thequality-control and data-cleansing functionsprior to inputting data into the proxy models.

An automated surveillance system comparescalculated results with measured results. Ifinformation from proxy models indicates that awell or surface component has failed to performas expected, an event alarm is triggered and

reported to the operator by way of the engineer’sdesktop. For the parameters that are set totrigger alarms, a deviation range can beestablished and adjusted as required. Once analarm has been triggered, the reservoir engineercan react in a timely manner to investigate thesource of the problem (above).

The level beyond monitoring and surveillanceis Level-III intelligence—an example of thedigital oil field. Although oil companies andservice providers have found it difficult toprovide a single definition for this term, thedigital oil field essentially provides a high level ofautomation, simulation modeling, decision-making tools (the faster, the better) and anintegrated approach that does not lose sight ofthe small details (or at least has a system tomonitor them). Schlumberger refers to this levelof field operation as the BlueField intelligentasset integration service.21

At the BlueField level, data are acquired andprepared for processing, and integrated-simulation models are run from variousperformance modules. System checks are carriedout at the highest level, and reports concerningthe health of the complete operation, usingdeterministic models, are generated anddelivered to the DECIDE! engineer’s desktop(next page). The service provides the ability tooversee scheduled automated tasks or thosetriggered by event alarms.

16 Oilfield Review

> Neural networks (NN) and problem identification. Multiple parameters are required to calculate the bottomhole pressure(BHP). Neural networks, functioning as proxy models, are trained with data from operations and can reduce the total iterationsrequired to produce results in a fraction of the time required by full-scale modeling programs. Measured BHP data (red) duringinjection and withdrawal are compared with NN-derived BHP data (blue). The two sets of data compare very favorably untilSeptember 2000 when the calculated BHP during injection increased and remained higher for the next four months because ofreservoir damage. The higher backpressure indicated well intervention would be required to maintain deliverability.

Measured BHP deviatesfrom calculated BHP

Pres

sure

, MPa

85

80

75

70

65

55

50

Jan 1999 April July Oct Jan 2000 April July Oct Jan 2001Time

Calculated BHPMeasured BHP

NN Architecture for Well Model

Reservoir pressure

Aquifer influx

Gas injection

Gas production

19. Onderka V, Dressler M, Severýn O, Giovannoli M andZangl G: “Expert System of UGS—An Efficient Tool forOnline Performance Management and Optimization,”presented at the 23rd World Gas Conference, Amsterdam, June 5–9, 2006.

20. Zangl G, Giovannoli M and Stundner M: “Application ofArtificial Intelligence in Gas Storage Management,”paper SPE 100133, presented at the SPE Europec/EAGE

23. “Supply on Demand,” http://www.falcongasstorage.com/fw/main/MoBay_Storage_Hub_LLC-28.html (accessedJanuary 15, 2008).

24. National Petroleum Council (ed): Hard Truths: Facing theHard Truths about Energy. Washington, DC: NationalPetroleum Council (2007): 36.

Annual Conference and Exhibition, Vienna, Austria, June 12–15, 2006.

21. Bouleau et al, reference 7.22. Foh SE: “The Use of Inert Gas as Cushion Gas in

Underground Storage: Practical and Economic Issues,”presented at the Gas Supply Planning and Management:1991 and Beyond Conference, Lake Buena Vista, Florida,USA, February 25–27, 1991.


Spring 2008 17

In the system implemented by RWE TransgasNet, automated tasks have the followingstructure: first run scheduled tasks or triggerautomation tasks, then run predictive data-mining proxy models and apply rules. Triggeringevents are either discrepancies from expectedtrends or violations of predefined constraints.The actions triggered by the alarm includesystem notification, retrieval and execution ofsurveillance software, exchange of data withthird-party software, initiation of subordinatetasks and generation of e-mail or text messagesto alert the operator of an error condition.

Along with the alarms, the softwareautomatically provides key performanceindicators to the operator at the engineer’sdesktop. It formats the data for visualization andprovides forecasts based on current performanceof the field. Reservoir performance modulesidentify bottlenecks, like facility constraints, andreport on optimization opportunities along withrecommended courses of action. With dramati -cally reduced analysis cycle time, the engineercan react almost instantaneously. Automateddata flow and transparently updated modelsallow the engineer to focus on systemoptimization and problem elimination. Proactive,intelligent reservoir management—a BlueFieldapplication—becomes a reality.

Into the FutureUnderground gas storage in depleted reservoirshas proved to be well-suited for many of theintelligent well and intelligent field technologiesthat are being developed for traditional oil andgas production. The UGS industry has achievedgreat success in the adoption of thesetechniques. The lessons learned by UGSoperators are being applied with greaterconfidence by the oil and gas production side ofthe business because these new technologieshave demonstrated their ability to providereduced costs and increased efficiencies.

Because UGS fields have long life expec tancies,they afford a long-term outlook for payback.Compared with conventional hydrocarbon fields,gas-storage fields do not experience the samedecline in their asset’s value as the reservoirdepletes because gas-storage fields can berepeatedly recharged. Retrofitting older facilitieswith modern intelligent-field equipment makesfinancial sense, increasing the value of the existing asset.

Maximizing the asset, above and below theground, leads to innovative approaches like thosediscussed in this article, but there are still moretechnologies and techniques to apply. For

instance, cushion gas can be the most expensivecomponent in a UGS facility and realistically willbe returned to the operator only when the field isdecommissioned. As an example, UGS in adepleted field with 20 Bcf [566 million m3] oftotal capacity would require 30 to 50% of the gasto remain in place as cushion gas. Borrowingfrom the bank analogy, that equates, at the highend, to 10 Bcf [283 million m3] left in an interest-free checking account. At 2008 price levels, thatcomes to more than US $80,000,000.

Even if the reservoir is operated in the mostefficient pressure and flow range, some cushiongas must be left in the ground to enable high-ratedelivery. Reservoir engineers have tested thefeasibility of injecting inert gas into the reservoirto function as cushion gas. This approach isespecially practical considering current naturalgas prices.22 The technique does, however, requiredetailed understanding of the reservoir storageproperties and flow characteristics, the conse -quences of mixing different gases and thelong-term effects of the inert gas on the reservoir.This is yet another example of UGS operatorsapplying novel reservoir management techniques.

As operators develop UGS fields and attemptinnovative approaches, the emphasis on reser -voir characterization, process optimization andautomated operation bring greater flexibility andopportunities to UGS projects. As an example,Falcon Gas Storage Company is applying much ofits experience with intelligent field technology in

UGS to the first offshore UGS facility in NorthAmerica.23 The reservoir has been characterizedand modeled using Petrel software. The surfaceequipment has been designed, and operationsare expected to commence during 2009. This high-deliverability, multicycle facility isdesigned to have a working volume of 50 Bcf [1.4 billion m3], with injection and withdrawalcapabilities of 1 Bcf [28 million m3] per day.

By 2030, global demand for natural gas isprojected to range from 356 to 581 Bcf [10 billionto 16.4 billion m3] per day, up from 243 Bcf/d [6.9 billion m3/d] in 2000.24 The Middle East hasby far the largest natural gas reserves with anestimated 2,566 Tcf [72.7 trillion m3]—or 41% ofthe world’s total. Russia, second in provenreserves, has extensive pipelines into Europe andhas proposed pipeline construction to China andother countries. As demand for natural gasgrows, new methods will be developed totransport, store and deliver it. Because thesupply is generally far removed from most users,UGS facilities are a major component inproviding a stable, secure source of natural gasfor industrial and residential consumption.

As the character of UGS evolves from peak-shaving to flexible applications, intelligent fieldtechnologies are assisting operators in the questfor greater efficiency, lower costs and innovativemethods of doing business. As a result, the digitaloil field has become a reality in the undergroundgas-storage industry. —TS

> Data in, decision out. SCADA data arrive in 15-minute intervals and are cleansed and qualitychecked. Data from simulation software are compared with model predictions. The engineer at theDECIDE! desktop receives results, forecasts and production information. This information is receivedautomatically or can be generated upon request.

SCADA system

System acquires data every 15 minutes

Dataacquisition

Datacleansing

Systemhealth check

Datastorage

010101010101010101010101010101010101010101

Run simulationWrite ECLIPSEschedule file

Datastorage

Simulationresults

Model validation loop

Run forecastRun optimization

Current day plusprevious 100 days To dispatcher

Current time plusprevious 24 hoursDispatcher request

Optimization and prediction loop–DECIDE! desktop

Generateproduction

curvesOptimization

results


intelligent well technology in underground gas...

Documents