By WAYNE KIRSNER, PE,Kirsner, Pullin & Associates,Atlanta, Ga.

I’ve got some bad news—the20-odd-year experiment withprimary-secondary design of

chilled water plants hasn’tpanned out. If you’ve designed alarge distributed chilled watersystem and monitored the opera-tion of the central plant, you al-ready know about the problems:the ∆T of the chilled water(CHW) returning to the campusplant is below the design valuefor which the chillers and pumpswere selected—in fact, it’s waybelow; the secondary CHW flowdoesn’t vary a hoot; and the ex-pensive variable-speed drive(VSD) purchased to vary the flowof the secondary pumps wasgreat for test and balance buthasn’t done much since (besides

heating up the central plantbuilding). The low ∆T at theplant causes the operators to runextra pumps and chillers to meetthe load, which, in addition to re-ducing the plant’s cooling outputcapacity, wastes energy. The sys-tem may be keeping the campuscool, but you know it’s inefficientand idling a lot of chiller capac-ity.

The problem described abovehas come to be known as “low ∆Tcentral plant syndrome.” To myknowledge, every large chilledwater plant serving distributedloads is afflicted with it to somedegree. The article “Trouble-shooting Chilled Water Problems

at the NASA Johnson Space Cen-ter” (HPAC, February 1995)1 de-scribes a typical situation. A cen-tral plant originally designed fora 16 F ∆T between the chilled wa-ter return (CHR) and chilled wa-ter supply (CHS) could only de-velop an 8 F ∆T because of lowCHR temperature from the cam-pus. This meant not only thattwice as much CHW as originallyintended had to be pumpedaround the 5-mile campus pipingloop but also that the seven 2000-ton chillers in the central plantcouldn’t be loaded much beyondhalf their capacity. Thus, opera-tors were usually forced to runtwice as many chillers to meetthe campus load, and the fric-tional loss in the mains due to theexcessive CHW flow made it

CHW PLANT DESIGN

November 1996 HPAC Heating/Piping/AirConditioning 73

The Demise of the Primary-Secondary Pumping Paradigm forChilled Water Plant Design

Accepting that low ∆Tchilled water plantsyndrome exists in

almost all bigdistributed chilledwater systems and

recognizing the need toseek design solutionsthat can cope with or

prevent it

VSD

Chiller Chiller

Constant-flowpumps

Crossoverdecoupler

1 Archetypal primary-secondaryCHW plant design.

1Wayne Kirsner authored the Febru-ary 1995 article cited above as well asthe article “What Caused the SteamSystem Accident that Killed JackSmith?,” HPAC, July 1995.

tough to deliver sufficient CHWto hydraulically distant build-ings.

The causes of low ∆T syndromeare not mysterious, but they areoften pervasive and thus can behard to remedy. Low ∆T can becaused by dirty cooling coils,throttling valves with insuffi-cient shutoff capability, resetCHS temperature, poorly con-trolled blending stations, and ofcourse, CHW bypassing out inthe system. But most often, lowsystem ∆T is the result of faultycontrols and improperly adjustedset points. This article, however,is not about the causes of centralplant syndrome. It’s about ac-cepting that the problem exists invirtually every big distributedchilled water system and thenrecognizing the need to seek de-sign solutions that can cope withit, if not prevent it.

So why can’t a standard pri-mary-secondary chilled water de-sign cope with low CHW ∆T?

Problem #1The primary-secondary control

scheme is “blinded” by low ∆Tcentral plant syndrome. Fig. 1 de-picts what I would describe asthe archetypal primary-sec-ondary chilled water schematicconfiguration. The primary fea-ture of the configuration is thedecoupled primary and sec-ondary loops, which allow con-stant flow through the chillerswhile permitting varying flow inthe system to save pumping en-ergy. Chillers are staged on andoff based on CHW flow throughthe crossover bridge (althoughthe sensor may be elsewhere).The sole indicator of system load,upon which control of the chillersand pumps depends, is chilledwater flow.

In a plant with low ∆T syn-drome, CHW flow is no longermuch of an indicator of load. Theamplitude of flow variation is justa fraction of the amplitude of loadvariation. Fundamentally then, aprimary-secondary control

scheme that depends on systemflow to gauge system load is virtu-ally blind to load variation.

Problem #2The primary loop is constant

flow. Constant flow throughchillers is a highly desirable fea-ture of primary-secondary chilledwater plant design, and mostchiller manufacturers still preferand recommend it. I’ve been con-vinced, however, that most mod-ern chiller controls no longer re-quire constant flow to keep thechillers out of trouble. Let me ex-plain.

When chiller vanes were con-trolled by conventional pneu-matic proportional controls, re-sponse time to changes in loadwas necessarily slow and gradualto prevent overshoot and huntingas the chiller controls tried toachieve leaving CHS set point.Hence, chiller capacity controlswould lag behind a sudden loadchange. If the change was a dropin load, the chiller would overcoolthe leaving CHW, dropping it be-low set point until capacity con-trol vanes could react to reducechiller refrigerating capacity. Ifthe drop in load was sharpenough, the chiller’s low evapora-tor temperature safety wouldknock the chiller off line, requir-ing a manual reset to restart thechiller. This is a situation to beavoided.

Now consider the response of achilled water plant designed forconstant flow versus one designedfor variable flow in the event thatload across a fully loaded chillersuddenly dropped in half. (This is

a severe upset, but it’s not far-fetched at all. Starting a secondchiller in a two-chiller plant,where identical chillers operate inparallel, typically results in theload to the active chiller beinghalved.2) In a constant-flow pri-mary loop designed to chill, say,55 F CHR to 45 F CHS, a 50 per-cent drop in load would manifestitself in CHR temperature risingto 50 F. (This might occur becauseapproximately half the primaryflow of 45 F CHS is recirculatingthrough the crossover bridge tomix with the 55 F CHR from thesystem.) The 50 F CHR enteringthe formerly fully loaded activechiller would initially be sub-jected to the full cooling capacityof the chiller until its controlscould respond to decrease capac-ity. The chiller would thus tend todrive the entering 50 F CHWdown toward 40 F.

Compare this upset condition toa variable-flow configuration.Starting a second equal CHWpump could cut CHW flow throughthe active chiller roughly in half.3

The active chiller would initiallycontinue to try to apply its full out-put capacity to half the mass flow,thereby doubling the ∆T of CHWpassing through it—i.e., it wouldtend to drive 55 F CHR down to 35F. This is pretty close to freezing. Ifthe design ∆T was larger, the CHWwould be driven down below freez-ing. In either case, a simple lowevaporator temperature sensorwould likely cause the chiller totrip off line to protect it from freez-ing. The constant-flow chiller, incomparison, whose leaving CHStemperature dips only half as far,would probably remain on line. Forthis reason alone, one can easilyunderstand why chiller manufac-

Chilled water plant design

74 HPAC Heating/Piping/AirConditioning Novemberr 1996

2For example, in a variable-flowplant, flow through the active chillerwill be cut roughly in two as the sec-ond chiller’s pump instantly usurpshalf the flow. In a primary-secondaryplant, approximately half the totalprimary CHW flow recirculatesthrough the crossover bridge and, oncethe second chiller’s compressor has be-gun outputting CHW, mixes with sys-tem secondary CHR to halve its ∆T.

3Assuming immediate system controlvalve response and the absence of cen-tral plant syndrome. If control valveresponse were slow or the system wereafflicted with central plant syndrome,flow would not suddenly fall to half,and thus the upset condition would befar less traumatic.

turers would prefer constant flowthrough chillers.

So what’s changed to invalidatethis argument? The low evapora-tor temperature control is more so-phisticated, for one. It’s no longersimply a low-temperature safetycutout. The Trane Company’s mi-croprocessor-based control, for ex-ample, integrates (i.e., sums) thenumber of degree-seconds (deg-sec) below the low evaporator tem-perature set point. Don Epple-heimer of Trane tells me that ifthis sum remains below 50 deg-sec,the control logic will not initiate asafety shutdown. This means:

◆ Evaporator temperature maydrop below freezing momentarily.

◆ The chiller’s capacity controlsare allowed time to catch up withthe load change.

In fact, the sophistication of thecontrol logic is such that Tranefeels confident in setting its lowevaporator temperature set pointas low as 30 F.

The upshot of this improve-ment, and to a lesser extent thecapacity control improvements, isthat chillers can survive a severeupset condition in CHW flowwithout tripping off line. In fact,and this is really the proof of thepudding, Mr. Eppleheimer saysthat Trane routinely tests itschillers to insure they can with-stand a 50 percent drop in CHWflow without tripping the lowevaporator temperature safetycutout. (Other manufacturershave different control strategiesfor handling upsets in appliedload. York’s chillers, for example,can accept an input that delayspowering down of the chiller com-pressor upon a large drop in leav-ing CHS temperature as long asthe temperature does not fall be-low 36 F. Carrier’s low evaporatortemperature control overrides thechiller capacity controller to closecompressor vanes should evapora-tor temperature approach 33 F. At33 F, the safety shuts down themachine.)

There is another benefit to con-stant flow through chillers. The

possibility of laminar flowthrough the evaporator due to lowCHW flow is eliminated. This con-dition can easily be avoided, how-ever, in a variable primary flowsystem. Burt Rishel of Systeconsuggests the best way to do this isthe old-fashioned way—with a by-pass from CHS to CHR opened viaa signal from a flow meter or dif-ferential pressure controlleracross the chiller. This mightseem to replicate the expense of aprimary-secondary crossoverbridge, but the difference is thatthe recirculation pipe is sized tohandle no more than about half achiller’s design flow, and its func-tion is less likely to confuse theoperator.

So the rationale for avoidingvariable flow through chillers is,in my opinion, no longer com-

pelling. But even if constant flowthrough chiller evaporators is nolonger essential for stable chilleroperation, what’s wrong with it?

A constant-flow primary CHWsystem with one nonvaryingpump per chiller cannot respondeffectively to low ∆T syndrome. IfCHR temperature returning fromthe system is below design andcannot be raised, a central plantoperator’s only option in respond-ing to a call for more cooling ca-pacity is to energize more pumpsand more chillers.4 What would bepreferable, of course, would be toincrease the CHW flow through

individual chillers to load themmore fully. But with the primary-secondary configuration shown inFig. 1, this is not possible. Pri-mary-secondary systems can beretrofitted, of course, in responseto low ∆T. New pumping capacitycan be added and flow throughchillers can be increased up to themanufacturer’s recommendedmaximum rate. Perhaps evapora-tors can even be converted fromthree-pass to two-pass, reducingpressure drop through thechillers. But obviously, the con-straints of the existing equipmentlimit the flexibility to cope withthe need to force more waterthrough the chillers. Further-more, retrofitting and addingequipment to accommodate alower ∆T within the context of theexisting constant-flow design re-

sults in permanently locking inthe higher CHW flow rate. Aretrofit of this kind effectivelythrows in the towel on the quest tofind and fix the root causes of low∆T out in the system.

So if not constant flow throughthe chillers, then what? A vari-able-flow CHW pumping schemecan respond to low system ∆T ifthe pumps are selected for excesscapacity. In fact, in my view, de-signers need to consider the de-sign ∆T for which they select the

November 1996 HPAC Heating/Piping/AirConditioning 75

Chiller Chiller

VSD VSD

Flow meter

DP

By

2 Typical variable-flow CHW plant design.

4Assuming pumps are not ganged in acommon header, of course.

5The maximum tube velocity recom-mended in the ASHRAE EquipmentHandbook is 7 fps; most manufactur-ers recommend 11 fps.

continued on page 77

November 1996 HPAC Heating/Piping/AirConditioning 77

chillers as a target value andthen provide for the eventual-ity that extra flow beyond thedesign target may be re-quired. In practice, thismeans:

◆ Selecting chiller evapo-rator tubes for tube velocitiesnot more than about 5.5 fpsat design so that flow can beincreased up to twofold if nec-essary.5

◆ Selecting pumps to over-pump the ch i l l e r s . Thebest scheme is to bank thepumps and provide themwith VSDs.6

With variable flow pumpingthrough the chillers, the crossoverbridge and the secondary pumpscan be dispensed with, so a typicalschematic layout for a simplebuilding can look like Fig. 2.Chillers are staged based on leav-ing CHS temperature. When achiller can’t hold leaving CHStemperature set point, a secondchiller is energized. Pump speedis controlled by a differential pres-sure sensor situated across thehydronically farthest coil. A flowmeter and smart controller open abypass valve should flow throughthe chillers fall below the manu-facturer’s recommended mini-mum.

The advantages of this systemare:

◆ It automatically responds tolow ∆T by increasing flow throughchillers.

◆ There ’ s on ly one se t o fpumps.

◆ Minimum chilled water flowis pumped.

◆ The sys tem i s s imp le r ;there’s no decoupling bridge.

Problem #3Secondary pumping is not the

most efficient pumping distribu-tion scheme. In Fig. 2, a single setof pumps handles the job of boththe primary and secondarypumps. But in a big system, asingle set of pumps is not alwaysdesirable. If the pressure neededto pump an entire campus islarge, it’s advantageous to placea second set of pumps, and per-haps even a third set, out in thesystem to avoid imposing highpressure on the equipment closeto the pumps’ discharge. Sec-ondary pumping, as shown inFig. 1, can achieve this objective,but it ’s not the most efficientpumping scheme. That’s becausethe same head is imparted to allCHW passing through the sec-ondary pumps, whether it’s mak-ing the short trip through theclosest building or the longesttrip through the hydraulicallymost distant building. The extrahead imparted to CHW passingthrough the closer buildingsmust be wasted across balancingvalves and/or throttling valves atthose buildings. Only the smallfraction of the total CHW flow go-ing to the most distant buildingis produced without wasted en-ergy.

A better way to pump distantloads is via distributed pumping,as illustrated in Fig. 3. BurtRishel of Systecon gives credit toWilber Shuster of Cincinnati forfirst proposing this pumping

scheme. Distributed buildingpumps assume the function ofthe secondary pumps. Eachpump is sized to deliver its build-ing’s flow at just the head neededto pump the building hydronicloads and draw the CHWthrough the mains from the cen-tral plant. There are no decou-pling loops at the buildings, so noCHW is bypassed. No balancingvalves are needed to eat up ex-cess head since there is none.Pump speeds are controlled byVSDs receiving signals from dif-ferential pressure switches at theend of the loop in each building.Pumping horsepower savingequals the sum:7

The primary pumps are VSD-controlled, as before, and can op-erate in series with the dis-tributed pumps or be decoupled asshown in Fig. 3. If decoupled, theVSDs would be controlled tomaintain slightly positive flowfrom CHS to CHR in the crossoverbridge and not let flow throughany chiller go below its minimumrecommended value. Chillers are

Chilled water plant design

VSD VSD VSD VSD

VSD VSD

DP DP DP DP

Chilled water plantBuilding 1 Building 2 Building 3 Building 4

CHR

CHS

3 Distributed campus CHW pumping.

6Of course, oversizing the pumps andbalancing them down with a throt-tling valve is not an option unless youroutinely wear your shoes on thewrong feet and when you tighten yourbelt, cut off your windpipe.

CHW flow

building building

ii N

N i

p

H H

×

−×

= −∑ 1 1

3960

,

( )( )

/η

7Assuming equal pump efficiencies forbuilding pumps and hypothetical sec-ondary pumps.

continued from page 75

crossover bridges at the buildings,but if it does become a problem,the pumps and chillers can effec-tively deal with it.

◆ It reduces head pressure im-posed on equipment.

◆ It’s simple and, more impor-tantly, looks simple to the opera-

tors who run it.The only unusual aspect of dis-

tributed pumping is that it re-verses the typical pressure gradi-ent in the system. The CHS mainis negative with respect to thepressure in the CHR main. Thus,every load must be pumped.

In conclusion . . .The traditional arguments for

desiring constant flow throughchiller evaporators no longercarry much weight; most modernmicroprocessor-based chiller con-trols can effectively deal with up-sets due to variable flow. More-over, constant-flow primarydesigns cannot respond to theneed to put more CHW throughchillers in the event that the dis-tribution system returns lowCHW ∆T to the central plant.

A variable-flow design withpumps either oversized and con-trolled by VSDs or banked can re-spond to low ∆T central plantsyndrome. Thus, for the samereason that we as HVAC design-ers provide freezestats upstreamof cooling coils, nonoverloadingmotors to drive pumps and fans,and tube pull space at chillers,boilers, and air-handling units,we need to design chilled waterplants that can anticipate thepossibility of low CHW ∆T andrespond to it. Therefore, I believeit ’s time to put primary-sec-ondary pumping back into ourtool bag of applications to ad-dress specific design situationsand adopt a new paradigm forchilled water system design. HPAC

BibliographyUniversity of Wisconsin’s seminar

on Chilled Water Plant Design, orga-nized by Harold Olsen.

Burt Rishel’s seminar speech, “Cur-rent Trends in HVAC Water SystemDesign.”

Gil Avery’s article “Designing andCommissioning of Variable Flow Hy-dronic Systems,” ASHRAE Journal,July 1993.

Al Utesch’s seminar speech at theUniversity of Wisconsin.

staged based on their ability tomaintain leaving CHS tempera-ture.

The advantages of this system,besides minimizing pumpingpower, are:

◆ It minimizes the potential forlow system ∆T by eliminating

Chilled water plant design

78 HPAC Heating/Piping/AirConditioning November 1996