See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/265377723

An Improved Cooling Tower Algorithm for the CooiTools™ Simulation Model

Conference Paper  in  ASHRAE Transactions · January 1999

CITATIONS READS
14 339

4 authors, including:

Dudley J Benton Mark Hydeman

McHale Performance Taylor Engineering, LLC
66 PUBLICATIONS   127 CITATIONS    19 PUBLICATIONS   808 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Chemical Reaction Analysis View project

Bubble Mass Transfer Analysis View project

All content following this page was uploaded by Dudley J Benton on 16 September 2014.

The user has requested enhancement of the downloaded file.

 AC-02-9-4

An Improved Cooling Tower Algorithm for

the CooiTools™ Simulation Model

Dudley J. Benton, Ph.D. Charles F. Bowman

Mark Hydeman Paul Miller

Member ASHRAE

ABSTRACT code. The accuracy of each CTSA was determined for a g iven
set of operating conditions by a comparison between the
An 1mpron ·d coolin}!. rower simulatiou alj!.oritlzm was
computed approach and that determined from vendor-
de,·e/oped.for rize Coo/ToolsT~ simulation mode/to support the
supplied performance curves for an array of cooling lower~
opumal desij!.n o(clzilled warer sysrems. The new algorirhm that ha ve been certified by the Cooling Tower Institute.
11·as de w/oped w predict more quickly and more at::cururely th e
In I 925. Dr. Fredrick Merkel ( 1925) proposed a theory
thermal performa nce and energy consumplion f!( a cooling
tower. A 11 arra.>' o_(previous~1· known ana~ytica l and empirical relating e va poration and sensible heat transfer where there i ~
algorithms and the ne w~)' developed empirical ulgoritlzm wen• counterflow contact of water and a ir. such as in cooling
evaluated on the hasis of accuracy and computational e.ffi- towers. Merkel made several s implifying assumpt ion~ that
nency. The accuracy o.f each algorithm wa.t determined.fiJr u reduce the govcming relationshi ps for a coun tern ow cooling
given set ofoperating conditions hy a comparison herwn•nthc lower to a single separable ordmary differen tial equation.
computed approach to rhe inlet wet-hulh lemperature Several authors have presented the derivation of the Merke l
(approach) and that delermined.fi-om l'c•ndor-.wppiied perfor- equation (Lefevre 1984; Feltzin and Benton 1991: Bowman
mance cun•es. The results o(rlzc evaluaTion indicate that the and Benton !996). This equation expresses the number o f
empirical algorithms rl!quirc• crmsiJerah(l' less computational transfer units (NTU) as a function of the integral o f t he temper-
time and provide morec.Y>IL>istent results than do tht• a/la~ytica! at ure difference divided by lhe enthalpy gradient in a cooling
ulgomhms. Tire new empirirul al~orirhm wa~.filund to he nmn· tower. Since the boundary conditions arc known. the in1egral
accurate and more efficient rhuntheprel'iously existing algo - may be integrated. Merkel used the four-poUl t Tchebychcff
rithms. method. Merkel posited tha t the bulk wa te r in con tact with a
stream of air is surrounded by a film of saturated air and that
INTRODUCTION the saturated a iris surrounded by a bulk stream o f air.! Jc made
the following assumptions:
The CoolTooJsTM simulation model emr>loymg the DOE:!
I. The s:~t ura lcd atr film i~ at the temper;llurc of the bulk
simulanon code i~ used for building energy ::malyses and opll-
water.
mtzations. An important com ponent of this model i~ the
ch 1lled water system that utilizes ·•package" type evapora tive ., The s:~turated atr film offer.; no resistance to heat transfer.
cooling towers. The expected performance of these evapora-
3. The vapor content of the air is proportional to the partial
tive cooling tower~ i~ based on cooling tower simulation algo-
pressure o f lhc water vapor.
rithms (CTSAs). A literature re view was conducted to iden tify
the existing CTSAs. An im proved CTSA was developed and 4. The heat transferred from the air to lhc film by convection
compared wi th the cxistmg CTSAs for speed and accuracy to is proportional to the hea l transferred from the film to the
detcnnme lhe best CTSA to be used by the DOE:? simul:!tion ambient atr by e vaporat ion.

J>udley J. Benton and Charles F. Bowman arc wi th Chuck Bowman Associates. inc., Kno;o;villc. -1 enn. !\1ark Hydeman ~~with 1 ay!or Engi·
nccring, Alameda, Calif. Paul Miller is with Pacifk Gas and Elcctri.: Com pany. San Ra mon. Calif.

THIS PRE?RINT IS FOR DISCUSSION PURPOSES ONLY. FOR INCLUSION IN ASH RAE TRANSACTIONS 200 2. V. 108. PL 1. Not to be reprinted .n w hole or,,.,
part W>IIIOU! wrinen perm<SSJOfl of the American Soci ety ol Healing. Refh!leral>r>ll and Air-Condilton ing El\ljineers. Inc.. 1791 Tufie Circle. NE. Atlanta . GA 30329.
Op•n,Ofls~ tu"1du1gs. conclus•ons . or recommendations expressed in this paoer are lhose ot the author (s} and do not necessaray reflect the v1ews of A5HRAE. Wr.tten
quesl •ons and commeols regardo09 thts paper should be receoved al ASHRAE no later than January 25, Z002.
 5. The spcci fie heat oft he a ir-water vapor mixture and the heat time~ the specific heat of water times the change in the water
of vapori7..ation arc constant. temperature plus the heat lost by evaporat ion) is equal to the
6. The Joss of water by evaporation is neglected. heat gained by the air (i.e .. the air flow rate times the change
in the air enthalpy). Both terms arc equal to the mass transfer
7. The force driving heat 1ransfer is the differential enthalpy coefficient times the enthalpy difference times the interface
between the saturated and bulk air. area per unit of volume times the incremental cooling volume.
ln J 943. a corporation plotted NTU a$ a funct1on of the Whereas Merkel used a simplified expression for the beat loss
coolmg tower liquid to gas (1.../G) ratio to plot cooling tower from the water to arrive at his equation. Lefevre used the
dcm;md curves. Numerous approaches have been devised in expression for the heat gain of the air. Lefevre arrived at an
an allempt to compensate for several of lhc above assump· expression for the NTU as a function of the gas to liquid ratio
lions. Mickley ( 1949) introduced temperat ure and humidity (that he assumed to be constant) and the air cnthalpies. Lefevre
gradients with heat and mas~ transfer coefficient:; from the applied a dimensionless correction factor to compensate for
water to the film of saturated air and from the film to the bul~ the model's shortcomings at higher water temperatures.
stream of aJr. Baker and Marl ( 1952) developed the concept of Vance ( 1984) presented mctl10ds for adjusting the perfor-
a "'hoi water correction factor:· Snyder ( 19 55) developed an mance of a mechanical draft cooltng tower for off-design air
emplflcal equation for an overall enthalpy transfer coefficient and water mass flow rates. Fulkerson ( 1988) reported heat
per un it of volume of fill material in a cross flow cooling towe r transfer and pressure drop data for counterflow cooling towers
based on tests that he conducted. Zivi and Brand ( 1956) at vendor test facilities. The ability of several computer codes
extended the analysis of Merkel to cross now cooling towers. to prcdictthc results of tests conducted by the Electric Powe r
Lowe and Chrisuc ( 1961) performed laboratory st udie~ on Research Jnsti tulc (EPRl) on eight crossflow and c1ght coun-
severa l types of counterflow fill. Halle!! ( 1975) presented the terOow fills wa$ reported by Bell ct a!. ( 1989). Benton ( 1989)
concept of a cooling tower characteristic curve where the 1\TU showed that both the Gauss and Lobatto methods of numerica l
is expressed as an empirically derived function of the LIG integration arc superior to the four-point Tchebychcff method
ratio. Kelly (1976) used the model of Zivi and Brand along for determining the number of transfer units. Fellzin and
with laboratory data to produce a volume of cross flow cooling Benton ( 1991} derived a more exact model and compared the
tower characteristic curves and demand curves. resu lts of this model to the Merkel equation. The Fcltzin and
Penney and Spald in~ ( 19 79) introduced a model for natu- Benton model did not include an empirical temperature
ral draft cooling towers using a finite dif ference method. correction factor. Desjardins (1992) analyzed the EPRJ test
Majumdar and Singhal ( 1981) extended the model to mec han- data by employing the concept of an ·'offset" hot water
ical draft cooling towers. Johnson et al. ( 1983) proposed a temperature as proposed by Mickley (1949) and the more
computer model based on the 1\TU-cffectivcness approach exact method of Feltzin and Benton . Twelve CTSAs were
used for heat exchangers. Bourillot ( 1983a. !983b) developed identified from lhls research for further consideration.
the TEFERl computer model based on heat and mass transfer
DEVELOPMENT OF THE IMPROVED CTSA
equations similar to Zivi and Brand. The TEFERI model
assu mes uniform water and air temperatures and flow rates at The objective of the improved CTSA was to accuratcl v
the tnlet and calculates the loss of wate r due to evaporation, so reproduce cooling tower performance and energy consump-
the water flow rate does not remain uniform .as it passe~ tion given minimal tower specifications. The first pha.<;c of
through the cooling tower. development was ba.<;ed on vendor-supplied cooling towct.
Benton ( 1983) developed the FACTS model. whtch pcrfom1ancc data. A second pha.~c will be based on field
employs an integral formulation of the equations for conser- mca!'urcmcnt~ taken on operating towers. The minimal
vatiOn of the mass of air and water vapor. conservation of rcqutrcd tower specif1cation~ arc a single design or operating
eneq,'Y. and the Bernou lli equation to arri ve at a numerical point whether the cooling tower is of the cross flow or coun-
~o lution apan from the Merkel analot,•y. FACTS can accom- ternow design. Several operating points can be used to
modate variable inlet water and air temperatures and hvbrid compute ru1 equivalent single des1gn point. Model selection
filb. but it assumes a constant water flow rate through the for inclusion in DOE:! simulation code was based on the
tower (Benton 1984). Benton and Waldrop (1988) and following criteria:
Bowman and Benton ( 1995) presented the results of compar- I. Computational speed (as many calculation~ arc required for
isons between FACTS and test data. FACTS is widelv used bv simulation and optimization)
utilities to model cooling tower performance. Majun;dar e t ai. ..,
Sunpliciry of daw input (i.e .. only data that arc commonly
( 1983) developed the YERA2D model. VERA2D treats air available to the intended user) .
llow in the cooling tower as two-dimcnstonal and steadv and
water flow as one-dimensional. . 3. Ability to simulate response (i.e.. variation of all m~jor
operating parameters)
Lefevre ( 1984) revisited the energy balance between
water and air that was the original basis for the Merkel equa- 4. Accuracy
tion. The heat loss from the water (i.e .. the water flow rate 5. Al~orith:11 avajlability (i.e .. free of legal cncun1brancc)
6. Compatible source code (VIZ.• FORTRA:\') based on the approach. Separate correlation cocffic tcnts were
developed based on cross flow and counterflow data sets.
7. Completeness (i.e_ not dependent on exces51ve auxih:uy
funcllons or extensive libraricl') The task of generating this empirical cxprcss 1on for the
improved CTSA is basically one of multi-variable regression.
8. Compaemess {i.e .• small enough to be included a~ an inter-
Here. the dependent variable is approach and the independent
nal function of a larger code) variables arc wet bulb. range. water flow. and fan power.
9. Computational stabiLity ( t.c .. not producing sporadic run- Many dcvclopcn. have tac kl ed this problem for empirically
tunc errors. such as diviSion by zero or square-root of :J computing cooltng tower performance. The preVJOU$ efforts
ncgat1vc number) identified have grouped the independent vanablcs and devel-
oped sequential regression models for the dependent variable
10. Robustncs~ (i.e .. consistently convergent with meaningful
or !>Otnc intermedtatc parameter used to compute the depen-
output for meaningful mput)
dent vanablc. All previous versions of the DOE2 CTSAt- usc
11 . Range of applicability ( VIZ . • must cover the nonnal range of ~omc fom1 of seq uential rcgrc1>sion on grouped parameters.
opcrat1on for package tYJ)C towers) Th1s approach treats each additionalmdepcndcnt variable or
Three of the CTSA!- 1dcnt1f1ed in the background group of tndcpcndcnt v:Jnablcs a$ a corrccuon. Grouping the
research. the TEFERI. FACTS. and VERA2D modeb. were independent variables presumes s1gm ficant intcrrclauonship!>
discarded because they fruled to meet criteria 1.2. 6. 8. and/or withm the groups. Separating the parameters presumes negli-
10 above. The following CTSAs were encoded and tested: gible mterrclationships among the groups. Overlapptng the
DOE2 versiOn :!.1 D. DOE2 version 2.1 E. DOE2 vcrs1on 2.2. group!- ( t.c .. one or more independent variables appcarmg in
the Merkel method. LeFevre method. Benton and Feltzin more than one group) increases the regression error. as each
implementation of the LeFevre version. the .. more nearly sequential regression must also m•nimizc the system<~tic resid -
exact.. (exact) method. the LMTD method. and the ;-\TV- u:~ls left over from the previous regression. This regression

effectiveness method. Of these CTSAs. which represent a problem is evidenced by an increasing number of coefficients
mixrure of empirical and simple analytical algorithms. the without significant reduction in error.
foiJowing were selected for further investigation: Simultaneous regression on a lithe independent variables
I. The DOE2 version 2.2. a 12-parameter multi-variable is the best approach to min im1zc the error over an enttrc data
curve fit. sct. Linear least -squares regression was used. A power expan-
sion in volving all the independent variables is the Simplest
2. The Merkel method with the four-point Tchebycheff intc- form. allowing interaction among all oft he variables. A zero-
grauon method. order expansion yields a single term and coefficient. A first-
3. The Benton and Feltzin implementation of the LeFevre order expansion of four independent variables y1clds live
method using the Lob:lllo quadrnturc mstcad of the lcrm!- and coefficient:.. A second-order cxpan~aon. inc ludmg
Romberg nun1erical mtegration method. every possible combmation up to :1 power of two. yield:-. I :'i
tcnns. A third-order cxpanston yield ~ 35 tcnm and so on. A
4. The Benton and Feltz in exact method. a set of simultaneous
third-order expansion proved to be the mos t satisfactory lor
nonlinear ordinary dtffercntial equations using the Rungc-
the present development. yielding the highest regression coef-
1\.utta method of intcgrauon.
ficient. The new CTSA is basically a higher-order. multi-
S. The :-..IU-cffcctivencs!' method. a modified vcrs1on of tl1c parameter enhancement of the DOE2 ratmg function~ ba~cd­
CTSA developed by tlte t:.nv1ronmcnLal Protection AJ:!ency on approach rather than rating.
tEPA).
Two problems arose in pract1cal application!- of such a
The accuracy of each CTSA was measured by computing regression involving so many terms. The first problem wa~ :J
approach temperature as a function of wet-bulb temperature. result of finite digital accuracy. The least-squares matrix
cooltng range. water flow rate. and fan power. based on a become$ increasingly graded as the num ber of terms
single des1gn point and by companng th is temperature to incrca~cs. A graded matrix IS one having very large and very
vendor-supplied data. l'onc of the identified cxtshng CTSAl- small terms that arc all Significant. With finitc-prcc1sJon
computc approach directl y from these other parameters: calculations. the small term$ are lost. and the resulting agree-
rather. the approach is requ1rcd as mput. As the accuracy wa~ ment ts poor. In order to minimize thi!' problem. 80-btt preci-
measured by the computed approach. the most logical way to sion was used throughou t. along with maximal
achieve the greatest accuracy was to develop a CTSA that preconditioning of the regression matrix.lntcl80-bit floating-
pro\' ides approach directly. For a fixed complexity. this would . num b ers .have a range o f10~09-' an d prov1'd c apprOXI-
pomt
also be the most cff1cient wa~ to compute tower perfonnancc mate)~ H< s ignificant figure:.. The second problem that arose
when 11 IS expressed as an approach temperature. Preliminary was that of crrat1c behavior. A htgh-ordcr rcgress1on may fit
testmg of the existing model~ described abO\'C: ind1cated tha t the data pomts quite well yet produce unreasonable result~
an empmcal algorithm would best meet the current objectives. between and outside the datll pomts (i.e .. interpolation and
Therefore. the improved algorithm is a regression model extrapolation). Data point we1ght ing was used to control
 erratic behavior. Erratic behavior is most pronounced at
extreme values of the independent variables (e.g.. at very high
range or very low wet bulb). Multi-dimensional graphics, such
as Figures I and 2 showing isosurfaces and slicing planes.
were used to identify erratic behavior. Where erratic behavior
was identified. neighboring data points were weighted more
heavily until the desired asymptotic behavior was obtained.
The desired asymptotic behavior is shown in Figure I by the
contours drawn in the s licing planes and in Figure 2 by the
curvature of the sheets of constant rating. This process neces-
sarily inc reases the residual at other data points. but it greatly
improves extrapoJation. which is a key objective of the
improved CTSA.
The crossflow vendor-supplied data contained little vari-
ation of the water flow and few points for low wet bulbs. The
FACTS computer code was used to extend these vendor data
and provide control points. Thts was accomplished by numer- FiKure I Coumer.flnll' approach data and regression.
ically modeling a tower of approximately the correct size.
adjusting its characteristics unti I it best fit the vendor data. and
then performing calculattons at very low and high water flow
O;,~~nfl;~
and low wet bulb. These additional control points arc shown A~*"*'~"-~~
as diamonds in Figure 3. whereas the vendor data arc shown s~ o'C<:M~ ll~"'"

as spheres.
AU of the existing CTSAs require some correction for fan
power. Five fan power corrections were identified. These
include the curves from D0£2 versions 2.1 E and 2.2. a nomo-
graph published by Wood and Betts ( 1950}. a curve obtained
from Hudson fan perfom1ance curves using the FACTS
model, and the one-third power law ( 1.c .. air flow is propor-
tional to the cube-root o f the fan power). The improved CTSA
does not requ ire such a correction. as this variation was incor·
(!
porated tn, the multi-variable regression. Tbc regression was
performed based on normalized fan power and normalized fan
speed. The coefficients differ. but the results are essentially
the same. Either form could be used for computational conve-
nience. FiKure 2 Counterflow rating daw and regression.

EXPERIMENTAL DESIGN
Three tower models from each of four vendors. or a total
of 12 model~. were used for testing the CTSAs. Three of the
models from one vendor had three variations. makin~,; a total
of 16 data sets shown in Table I. All tower$ were "package''
type with heat capaeitiesranging from46 to 1667 tons. Towers
I. 2. and 3 are manufactured by vendor A. Towers I and 2 had
the same design point but are crossflow and counterflow.
respectively. These selections speak to the abil ity of the
CTSAs tO account fordiffcrent flow orientations. Tower] was
selected because it IS the largest model in the same type group
as tower 2 . Towers 4. 5. and 6 arc manufactured by vendor B.
Towers 4 and 2 have the same design point and configuration.
thus providing a drrec t comparison between these two
vendors. Towers 5 and 6 are a largc-rangc/closc-approach and
smaller-range/larger-approach tower. respectively. providing
another test of the CTSAs. Towers 7. 8. and 9 are manufac- Figure 3 Cms.~/1nv• daw. re~n?.~sinn. ami conrrnl points.
tured by vendor C. Tower 8 is the closest available model to

..
 the baseline design pomt and another crosstlow reference

TABLE 1
Vendor Towers

Water Flow ..an Power Wet Bulb Range Approach Heat Load
Tower Vendor T y pe (rif/s) (kW) (OC) (OC) (OC) {kW)

I A XF 0.06309 11.2 2~. 6 5.6 3.9 11 70

2 A CF 0.0630tJ 29.{< 25.1\ 5.6 J.<) 1170
3 A Cf 0. 15773 !<9.5 12.!< 11.1 11.1 5859
4 B CF 0.06309 14.9 25.1\ 5.6 J.<> 1170
5 8 C"r 0.022()!< 11.2 I ~-6 I 1.1 ~ -n H19
6 8 CF 0.31 545 74.6 !5.6 5.1\ !C~ 5R59

7 c Xf 0.03842 7.5 25.0 5.6 39 713

8 c XF 0.05886 1!.2 25.6 5.6 3.9 1093
9 c XF 0. 10372 29.~ 25.t> 5.6 3.9 192(>

lOA 0 CF 0.01 457 2.2 25.t> 5.6 3.<> 271

lOB D CF 0.0 1685 37 25.t> 5.6 3.9 313
IOC D CF 0.01893 5.6 25.0 5.6 3.9 351
I lA D Cf 0.01022 2.2 25.t> 5.6 3.9 190
118 D CF 0 .01192 3.7 25 6 5.6 3.9 221
IIC D CF 0.0 1344 5.6 25.6 5.6 3.9 250
12A D CF 0.0087 1 2.2 25.6 5.6 3.9 162
128 D CF 0.01022 3.7 25.6 5.6 3.9 190
12C D CF 0.01 173 5.6 25.6 5.6 J.C) 218

poinl. Towers 7 and 9 prov1dc adduional var~auon of param- fiel d da ta. 1hc inpul rcquircrncnrs arc the same. It 1s cxpectco
eters. Towers I 0, II. and 1~ arc manufactured by vendor D. that lhc CTSA!- would be used to compute an average implied
These models come with three fan motor size!> each (Lc.. A, H.
and C). The single cell "box·· size ts the same for all models. TABLE 2
Towers I0. !I. and 12 arc identical except for the fil l. tlJUs Comparison of Speed.
providing a differen t test of CTSAs.
DOF.2 Ver..ion 2.2 1.00
RESULTS Merkel 5.49
The five existtng algorithms and the new algorithm all NTU-Effccllvcncss 3. 17
sattsfy the basic selection criteria: however. they do vary in
fcltzin 101.7
computational speed and accurac y. The CTSAs can be sepa-
rated mto two categories: analytically based and empinca11) E.xact 100.9
based. Typically. the analytically based CTSAs an.' signifi- New Ce>untcrllc\\' 0.0013
cantly slower than those thar arc empirically based. The
New Crossflow 0.001:'
analytically based CTSAl> can further be separated into two
catcgones: those requ iring numerical integration of one or !\otc· A :. the ntv. tiS.-\ fN"O\"Kie\ appn»th dtrc:rdy. ch~r.:- arr con.."tdc"Dhlt<
more differential equations (viz .. Feltzin :lnd Exact) and those computauonal s 'vU"<;:\... t \'cn t.hou~h thr fll·nctton ttseJr in\.-olve~ .3.._ c-on$-Qnt~
and 2oJYiroiUmattty JI W: umc ~ f.$ many open.h Otb. T ht'" tlC \\" CTSA ptovtdc.!rio nc1
requiring evaluation of a simple equation (Merkel and ;>\TU· compucauon.J.l b t nefu when h cnuvcly com?utmg av.oul:a.blc water flow np;u·-
effec11veness). The computational speed of each of the CTSA!> IIV Or r eqUired fan r ower.
compared 10 the base case IS listed in Table 2.
The only required user input data is the design point. test poinl (analogous to the design potnt) when analyzing field
which mcludes the fan power. water flow rare. wet bulb, range. da ta from a single tower. The acc ur-<~cy of each oft he candidate
and approach and whether the tower is crossilow or counter- CTSAs has been rested agains1 the vendor data. This compar-
flow . Alrhough the CTSAs have not yet been tested against ison includc5 669 data points for the counternow lower.- and
34 7 data points for the cross flow towers. The resu lt~ arc Figure 4 shows a comparison between the error computed
summanzcd in Tables 3 and 4. by the DOE2 version 2.2 CTSA and the new CTSA for Tower
I. a c ross flow tower. operating at the rated now of I 000 gpm.

TABLE 3
CTSA Accuracy for Counterflow Cooling Towers ("C)

CTSA DOE2 V2.2 Merkel NTU- Effcct i vcncs~ Fcltzin Exact Cnew
Avence Error 0.0 0. 1 0.2 0. 1 0.1 0.0
Maximum Error 1.7 1.6 3.Q 1.4 1.4 0.9
Standard Deviation 0.4 0.4 1.0 0.4 0.4 03
95•1. Confid ence Interval· 0.& 0.<> 2.0 0.8 0.& 0.~

Note TheY~~- cnnrutcnC'~ uucrv.:.l •~:. r c ,•c:.hn~ ~tah~CIC21 mcuur..: Clf lhc ( · r~A ·, 3CC:UrJry ~c'H'·''' h Hli:t.: ' lnl(l ~ccount the dt.:\,:tUC'In and the ~mph.: :.aa: Ti'k: Y~~. ctm•
f•dcrc~ 1/Hcrv:.)
lor the 11(\\ l'T~/\ L\. :tppro:umatcly h WHl'urd\ Ihat or
O()C~ v·crMI.'In : . :

TABLE 4
CTSA Accuracy for Crossflow Cooling Towers (0 C)

CTSA DOE2 V2.2 Merkel NTU-Effectiveness Fcltzin Exact Cnew

Average Error -0.2 0.2 ~-~ 0.2 0.2 0.1
Maximum Error 2.2 1.4
.,-··...
~

1.4 1.4 1.2

Standard Deviation 0.4 0.3 0.6 0.3 0.3 0.3
.
95% Confidence Interval 0.8 I 0.6 I.:! 0.(1 0.6 0.6

Nate. The qs~. coafidcncc •ntcn·a.I u .1 rcvnhn~ stahJr-t1nl mc;u.urr or I he CTSA ·1 accuraC'"' ~C'auM: rt u\cc:~ mto :account tht drv1:auon .u:d the: ~pJc stZa: The 9~· . con·
fidcncc wUcf'\"JI for the nt\\ l"TSA '"' 1J)fY01UftUtdy lWO-lhlrd.' Lh-tt or DOC~ vct.szon 2 2

TABLE 5
CTSA Accuracy for Crossflow and Counterflow Cooling Towers (0 C)

CTSA DOE2 V2.2 Merkel NTU- Effcctr vcncs~ Feltzin Exact Cnew
Tower 1 (Xf) Std. Drv. 0.20 0.16 0 .4<1 O.D 0. 13 0.11
Tower 2 (CF) Std. De\·. 0.29 0.39 054 0.36 0.36 0.26

The error IS the difference between the cold water temperature

predicted by the CTSA and that predicted by the cooting tower
vendor. For this cooling tower. the new CTSA is well behaved
in comp:mson with OOE2 version 2.2. The 95% confidence
interval ts approximately half that of DOE2 version~-~ 10.19
n .0.36J.
A companson may be made between tower5- I and 2.
whtch have the same des1gn point but arc crossnow and coun-
terflow. respectively. This comparison. shown in Table 5.
md tcates that all of the CTSAs. including the new CTSA.
more accurately predict the performance of the crossnow
tower than the counterflow tower. These resu lts include 96 12 14 16 18 20 22 24 26 28
Wet Bulb Temperature (C) I
data pouus. Figures 4 and 5 show the computed error a s a func-
tion oft he wet-bulb temperature for an array of cooling ranges
Figurt' 4 DOE~ w r.tion ~- ~ CTSA vs. tmn!r 1 datu.
for DOE ~ ve rs ion 2.2 and the new CTSAs. T hese results show
that for the counterflow tower. the error computed by the new
CTSA mc reases with incrcasmg range as do the orhcrCTSAs
(~ec Fit!urc 5). For rhe case of tower 2. the new CTSA i~ no r Table 6 and Ftgurcs 5 and 6 show a comparison between
as well behaved a s the DO E~ version 2.2 CTSA at hi1;hcr tower~ ~ and 4. These counrcrflow tower:- have the sanK
ranges.

'\ 1 11 >'1-4
 TABLE 6
CTSA Accuracy for Vendors A and B (°C)

CTSA DOE2 V2.2 Merkel NTU-Effcctivencss Feltzin Exact Cnew

Tower 2 Std. Dev. (96 pts.) 0.2 9 0 39 054 0.36 0.36 0.26
iTower 4 Std. Dev. (8S pts.) 0.27 0 . 14 0 .46 0.15 0.1 5 0. 16

TABLE 7
Accuracy for Large-Range/Close-Approach and Smaller-Range/Larger Approach (•C)

crsA OOE1 V2.2 Merkel NTU- Effectivcncs~ 1-'eltzin Exact Cncw

Tower 5 Std. Oe~·. (4S ph.) 0 .1 9 0 .2!; 0.71 0 .29 0.29 0.22
Tower 6 Std. l>e\'. (53 pt>.) o. 1o 0.06 0 .06 {),07 0.()7 0 .11

TABLE 8
CTSA Accuracy for Vendor C Towers (•C)

CTSA DOE2 V2.2 Merke l NTU-Effcctiveness l-'c1t7.m Exact Cncw

To ...·er 7 Std. De\'. (84 pts.) 0 .28 0 .41 11.45 0 .42 0 .42 0.32
To-..·er 8 Std. De\·. (84 pts.) 0 .46 0 .2Q I 0.60 0.21! 0.2}; O.J4
Tower 9 Std. De\'. (83 pts.) 0.53 0 .19 I 0.67 0.1 7 0.17 0 .33

TABLE 9
CTSA Accuracy for Vendor 0 Cooling Towers (°C)

CTSA DOE2 V2.2 Merkel NTU-Effectiveness Fclt7J n Exact Cncw

Tower JOB Std. De,·. (82 pt.~.) 0.43 0.37 0.96 031 0.31 032
Tower liB Std. De". (96 pts.) 0.32 0 .42 0.85 0.35 0.35 0 . 11<
!Tower 128 Std. De,·. (94 pts.) 0 .62 057 0.83 0.54 0 54 0.29

o.a1 -=====-=
-r ·+-R..2.7s .
1
0.8 !
I
I
I
!

i
I

0_6 -;......... R..S.56t i = -- -. -- -- - --l--- 0.6 1

I f i
-
._fi'"8 33
__._ • •
J

••-j ,- - • I • -~
0 .4
0.4- ~R~1t 1 ~ ! ' ~ I • I
2:(5 020 __•.____.~--··~1--~i~~-~
I
• t
0 02 .:. ~
~---------
(50~=~~ ~ ~ -02 -===-----------c..·! --.- - - --ii;:... ~ !r
~~-~ -l :~ ~ I~
+-'

-0.4 • 'I
-o.s I

l'
-0-~ - - ]_ _ _ _ _: l
12 14 16 18 20 22 24 26 28 :
i2 14 16 18 20 22 24 26 28 Wet Bulb Temperature (C) I
Wet Bulb Temperature (C) f

Fi~:ure 6 DOEJ vtnion J.J CTSA ~·s. tower J clato.

Fi~:ure 5 lmprol'l!d CTSA l '.t. tower 1 data .

CTSA does a better job of predicting the performance of the

small-ran£ellargc approach tower than the large-range/close-
design point and provide a direct compariwn between the two approach tower.
vendors.
A comparison may be made among towe rs 7. 8 and 9. all
Table 7 show!- a comparison between towers 5 and 6. vendor C towers w1th the same wct-bulb\approach\rangc
Towers 5 and 6 arc a larg.e-rangelclose-approach tower and a design condit ions but with increasing water flow. fan power.
smallcr-rangel!arger approach tower. respectively. Thts and heat load. This comparison. shown in Table 8. s uggest~
com pan son suggests that like the analytical CTSAs. the new that unlike the DOE2 version 2.2 CTSA. which provide!-

•\ (' -1! ~ -· • ...:

 1 1 ·- -
t-• - r-- I I
0.8
0.6 • I
0.8
0.6
a:R~78
•ReS 56 I
• I

.
-...
0 .4 "' •
• R•8 33 I
£ 0.2 • • I
....
•"' ..
I •
•I
0.4 • R"11 .11
! I 1
I
f-t~ -.-
......
0 0
'
•• II! i
£ 0.2
~ 0
! ; ••.I •
fll ., • .......__.
• •••••••••• "'
s.:·•R•5.56
R-2.7~~
-0.2 t
•
sE:~:! t- :·~
i : ~ =1~ ~
. ~
·li ~ ~:.
w I
-0.4
I

:~:: L-, -~-. . .,~r--+--+~--+-i

&R•833 1
-0.6
-0.8
ieR•11 1! : I c.

-1
i I : '
--+--
12 14 16 18 20 22 24 26 -1 - -: - - - . ! ----+---__!____J.___ -+-___,
Wet Bulb Temperature (C) 12 14 16 18 20 22 24 26
Wet Bulb Temperature (C)
. ......... .... tlo.o\ _ , _ • .,.,.
..-..-.~

Figure 7 improved CTSA •·s. tower:! data .

Figure 9 Jmprn•·l!d CTSA 1·s. tower 4 dwa .

0.8
0.6 1

£
0.4 •
02 _ __;__ __ _; -_ _;_.,---+-=--'-A--=
1--
·+-~
0.8 rl . . . . R-~ 44 I
I I
I i II I

0.6 tl-e- R-ssal '

.._ R-667J
o o----~---4-~~~ ..~..~~~~._~._-4 I I l !
~ ~.2--------~~-;~~~--~L-------~--~
~.4 --------~--~~~-r~~~~----~----
-
£
~
0 .4
02
0
I
' .-
I

--.....;
: I I
I
i
!
·0.6 ------"----i--~----+--+---
.E -0.2 ' I i
'
j

.I
·0.8 - - - -- -+----;--r---:---+--......;.-- -04 ~--~--~---r--~---.--~-------­
·1 --~---~---~-~-----+- ·0.6 f----i---+---+---+---+----+----+---1
12 14 16 18 20 22 24 26 28
-0.8 +-----------+-----+- ---1f--.:__-
Wet Bulb Temperature (C)
•••nL CJ VUCT'SA. & T....,.•O.ta -1 +----------~------~-~--~-- I
12 14 16 18 20 22 24 26 28 j:
Figur~8 DOE:! version 2.:! CTSA vs. tower 4 data. d
Wet Bulb Temperature [CJ I'

Figure I I lmpro1·1!d CTSA vs. tower II data.

'I.I A comparison may be made among towers I 0. II. and I ~­

all vendor D lowers with the same wet-bulb\approach\rangc
design condition~ and fan power but with decreasing water
flow and heat load. Towers I 0. II. and I 2 arc physically idcJ1=-
tical. except for the fill. This comparison. shown in Table 9.
suggcsg · th:lt the new CTSA more accurately prcd Jct~ the
perfom1ancc of rhese towers than the other CTSAs.
·0.6 1

-o.8 j ;I DISCUSSION
·I ------------~---------~ ----J
12 14 16 18 20 22 24 26 28 For the span of flows. we t-bulb temperatures. cooling
Wet Bulb Temperature (C) ranges. and approaches to wet-bulb temperature considered in
this study. the maximum error in computed approach over the
Figure /0 DOE:! •·ersion 2.:! CTSA vs. Iowa/'/ data. entire range of data was from 0.9°C to 2.2°C for all of lhc
CTSAs except for the !\'TV-effectiveness CTSA. Differences
exist among the CTSAs and between the empirical and the
poorer corrclauons as heat load and fan power increase. the analyt1cal CTSA~ that arc worth noting. For example. Llu.:
new CTSA provides consistent results independent of these analytical CTSAs requi re considerably more computational
parameters. Figure 9 shows that the new CTSA il' well time. Somewhat surprisingly. all of the CTSAs appear to do <1
behaved for rower 8. bcllcr job of prcdictmg the performance of crosstlow than
counterflow towers. This may be due. in part. to the fact that
the error appears to increase with increasing cooling range for Bourillot. C. 1983b. TEFERJ. Numerical model for calculat-
counterflow towers. The empirical CTSAs appear to provide ing the performance of an evaporat ive coolmg tower.
more consistent resuhs over a wide span of cooling range$ EPRJ CS-3212-SR. Palo Alto, Calif.: Electric Power
than do the analytical CTSAs. Research Institute.
Bowman. C.F.. and O.J. Benton. 1995. Cooltng tower per-
CONCLUSIONS
formance. Thermal Performance Improvement Seminar.
Care must be exercised in cxtrapolatmg the result~ Palo Alto. Calif.: Electric Power Research Institute.
beyond the range of flows. wet-bulb temperatu res. cooling Bowman. C.F.. and D.J. Benton. 1996. Oriented spra y-
ranges. and approaches considered . Neither the em pirical nor assis ted cooling tower. Hou~ton . Tex.: Cooling Tower
the analytical models considered appear to cxhtbit any partic-
Institute.
ular advantage in extrapolating results beyond the range of
Braun. J.E .. S.A. Klein. and J.W. Mitc hell. 1989. Effective-
available data.
ness models for cooling towers and cooling coils.
The purel y empsncal CTSAs were shown to provide
AS/IRA£ Transocfions95(2): 164- 174.
accurdey comparable to the analytical ones. In general. th e
new CTSAs arc more accurate and better behaved than DOE~ FWC. 1943. Cooltng tower performance. Bulletm CT -43-~.
vers10n 2.2. This conclusion applies to both crossnow :~nd Foster Whee ler Corporation.
counterflow cooling towers and to all of the cooling tower Des_1ardins. R.J . 1992. Using the EPRI test data to verify a
v endors. There arc a few exceptions to thts conclusion. For more accurate method of predicting cooling tower per-
example. the error for the new CTSA is slightly higher than formance. Houston. Tex .: Cooling Tower Institute.
that ofDOE2 verst on 2.2 for tower 5. and the new CTSA is not Fcltzin. A.E .. and D.J. Benton. 1991. A more nearly exact
well behaved for tower 2. However. overall. the e rror for the reprcsentah on of cooling tower theory. Houston. Tex.:
new CTSA is approx imately two-thirds that ofDOE2 version Cooling T ower Institute.
2.2. Fulkerson, R.D. 1988. Comparauve eva luation of counter-
flow cooling tower fills . Houston. Tex.: Cooling Tower
ACKN OWLEDGMENTS
Institute.
The authors acknowledge the financial support of Pacific Hallett. G.F. 1975. Performance curves for mechanical draft
Gas and Electri<: Corporation in making thls presentation cooling towers. Jnumol oJ EnJ!ineuing for Power. p .
posstble. The authors also express appreciation for the 503.
vendors who provided the performance data.
Johnson. B .M .. O .K. Kreid. and S.G. Hanson. 1983. A
method of comparing performance o f extended-surface
REFERENCES
heat cxcbm1gers. Heat Trw1.~{er Engi11eering 4( I ).
Baker. R.. and L.T. Man. 1952 . R(:frif!t'rotitm Engineenllf!.
Kelly. 1'\.W. 1976. Kelly's hanJhrmJ.. f~{ cm.~sflow ctwlin~
p. 965.
tmn'r p('r.formonce. Kansa~ City. M1ssouri : Neil W.
Bell. D.M .. B.M . JoluHon. and E.V. Werry. 1989. C oolin); Kell y and Associates.
tower performance prcdtction and improvement. EPRI Lefevre. M.R. 1984. Eliminattn!; the Merkel theory approxi-
GS-6370. Palo A Ito. Calif.: Electric Power Re search matiom:- Can 11 replace the empirical "temperature cor-
lnstnutc. rection factor"? Houston. Tex.: Cooling Tower lnstttutc_
Benton. D.J. 1983 . A numerical simu lation of hc:u transfer m Lowe. II.J.. and D.G . Christie. 1961. Heat transfer and pre~ ­
evaporative coolmg towers. Report WR28-i-900-1 I 0 . surc drop in cooh ng tower packing and model studte~ of
Knoxville. Tenn.: Tennes~e Valley Authority. the resistance of natural-draft towers to atrllow.Imernu-
Benton. D.J. 1984. Computer simulation of hybrid fill m tinnol Division of Heat Transfer. part V. p . 933 . New
crossflow mechantcal-indueed-draft cooling towers. York: ASME.
Wtnter Annual Meeting. ASME. New York. Merkel. F. !925. Verduftungskuhlung. VDI Fnrschungsor-
Benton. D.J .. and W.R.Waldrop. 1988. Computer simulation h(•if<'fl. no. :!75. Berlin .
of transport phenomena m evaporative cooling tower.-. Mickley. II .S. IQ4Q. Clwmical l:,"ngineerinf!, Pmgrl!ss 45 :
Jnumal of Eng ineainp. for Gus Turb111~·s ami /'mwr 739.
110: 190. Majumdar. A .K .. and A.K. Singhal. 1981.1'£R.A:!D- A com-
Benton. D.J. 1989. Lcuer to CTI ~on-Merk el Comm ittee puter program for t.,..o-ditnt'tlsinnal una~t·.~is of .flow.
Members. July 24. hear and mas.~ transf(•r i11 e~>opnrutil·(· conli11g towers.
Bourillot. C. 1983a. On the hypotheses of calculating the vol. 11-Uscr"s manual. Palo Alto. Calif.: Elcct nc
water flow rate evaporated in a wet coohng tower. EPRI Power Research Institute.
CS-3144-SR. Palo Alto. Calif.: Electric Power Research Maju mdar. A. K .. A .K. Singhal. and D.B. Spalding. 1983.
lnstllute. VERA :!D: Prof!. rom .for 2-d ano(vsis of.flnw. heat. and

1\C-02·"·~ <)
 mass tralt~{er in evapt>rattve r·nolinl!. f()WI!rs- Vt>/ume I : Snyder. 1\:.W. 1955 . Effect of air rate. water rate. tempera-
Mathematical formulatinn. .wlu/lnn prncedure. anJ lure. and packing density in a cross-flow coolmg tower.
applicarinns. EPRI CS-~923. Palo Alto. Calif.: Electric Chemical Engi11eeri11~ Prngre.u 52(18): 61.
Power Research Institute. Vance. J.M. 19&4. Determination of fan flow and water rate
adjustment for off-design cooling tower tests. Houston.
Penney. T.R., and D.B. Spalding. 1979. ValiJatimz t!{ cool- Tex.: Cooling Tower Institute.
ing rower a/la~r=er (I' £R.4). vols. I and 2. EPRI Reporl Woods and Bell~. 1950. Engineer 189: 337.
fP-12 79. Palo Alto. Calif.: Elcctnc Power Research Ztvi. S. M.. and 8 .1:3. Brand. 1956. An analysis o f the cros~­
lnstitulc. flow coolill£ tower. R(~{rig(•ration £11gine<•ring.

10

View publication stats