SERI/TP-254-1928UC Category: 59c

Empirical Validation UsingData from the SERI Class-AValidation House

Ronald JudkoffDavid N. Wortman .Jay Burch

April 1983

To be presented at theASES 1983 Annual Meet ingMinneapolis. Minnesota1-3 June 1983

Prepared under Task No. 1053.00WPA No. 304-82

Solar Energy Research InstituteA Division of Midwest Research Institute

1617 Cole BoulevardGolden, Colorado 80401

Prepared for theU.S. Department of EnergyContract No . EG-77-C-01-4042

Printed in the United States of AmericaAvailable from:

National Technical Information ServiceU.S. Department of Commerce

5285 Port Royal RoadSpringfield, VA 22161

Price:Microfiche $3.00

Printed Copy $ 7.00

NOTICE

This report was prepared as an account of work sponsored by the United StatesGovernment. Neither the United States nor the United States Department of Energy,nor any of their employees, nor any of their contractors, subcontractors, or theiremployees, makes any warranty, express or implied, or assumes any legal liabilityor responsibility for the accuracy, completeness or usefulness of any information,apparatus, product or process disclosed, or represents that its use would notinfringe privately owned rights.

TP-1928

EKPII.lCAL VALIDAI'IOlf USING D&rA ROM mESEB.I CLASS-A VALm&7IOlI BOUSE

~aa1d JuelkoffDnid B. Worman

~., lurch

Solar "1'17 ...earc.b Iut.1b1t:e1617 Cole Boul.eYari

Go1AJell. CO 80104

A residential test building at the SERIInterim Field Site was monitored at theClass-A level during the spdng of 1982. Thebuilding was also modeled on three buildingenergy analysis .1JIIulatioftS-DOE2.lA, Bt.AST3.0. aDd SERIlES--usiaa ...sured weath.r datafrom the test period and the location. The••••ured enerlY pertoraAnce data, and thatpredicted by the siDlulatiol18. _re coapared.More cot'rect input Ules for the codes weredeveloped using measured values of inputparameters. and the results ",ere also com­pared with the measured performance data.The C01llPari80na show that input errors cancontribute to predicted auxiliary energyr6quiresaenu which are on the order of 60% ormore higher than the measured loads, and thatiaprov_enu in input variables could reducethese errors 8ianiflcantly.

1. Ilft'IODIJC'fIOR

The usefulness of building energy analysissimulat1.ona (BEAS) depends to a lara. excenton the confidence _ have in the results ofsuch s1111Ulat10na; 1.e., their validity andaccuracy. Our group at the Solar EnergyResearch Institute (SEal) has developed andimplemented SOlDe :Dethode of evaluatiDg lEASand of determining the limitations on theirapplication to building energy analysis.

The validation work reported here consistedof monitoring a residential building at avery high level of detail, known as theelasa-A level (Ref. 1), and using ehese datato test the validity of several SEAS. Previ­ous validation work at SERl included cOlllpara­ttve studies and analytical verification (seeRef. 1). These comparative studies haveshown that there 1s significant disagreementin the energy performance predicted by dif­ferent SEAS for the salDe simple building,primarily in auxiliary energy requirements.However. the comparative studies alone couldnot show if any of the BEAS are correct,because there was no standard against which

1

to compare the results obtained from the var­ioul codes tested. Analytical studies haveshown that selected _jor individual thermalmechaniS1lle, including conduction tbroughwalll and the interaction of solar radiationw1th "thermal 111&18 J are correct as modeled.However. no test. were available to determinebow well theae .chani.. work wben otbermeehani... are present. nor how _11 otheri.portant thermal mechanislDs, such IS radl­tion heat transfer between surfaces, behave.

Frequently sampled and calibrated meaauredvalues of temperatures, heat fluxes. andother energy flows are needed to evaluateBEAS. to eliminate the proble1llSthat resultfrom usiog either comparative studtes oranalytical verification. Empirical data pro­vide the standard of reference against whichbuilding energy performance predictions fromvarious BEAS can be compared. The data alsocaD provide detailed information about thethermal mechanislDS in a building. Such dataare needed to verify the accuracy of themathematical formulation of individualexchange mechanis_ currently 1n the eedes ,and, perhaps more tMportantly. to ~prove theformulation of more complex mechanisms.

Validation work has focused on the SERl Vali­dation House. a four-zoned, single-storyranch house over a erawlspace (Fig. 1). Thebuilding 1s located at the SERI Interta !ieldSite near Golden, Colo. The building wasmonitored with an automated. digital data

1 4

2 3

Fig. 1. Diagram of test building showingzone numbers

acquisition sy.tem that take8 over 230 chan­ne18 of thermal and other data. Data frailthe house vere compared With building .n.raYperformance predictions from three II&theaati­cal IIOdela: B1.AST 3.0, SEllIIlES, and DOE2.LA.These COliparisOD8 were originally perfor.edu.ing .tandard, referenced th.r.ophyaicalproperty data. Additional comparisoM ,..r.alao sade using ....ur.d valuea for select.dsat.rial and building properties. Thea.acldi tional caa.s includ.d using _ ..ur.d nl­ues for infiltration rate.. grOUlld t8llpera­ture. .nd zone set points.

Bec.use of the char.cteristica of the teathouae. the acope of this work and concluaionadr.wn frOll it are lillitecl to resid.nti&!­ac.le, akin-lo.d-doainated buildings. Whilecertain conclu.ions could be extr.pol.ted to.pply to other type. of building. , thi. 1&not r.comaend.d. p.rticularly for co...rcial­• cale buildings in which the perf01'llllnce of_ch.n1cal ay.tems is IDOre illlportant than the.kin load. The sy.tem and pl.nt _chani....in the BEAS usecl in the studi.. reported onhere were given Il1n1111al treae-nt and gtnUl~

ally vere not t.ated.

ThiB r.port preaentB only SOll8 of the resultsfrom our empirical validation vork. Horedetailed information can be found in Worc.an.Burch, and Judkoff (2).

Z. VALIDAnOil S!UDUS

Figure. 2 through 7 .how prel11dnary resultsfrOil the validation • tudy on the DOI:-2.lA.BLAST-3.0, .Del SEIl.IR.ES cOlllputer progr....Theae reaulta are presented in terma of ninecuea. The first, or baae case, uses hancl­book or .saumed value. for all theraophysicalproperty and other input.. Thi. correapond.to the data .ource. used in the input filesfor the code. u they are normally usecl.eaa.. 2 through 8 use selected _a.ured val­ue8 of individual variable. in the inputfil.s. Ca.e 9 combines all of the change. inCa.es 2 through 8.

1. JlIuIe ea.e. Handbook or assUlled value.are used for all thermophy.ical input8.Meteorological and geometric inputs aremea8ured.

2. 1Dfiltr.tlon. Same 88 base case.except that hourly zonal infiltrationrates vere measured aDel used to gener­ate the infiltration input for the com­puter codes.

3. Groaad Te.per.t::ure. Same as base ease ,except that measured ground temperatureva. used as input to the ground coupl­ing 8ubroutine8 in the codes.

4. GroaDd Albedo. Same u base caae ,excePt that mea8ured ground albedo "'a8used in the calculation of radiationincident upon glazed surf.ces.

5. Set Point. Same a8 base ease , except

2

TP-1928

that a correction ",as made to the ther­IIDstat set-point based on the averagetemperature of air in the zone when theheater actually turned on.

6. Vall. aad IIoof CoDducunce. Saate 88base c.... .xcept that lIIeasur.d wallaDel ceiling conductanc•• were used.

7 • ~ eo.IactaDee. This case vas notrun. b.cause _.ured Window conduc­t.nc.a vere the s..e .. thoae given bythe ASHRA! Handbook of Fundamentals(3).

8. AMorpt:1Y1ty. This caee was not run,b.c.use the measured .olar spectrumab.orptivity on opaque surfaces was not.ignificantly different fr01ll a••umedv.lu.s.

9. -'-ed. All of the mea.ured valuesin Cases 2 through 6 were used. Thl•case repreaents the highest degree ofcontrol ovar external error sources,.nd presUll&bly ahould yield resultsclo••st to the measured temperature .ndenergy performance of the building.

3. DSULTS

Figure 2 shows the total heating load (inKWh) for the week of 20-26 April 1982. Theloads predicted by the DOE-2.lA, BLAST-3.0,and SEIlBES computer prograll8 are shown .longvith the _asured load for Cases 1-9. InCaae 1. where handbook input values wereused, cod. ,.pr ed i c tions were higher by 46%­62%. cOllpar.d vith lIeasured loads. Inease 5. where the correction was made for theactual thermoetat 8et-point. code prediction.vere high by 21%-50%• In Case 9, where allknown _asured input values were used, thecode predictions were low by 11%-33%. Gener­ally. predictions were most accurate forCase 9.

Figure 3 shows the root mean square (rms)difference between measured and predictedtemperatures in Zone 2 of the house for allnine cases. Zone 2 is the southern liVingro01ll and bas a 1188sive floor 8urface. Ingeneral, results for Zone 2 are typical ofthe results for the whole building. Case 1has rms error8 of between 10 and 1.20 e.Case 5 bas rill errors of from 0.60 to 0.90 C.ease 9 has the largest rms errors: from 0.90

to 1.60 C.

Figure 4 shows the Zone 2 measured peak heat­ing load and the peak heating loads predictedby the three computer codes in Ca8es 1-9.ease 1 predictions of peak load are high by36%-45%. Case 5 predictions are high by Z7%to 40%. Calle 9 predictions are the 1IIostaccurate and fall within *5% of the measuredpeak load.

Figure 5 shows the peak load f or the wholehouse. The pattern is similar to that

TP-1928

ZONE 2 PEAK LOAD..-r----,r----,.---r----,--...,.--.,.--...,--..,..--.,.---,

-:l-l--4---,!:--..;...-....--i---+--.-+---+---1---t3:~-o~ --I--4--4---+---+--+---+---+--+--+--t-':ll'-eW-'r----I---lf---+--+--+----1I---+--+--+---tLei...

o+--+---t~-_+--_t_--+_--'I--_+--+--+_-_t

234 5 • 7 • •CASE NUMBER

a =BLAST3.0 0 =SER IRES

• =DOE2.1A + =MEASUREDFig. 4. Predicted a'lld measured heating loads for Zone 2

WEEK PEAK BUILDING LOAD.,..--,r----r--..,----,r----,.---r----,r---...,.--.,.---.

.+--f----tf---+--+--+---I---+--+--+----i-3:~-"+--t----tl---+--+--+---I---+--+--+----io-eo-'"+--t----tf---+--+--+---I---+--+--+----i:ll'

~

o-t---+---t---+--+--+---I--_+--+--+---f234 5 8 7 I •

CASE NUMBERa =BLAST3.0 0 = SER IRES

•=DOE2.1A + =MEASUREDFig. S. Peak building load for the week studied

observed fDr Zone 2; Case I predictions arethe least accurate and Case 9 predictions arethe most accurate.

Figure 6 shows the hourly temperature profilepredicted by the DOE-2.lA code in relation tothe measured tet:lperature profile for Case 1.Zone 1. Zone 1 was primarily a free-floatingzone during the measurement time periodbecause tamperatures remained above the ther­mostat set-point from hour 36 to hour 168.

4

We observe from Fig. 5 that predicted temper­ature tends to overshoot measured temperatureduring the day and undershoot meallured tem­perature at night.

Figure 7 shows the same information forCase 9 as for case 1 in Fig. 6. In Case 9.we see that predicted temperatures overshootmeasured temperatures by even more during theday than in case 1; however. they undershootby les8 at night than in Case 1.

TP-1928

41 it tie lIDTIKE (HOURS)

---DQ~ MEASURED_~A::.~JU;L

..G

Ba0::

Se

51 toll

tla::~

is ~

of! fi.e..t: ::-

Q w...,0

164 • II &&

:u

~80::aetoli0::~

~15. I'e,

S•

0 .. 41 '71 .. 110TIME (HOURS)

__..DQ~UL VEASURtD

-~~

Fig. 6. Predicted V5. measured hourlytemperature profile for Case 1,Zone 1, days 110 to 116

Fig. 7. Predic~ed vs. measured hourlytemperature profile for Case 9,Zone 1, days 110 ~o 116

There is 101le apparent inconsistency in thedata with respect to the presumption thatease 9 would always yield the .:)st accuratepredictions. The Mit obvious i8 seen inFig. 3 where ease 1 and Cale S exhibitedsmaller ~ telllperature errors than Case 9.'1'hil trend is the reverse of that I.en inFig. 1, where, as expected, the most accurateload prediction was obtained in Case 9. The1Il0st likely hypothesis at this time is that(a) the amount of solar energy absorbed inthe building 1& being overpredicted in allcases, and (b) the conductive 10lses throughvalls and roof are being overpredicted in-Cases 1-5.

This bypothesis is partially supported by thelarae (approxiaately a factor of tvo) differ­ence found between _asured and uSUlled walland roof resistances, as shown in Table 1.

Table 1. Differtntee 1etwee1a Meuured &adAs......t Vall aDd. ~of a-Valaes

building. This explana~ion 1s condstentwi th the hourly temperature profUes seen InFigs. 6 and 7, where the Case 1 predictedtemperature was high in the day and low attdght, and the Case 9 predicted temperaturevas even higher during the day but not so lowat night. This also explains how the heatingloads in Case 9 would be the 1II08t accurateeven though the rms temperature errors inCase 9 were the greatest.

The large t1IIS temperature errors were causedprimarily by overpredictions of daytime tem­perature. Greater accuracy in load predic­tion was still possible, however, because atnight the performance of the building wasprimar1lygoverned by conductive skin 1088es.The overprediction of solar radiationabsorbed resulted 1n the 11%-33% underpredie­tion of loads in Fig. 1, Case 9. Finally,the high degree of accuracy in the Case 9peak-load predictions is also cODsistent withthis explanation. At those times when storedsolar energy was most depleted and envelopeconduction was IDOst dolll1nant, the code pre­dictions were most accurate.

Average measuredvall resistance 3.05 (17.3)

Aseuaed wallresistance fromASHRAE 1.56 (8.83)

Average measuredceilingresistance 13.19 (75.03)

Assumed ceilingresistance fromASHRAE 7.04 (40.00)

The smaller rms temperature errors inCases 1-5 could, therefore, be explained bythe offsetting effects of too high an enve­lope conduc tance and the code's calculationof too lZUch solar radiation absorbed in the

5. COHCLUSIOHS

The work discussed in this paper 18 part of amultiyear, multilahoratory effort: on the partof DOE to improve our knowledge of the mecha­nislIl8 governing the thermal behavior ofbuildings and to formulate that knowledge inmathematical terms. This effort would alsodevelop an analysis of the thermal behaviorof buildings and predictions of that behaviorby collecting high-quality. detailed data andapplying rigorous nlldation techniques. Themain benefit of efforts to validate calcula­tion techniques now built tnto the modelexamined 1s to establish their degree ofaccuracy and their validity for predictingthe behavior of buildings. The validationwork 1s far from complete. However, a numberof early conclusions can be drawn that shouldhelp to guide future activities.

5

TP-1928

Installation of a simpler windowassembly in the test-house.

Development of a measurement tech­nique to determine the amount ofsolar radiation absorbed in thebuilding.

Deteradnation of the sensitivity ofoutput accuracy to isotropic-versus­anisotropic sky models.

• The methodological approach used in thiswork for skin-load-dominated buildingsshould be expanded to include themechanical systems in co_ercial build­ings.

standardwall con­errors in60%, evendata are

• Input assUlIIptions based onengineering references such asductance can cause predictionauxiliary load on the order ofwhen measured meteorologicalused.

• Accurate zone air temperature predictiondoes not guarantee accurate load predic­tion, nor does it guarantee an accuratetemperature prediction on the nextbuilding studied. There is evidence ofcompensating errors giving a false senseof confidence. Any validation methodol­ogy 1IIUlIt account for the possibility ofhidden compensating errors.

• Even when most input inaccuracies areeliminated by using measured theraophy.­ical input data, prediction errors inthe auxiliary load, ranging frOlll 11%­33%, have still been found. This canhave a large impact on building and &vACsystem design options.

• Consistent differences were observed inthe heating load predictions for thethree codes for all cases even thoughinputs were developed to ensure compar­able buildings for each code.

• A more detailed level of analysis andexperimentstion will be necessary todeteradne 1£ these inaccuracies are dueto unknown remaining external errorsources or to internal error sources.This additional work should include:

Corroboration of conductances mea­sured in the walls and ceiling withan ASTM standard large sectionRclamp-onR guarded hot-box.

6.~

This work was sponsored by the U.S. Depart­ment of Energy, Systems Research and Develop­ment Branch, Pass.1ve and Hybrid Solar BeatTechnologies.

1. Judkoff,.R., Wortman,D., O'Doherty, R., andBurch, J. A Methodology for ValidatingBuilding Energy Analysis Simulation,SERI/TR-254-1508, October 1982.

2. l~ortman, D., .~urch, J., and judkoff, R.Empirical Validation of Building EnergyAnalysis Simulations Using Class A DatafrOlll the SEar Validation Rouse, SERI/TR­254-1840, Solar Energy Research Institute,Golden,'Colo. (forthcoming).

3. ASHRAE Handbook of Fundamentals, 1977.

6