SERltTP-254-3371UC Category: 232DE88001173

International Energy AgencyDesign Tool EvaluationProcedure

R. Judkoff (SERI)

S. Barakat (National Research Council, Canada)D. Bloomfield (Building Research Establishment,

United Kingdom)B. Poel (Bouwcentrum, Holland)R. Stricker (Fraunhofer Institute of Building Physics,

W. Germany)P. van Haaster (Bouwcentrum, Holland)D. Wortman (Architectural Energy Corporation, USA)

July 1988

Prepared forAnnual Meeting of ASESCambridge, Mass.June 20-24, 1988

Solar Energy Research InstituteA Divisionof MidwestResearchInstitute

1617Cole BoulevardGolden, Colorado 80401-3393

Prepared for theU.S. Department of EnergyContractNo. DE-AC02-83CH1 0093

NOTICE

This report was prepared as an account of work sponsored by the United States Government. Neither theUnited States nor the United States Department of Energy, nor any of their employees, nor any of theircontractors, subcontractors, or their employees, makes any warranty, expressed or implied, or assumes anylegal liability or responsibility for the accuracy, completeness or usefulness of any information, apparatus,product or process disclosed, or represents that its use would not infringe privately owned rights.

Printed in the United States of AmericaAvailable from:

National Technical Information ServiceU.S. Department of Commerce

5285 Port Royal RoadSpringfield, VA 22161

Price: Microfiche A01Printed Copy A02

Codes are used for pricing all publications. The code is determined by the number of pages in the publication.Information pertaining to the pricing codes can be found in the current issue of the following publications,which are generally available in most libraries: Energy Research Abstracts, (ERA); Government ReportsAnnouncements and Index (GRA and I); ScientifiC and Technical Abstract Reports (STAR); and publication,NTIS-PR-360 available from NTIS at the above address.

INTERNATIONAL ENERGY AGENCY DESIGN TOOL EVALUATION PROCEDURE

R. JudkoH, Solar Energy Research Institute, USAS. Barakat, National Research Council, Canada

D. Bloomfield, Building Research Establishment, United KingdomB. Poel, Bouwcentrum, Holland

R. Stricker, Fraunhofer Institute of Building Physics, W. GermanyP. van Haaster, Bouwcentrum, Holland

D. Wortman, Architectural Energy Corporation, USA

ABSTRACT

Detailed state-of-the-art building energy simulation modelsfrom nations participating in International Energy Agency(lEA) Task VIIl are used to develop a quantitative procedureto evaluate more simplified design tools. Simulations areperformed with the detailed models on a series of cases thatprogress systematically trom the extremely simple to the rel­atively realistic. Output values for the cases, such as annualloads. annual maximum and minimum temperatures. and peakloads. are used 'to set target ranges with which the result.sfrom more simplified design tools can be compared. Themore realistic cases, although geometrically simple. test theability of the design tools to model such combined effects asthermal mass, direct gain windows. overhangs. internally gen­erated heat. and dead-band and set-back thermostat controlstrategies.

INTRODUCTION

With the increased acceptance of personal computers hascome a proliferation of building energy design tool software.A survey among lEA participating countries listed 215 suchdesign tools. 156 of which were developed in the UnitedStates alone (D. There is little jf any objecrive quality con­trol of this software. It is essentially a buyer beware situa­tion. An evaluation of a number of design tools carried out inlEA Task VIII showed large unexplained differences betweenthese tools, even when run by relatively "sophisticated" users(2). It is important that the design industry not become disil­lusioned with these tools. since the potential for energy sav­ings and comfort improvements are great through their use.

In recognition of this problem. a joint effort was begun in1986 under lEA Task VIII to investigate the possibility ofdeveloping a quantitative design tool evaluation procedureand to provide a technical basis for doing so. In this paper.lVe summarize the results of that effort. A full-length lEATechnical Report will also be available on this topic by thetime this paper is published. Based on this work, several ofthe participants have indicated the intention to develop eval­uation procedures "customized" for conditions in their owncountries. Recently, concern about this issue in the UnitedStates prompted the formation of a Standards SubcommitteelVithin ASHRAE TC-II.7. The goal of that subcommittee is toformulate standard test procedures for building energy designand analysis tools.

APPROACH

It was not the intention of this project to create proceduresfor "validating" design tools. This would have been a large, ifnot impossible. undertaking, and was beyond the scope ofTask VIII. Instead. the strategy was to use a number ofdetailed comouter programs to generate "reference" dataagainst which more simplified toots could be compared. Thereference programs were selected by experts in each of thepar ttcipating nations as represenung the current state of theart in building energy simulation. These reference programs

had been subjected to a number of validation studies bothwithin the context of lEA Task VIlI and bv individual coun­tries 0,11). It is well known from these studies and othersthat even the reference programs freouently differ dependingon the climate and building-type modelled (5). However. inthis project WI!! chose to accept legitimate internal modellingdifferences between the reference codes to establish usefulranges of target output values for the simplified designtools. Legitimate modelling differences are those not due toinput errors or code bugs.

The objective of the project was to see if a series of casescould be developed for which agreement between the refer­ence programs was adequate to help in determining A) theoverall credibility of a design tool and B) the appropriatenessof a design tool for a given application. Additionally, it washoped tha t the reference cases would be useful for diagnosticpurposes when a design tool disagreed significantly with thereference simulations. While such disagreement would notnecessarily mean that the design tool was faulty, it would in­dicate that the design tool disagreed with the current state ofthe art in building energy simulation as defined by the projectteam.

SPEClFICAnON OF THE CASES

A series of buildings were specified that proceeded from thethermally simple to the realistic one parameter at a time.The cases were defined so that thermal properties, geometricproportions. and thermal responses were meaningful in termsof actual residential buildings. Since this was an internation­al effort, the building specification represented a compromisebetween American and European construction. In general,the buildings contained more thermal mass than common inthe Untted States. .

The cases were jointly defined by the project team to test theability of design tools to calculate shell ioads in residential­type buildings. Such leatures as thermal mass, direct gainwindows, window overhangs, internally generated heat anddead-band and set-back thermostat control strategies wereincluded to emulate those found in a variety of residentialbuildings. We did not define cases that required simulatingmechanical equipment operation. The mechanical equipmentwas assumed to be 100% efficient and adequately sized tomeet the peak load.

Figure I shows the basic building geometry. The buildinggeometry remains constant for all cases except for size andorientation of windows and the presence or absence of over­hangs. Table 1 lists the key thermophysical properties of thebuilding. These properties either define a heavyweight orlightweight case. The floor was assumed to be thermallyisolated from the ground because the coupling between thebuilding and the ground is not well modelled even in the de­tailed simulation programs. Material thicknesses were ad­justed so that the conductance of the walls, roof. and floorwere equal in the light and heavy cases. Table 2 lists thecharacteristics of each case.

INPUT EQUIVALENCY AND REFERENCE CODEDIFFERENCE5

All the project participants were experienced in buildingenergy simulation. They were aware of the need for rnam­taining strict equivalence between the input files of thevarious programs. We jointly developed the building specifi­cations from which the input files would be derived. Weattempted to document the specifications as unambiguouslyas possible. Even so, several iterations were required untilwe could reach mutual agreement on the equivalence of ourinput files. Input equivalence is not always a straightforwardconcept, especially where the modelling approach is very dif­ferent between codes. Even where no input errors exist, le­gitimate differences in interpretation can lead to significantdifferences in simula tion resul ts, In some instances, we ran

The Basic Building Geometry

CanadaEnglandAustria

EnglandEngland and HollandWest GermanyUSA

EASI and ENERPASSBREDEMEBIWAN

ESP and HTB2SERIRESDEROBBLAST-3.0 and DOE2.IC

parametric studies to better understand the consequences ofdifferent modelling approaches. Some of these parametricsare discussed later in the paper.

The design tools were run voluntarily by groups interested inthe procedure. These were:

During the project, we sometimes observed relatively largeoutput differences between the "reference" codes. A seriousinvestigation of those differences would have been a largeproject and was beyond the primary mission of the workinggroup. As a group, we tried to weed out input-induced dif­ferences. Internal modelling differences were consideredpart of the legitimate output scatter. Where the referencecodes differed markedly, a large target range would exist forthe simplified design tools. Where the reference codesshowed close agreement, a tight target band would exist forthe design tools. In several instances, participating countrieswithdrew a reference code for debugging when an internalproblem was suspected. Thus, the procedure proved usefulfor identifying problems in the reference codes as well as thesimplified design tools.

The reference codes originally selected were:

!c...,

1

207m

.;~/.

,/,,/

~~~/1

6m

Fig. 1

TABLE 1. THERMO-PHYSICAL PROPERTIES

LIGHTWEIGHT HEAVYWEIGHT

Element Thickness R-Value Element Thickness R-Value(MM) (m2K/W) (MM) (m2K/W)

EXT. WALL 1.92 EXT. WALL 1.92Plasterboard 12 Plaster 16Batt 47 Cone. Block 100Cavity 29 Foam 50Plywood 9 Cavity 50Cavity 50 Brick 102Brick 105

INT. WALL INT. WALLPlasterboard 12 Plaster 16Cavity 50 Cone. Block 100Plasterboard 12 Plaster 16

FLOOR 25 FLOOR 25Timber 25 Screed 50Insulation 1003 Cone. Slab 150

Insulation 1000

ROOF 3.13 ROOF 3.13Plasterboard 10Batt 100Cavity 25 (Same as lightweight)Fiberboard 13Asphalt 19

WINDOW GLASS .37 WINDOW GLASS .37Double Pane Double Pane

TABLE 2. CHARACTERISTICS OF THE CASES

Set-PointsHeat Cool Glazs Infilt Intgen Base Other

Case II (OC) Mass (rn ) (ach) (W) Case

0 20 20 LW 0 S 0 0I 20 20 LW 0 S I 02 20 20 HW 0 S I 0

3 20 20 LW 9 S 1 04 20 20 HW 9 S I 07 20 27 LW 0 S 1 08 20 27 HW 0 S I 09 20 27 LW 9 S I 0

10 20 27 HW 9 S 1 011 FLOATING LW 9 S 1 012 FLOATING HW 9 S 1 0

1) 20 27 LW 4 S 200 914 20 27 HW 4 S 200 1015 20 27 LW 4 E 200 1316 20 27 HW 4 E 200 1417 20 27 HW 9 S 200 10 overhang Denver only18 FLOATING HW 9 S 200 10 overhang Denver only19 SET-BACK LW 4 S 200 1)

20 SET-BACK HW 4 S 200 1421 20 27 HW 9 S 200 10 adiabatic E,W,N walls

22 20 27 LW 9 S 200 9,1323 20 27 HW 9 S 200 10,1424 SET-BACK LW 9 S 200 1925 SET-BACK HW 9 S 200 2026 FLOATING LW 9 S 200 22 Denver only27 FLOATING HW 9 S 200 23 Denver only

Where: LW = LightweightHW = Heavyweight

Infilt = Infiltration in air changes per hourIntgen = Internally generated heat from lights, appliances, etc.

E,W,N,S = East, West, North, South20 20 = Heat on if temp <20°C

Coolon if temp> 20°C20 27 = Heat on if temp <20°C

Coolon if temp> 27°CFLOATING = Temperatures in building allowed to float freelySET-BACK = Set-back thermostat control trom 2300-0700 hrs.

Set-back temperature = lOoC(A blank cell contains the previous value in the column.)

RESULTS

Simulations were performed for each case using annual hourlystatistical weather data from Denver, Colo., and Copenhagen,Denmark. The winter Denver climate is generallv clear andcold with about 3600 degree days (·CDD), while that inCopenhagen is cloudy and cold with about 3800 ·CDD. It isnot possible to show all the results in the context of thisnecessarily short paper. Here, we show selected results con­centrating on heating in the Denver climate. For completeresults, please refer to the previously mentioned lEA report.

Figure 2 shows annual Denver heating loads obtained withfive detailed simulation programs for cases 0 through 10.This sequence of cases starts with an opaque box in whichsolid conduction is the dominant mode of heat transfer andfinishes with a passive solar example (I9% south glass-to-

floor area ratio) in which many dynamic heat transfer mech­anisms are simultaneously operative. For most of the cases,agreement is quite close. However, in case 4, SERlRES andESP represent high and low outliers. respectively, with a dis­agreement of about 40%. The difference between the maxi­mum and minimum result obtained for each case establishesthe target band for that case.

Figure 3 shows annual Denver heating loads for cases 13through 21. These cases are variants of a relatively realisticbuilding with such features as overhangs, east window, andnight-setback thermostat added. The "overhang" cases con­tain a 9-m~ window, while all other cases in this series con­tain a 4-10 Window. Most of the codes agree quite closely inthese cases.

Figure 4 shows annual Denver heating loads for cases 22

12'iil 11"tlc 1001(/)

:J 9

110

§. 8s: 7

1~

6fj III.><

II0>

~I.S 5

. 11111[[[111(;j

4ll>

11

~

3iii

I:'. I II : II:J 2 LI

c 1 JI!I,lrlllll:1c1 I:_<:

00 2 3 4 7 8 9 10

CJBLAST CJDOE OSERIRES r::-:!HTB [··,'·;ESP

Fig. 2. lEA Reference Codes Cases 0-10(Denver TMy)

E:]ESP

25

C.-:JESP

Denver

OHTB

24

C:lSERIRES

23

c: IDOE21C

12

11

10

987

6

; fiT 1'·[Thj••.·.i1 ..•••..o .1_. Lid .: jliffij

22

[',' BLAST3

Fig. 4. lEA Reference Codes Cases 22-25(Denver TMy)

12 ---, ---­

11

10

987

: 't/l~I')"3 , / '~' ":>:.2 ";-, ::::

~ .. ,_,~_" .:/.~.L:-,-' ....l:.-'C---,-,--,-,-,--,,-,,-.c...:...c.......:J

Copenhagen

r-~iBLAST ODOE f.::.1SERIRES

'iil"tlC01(/)

:Jo§.

(/)

"tlC01(/)

:Jo£s:~.><0>.Stoll>s:iii:JCC<:

19 20

c::HTB

13

CJBLAST

Fig. 3. lEA Reference Codes Cases 13-21(Denver TMY)

Fig• .5. lEA Reference Sodes Solar Gain(Annual gain through 9-m window)

through 25. These cases are simila'2 to those in th~previous

series except that the window is 9 m instead of 4 m •

Figure '2shows the annual solar energy transmitted throughthe 9-m south-facing window and absorbed in the building.Both the DOE and ESP codes contain anisotropic sky algo­rithms that tend to place more solar energy on the south thanthe isotropic algorithms in BLAST or SERIRES. However. theanisotropic model in ESP apoears to place significantly moresolar radiation on the south than does the anisotropic modelin the DOE code. This would explain the consistently lowerESP heating load predictions in figures 2. 3, and 4. This doesnot necessarily indicate a fault in ESP. No sky modellingalgorithm generally accepted as "correct" has yet beendeveloped.

Figure 6 shows annual heating load "target" ranges in hori­zontal black lines for cases 0 through lOin the Denver cli­mate as established from Figure 2. The target ranges weretaken Irom the maximum and minimum reference code pre­dictions for each case. Also displayed are the outputs fromthree design tools. The most obvious out-of-range predictionis from BREDEM in case 10. Cases 9 and 10 are the light andhea~ variants of a fairly realistic case with a south-lacing9-m winoow and a 20°-27°C banded thermostat controlstrategy. BREDEM shows no change in heating load predic­tion Irom the lightweight to the heavyweight case.

Figure 7 shows identical cases as those in figure 6 except thatthe climate is Copenhagen. Here, none 01 the design toolsare lar out 01 range. The significance to the designer is

obvious. Some tools will not be appropriate for certain appli­cations. In this example, BREDEM is not appropriate for pas­sive buildings in sunny climates. but it is quite adequate forcloudy climates. The significance to design tool developers isalso quite clear. Know the limitations 01 the method.Document major limitations prominently so as not to misleaddesigners.

Figure 8 sho....s the differences in heating loads predicted bythe reference programs lor the light and heavv buildings.These can be thought of as load savings due to the addition ofthermal mass. In some ways. the responsiveness of a designtool to parametric changes is more important than the magni­tude 01 the load prediction lor anyone case. The designerwants :0 kno.... what eflect design changes will have on thethermal perlormance of the building. Does the design toolshow the right trend by approximately the right amount? Ifwe added BREDEM to ligure S, it would be immediately ap­parent that this tool shows no response to changes in buildingcapacitance. This may not be a serious drawback for build­ings with limited window area or in predominantly cloudyclimates. However, in manv other instances. use of a toolthat is insensitive to thermal mass could mislead thedesigner.

Note in figure & that ESP, which has been somewhat lowerthan the other reference programs in the prediction of annualheatmg loads. agrees quite well with the other programs in itsresponse to thermal mass. This was also true for many of theother sensitivities studied, including set-back thermostat con­trol, shading, and window size.

Case 9­Case 10

CJHTB EJESP

Case 7­Case8

Case 3·Case 4

12

1110

987

6

54

321o ,.T.!·.;.'::I;''' -_.

Case 1­Case 2

Fig. 8. lEA Reference Codes Mass Response(Denver savings due to added mass)

E"l BLAST C.l DOE r:.'.lSERIRES

. f '.

SENSITIVITY STUDIES

12 Ien 11"0

~ 10<Il

is 92. 8

7

65

i I II f~ ,l ~ - ,1.1 I, ._ ,1.11... Ui 1 1.11

o 1 2 3 4 7 8 9 10

{:=JEBIWAN Cl BREDEM C.1 ENERPASS

Fig. 6. lEA Design Tool Examples (DenverTMY cases 0-10)

I i1r n ~ n n _. '-'~~-~1:1 -ig 11 r:-:n tI!!J -,.-.-.. I

~ ITsj!.

Ol.E

~ I

j ! LIII, II "II 111o 2 3 4 7 8 9 10

CZIEASI 0 EBIWAN Ci BREDEM C:.-! ENERPASS

Fig. 7. lEA Design Tool Examples (Copen­hagen TMY cases 0-10)

Figure 9 shows annual peak heating loads in the Denver cli­mate calculated using three of the reference codes. In allcases with a night set-back thermostat control strategy(19.20 and 24,2'), SERIRES predicted substantially greaterpeak loads than the other programs. Since the developmentof SERIRES was sponsored by the Solar Energy ResearchInstitute (SERIl, the SERI participant decided to investigatethe disagreements pertaining to cases 4, 19, 20, 24, and 25.

The SERIRES simulations had been performed by researchersin England and Holland. A careful review of the input filesfor those cases revealed no obvious input problems. However,an interpretive input issue was revealed concerning the inter­ior surface coefficients. The SERIRES Manual advises usingthe constant combined radiative and convective surface co­eff icients as defined in the ....SHRAf: Handbook of Fundamen­tals. The European researchers used the values from theCIBSE guide, which are essentially identical to those in theASHRAE Handbook. The specifications had called for thethermostat control to be based on the zone air temperature.The control temperature in SERIRES is the conductanceweighted average of the interior surface temperatures. Thus,when the normal ASHRAE combined coefficients are used,the control temperature in SERIRES is more like a radianttemperature. This would be similar in physical reality topainting the thermostat black versus silver. In actuality, realthermostats behave somewhere between these extremes, re­sponding to both the radiant and convective environment.

In BLAST and DOE, the thermostat control temperature ismodelled more like an air temperature, Both BLAST and

DOE calculate the interior surface radiation hourly based onalgorithms of varying sophistication. In most instances, thesesomewhat subtle distinctions have very little effect on ther­mal performance predictions. However, under certain condi­tions, the effect can be significant.

For the set-back cases, the standard combined coefficientused in SER1RES couples the building mass to the thermostatcontrol node more closely than in DOE or BLAST. This be­comes very noticeable in the calcula tion of peak loads, whichoccur at the "set-up" hour. The more closely coupled thebuilding mass to the control node, the harder the heating sys­tem has to work to bring the control temperature up to thenew thermostat setting. Since the heating systems weresized to always meet peak load, this shows up as a large dif­ference in the peak load.

To test this hypothesis, we reduced the interior surface coef­ficient input values in SERIRES by the amount normally asso­ciated with the radiative portion of the coefficient (about 2/3radiative to 1/3 convective). We increased the wall conduc­tance by the same amount to keep the overall building loadcoefficient identical.

Table 3 shows the influence of these changes on annual andpeak loads for case 25. As anticipated, the peak loads cal­culated with SERIRES are now close to those calcuiated withDOE and BLAST. The effect on annual loads is negligible.They remain close to those for DOE and BLAST, as in theoriginal run.

TABLE 3. RADIANT VS. CONVECTIVETHERMOSTAT CONTROLCase 25 Denver

Annual Annual Peak Peak(kWh) (kWh) (kW) (kW)

SERIRESor iginal 2212 1549 8.95 3.48

SERIRESmodified 2156 1416 6.45 2.67

BLAST3.0 2001 1291 6.27 2.59

DOE2.IC 2149 1463 6.14 2.69

ESP 1565 1435

For case 4, the source of the disagreement is the same, butthe effect is somewhat different. Here, it is the ability tostore transm itted solar energy tha t is affected by the valuechosen for the interior surface coeff icient. The large effectth is has on the annual hea ting and cooling loads is an artilactoC the unrealistically t ight thermosta t control strategy lorthat case. Since the control temperature cannot dril t , theheating and cooling system response is very sensitive to t!lecoupling between the control temperature and the buildingmass. When the standard comb ined interior surface coeff i­cient is used in SERIRES, the coupling is effectively t ighterthan in BLAST or DOE.. Thus, less solar energy is stored inthe mass increasing both heat ing and cooling loads. For morerealistic thermostat control strateg ies, this effect would bemuch less apparent. To test this theory, we ran case 4 withthe same input changes used in case 2.5.

Table 4 shows the influence of these changes on the case 4results. As expected, the heating and cooling load predictionsfrom SERIRES now resemble those Irom BLASTand DOE.

A similar sensitivity stud y was conducted by Bloomfield inEngland uSing the ESP model. ESP allows the user to specilythe degree to which the thermostat responds to mean radianttemperature versus "air" temperature. For cases I and 9 inthe Copenhagen climate, annua l heating load differences of1096 and 2896 were found, respect ively, when thermostat re­sponse was based on the air temperature versus a mixed radi­ant and convective temperature (2/3 radiant, 1/3 aid.

CONCLUSIONS

The work conducted to date demonstrates the feasibility oCusing detailed building energy ana lysis simulation programs toquant itatively evaluate more simplified des ign tools . Agree­ment among the detailed "re ference" programs was suff icientto establish reasonable target ranges aga inst which to com­pare the OUtput Crom simplilied des ign tools. Cases for whichagreement was not close indicate heat transfer phenomenathat are somewhat beyond the current modelling state of theart. Target bands were consequently wider , presenting a lesssevere test for simplified tools . This is legitimate. since thesimplified tools should not be expected to surpass detailedtools in modelling capability. The prototype design tool eva l­uation procedure was successCul in uncovering limitations inthe sample design tools tested. The procedure was also suc­cessful in illustrating the significance of algorithmic differ­ences in the reference programs.

The exact nature of the tests and the definit ion of whatconsti tutes passing a test remain to be defined. This willprobably be done within the con te xt of individual countries.engineering societies. or trade assoc iations. Several of theparticipants in this internat ional ellort have expressed the

TABLE 4. RADIANT VS. CONVECTIVETlIERMOSTAT CONTROL

Case 4 Denver

Annual AnnualHeat Cool

SERIRES original 5617 5839

SER1RES modified 4729 461 1

BLASTJ.O 4738 4242

DOE2.IC 4609 4754

HTB2 4956 4416

ESP 3915 4779

int ention 01 their respective governments to pursue domesticversions of the design tool evaluat ion procedure. In theUnited States, ASHRAE has formed a subcommittee to form­ulate standard procedures. It is hooec that this work will beuse lul as an ini tia l basis for those efforts.

It should be re iterated that failing a test does not necessar ilyindica te a Caultv too l. However. the tool oevelooer shouldunderstand why 'the tool disagrees and should warn users oCimport ant limitations in the tool. Conversely, passing all thete sts does not cernotetetv validate a too l. The tests represen ta lair Iy course Iilter, Simplilied ees ign tools should be passedthrough that /ilter as a minimum quality assurance proce­dure. The target ranges and types 01 tests will evolve as thesta te of the art in building energy anal ysis simulation prog­resses, and our understanding of building physics improves.The outputs of the "relerence" programs do not representtruth. They do. however, collect ively represent our best cur­rent Objective knowledge on the ca lculation 01 build ing ther­mal behavior.

Design tool evaluation procedures will provide pract it ionersII.-ith a ra tional basis for se lecting the appropriate tool lor agiven application.

REFERENCES

(I) Design Tool Survey, lEA Task VlIJ. P.R. Ritt lemann, S.F.Ahmed. Burt Hill Kosar RitUemann Assoc., Butler, PA.,USA. May 198'.(2) Design Tool Comparison lEA Task VIIl Working Document.P.R. Rittlemann, S.F. Ahmed. Burt Hill Kosar RittlemannAssoc., Butler, PA., USA.0) Simulation Model Validation Using Test Cell Data, lEATask VIII. Edited by Ove Mork. Techn ical University ofDenmark . June 1986.(4) Va lida t ion of Building Energy Analysis Simulat ion Pro­grams at the Solar Energ y Research Inst itute, R. Judkof},ENERGY AND BUILDINGS, Elsev ier Sequo ia S.A., Lausanne,Switze rland. Volume 10, 1985.r» International Energy Agency Building Simulation Compari­son and Validation Study, R. Judkoff. Proceedinl!s Interna­tiona l Climatic Architecture Congress, Louvarn-L a-Neuve ,BelgIum. July 1-3, 1986.

ACKNOWLEDGMENTS

This work was supported by the United States Department ofEnergy Solar Buildings Program. ft was cosponsored underTask VlIl of the International Energy Agency . We also wouidlike to thank Manfred Bruck, Austria; Stephen Carpenter,Canada; and Hans Erhorn , West Germany; and Don Alexanderand Peter Lewis, United Kingdom, for their contributions toth is pro iect,