rp-942

7/31/2019 rp-942

1/127

Loughborough University

Final Report

Comparison of Load Calculation ProceduresASHRAE/CIBSE 942-RP

J.D. Spitler

S.J. Rees

P. Haves

M.G. Davies

M. Holmes

25 February 1998

7/31/2019 rp-942

2/127

Final Report

Executive Summary............................................................................................... 1-1

Quantitative Comparison of U.S. and U.K. Cooling Load Calculation Procedures -Methodology, J.D. Spitler and S.J. Rees ................................................................ 2-1

Quantitative Comparison of U.S. and U.K. Cooling Load Calculation Procedures -

Results, S.J. Rees, J.D. Spitler, P. Haves ............................................................... 3-1

Qualitative Comparison of U.S. and U.K. Cooling Load Calculation Procedures, J.D.

Spitler and M.G. Davies ........................................................................................ 4-1

A Diagnostic Test Method for Cooling Load Calculation Procedures - BUILDTEST, S. J.

Rees, J.D. Spitler................................................................................................... 5-1

Appendices

On the Equivalence of the Radiant Time Series Method and the Transfer Function

Method, J.D. Spitler, D.E. Fisher........................................................................ A1-1

Comparative treatment of wall conduction and room internal heat transfer, M.G. Davies

.......................................................................................................................... A2-1

Notes on Modifications made to the BRE-ADMIT program, S.J. Rees ............... A3-1

Comparison of U.S. and U.K. Cooling Load Calculation Procedures to the Comit

Europen de Normalisation Draft Standard......................................................... A4-1

7/31/2019 rp-942

3/127

942-RP Final Report Executive Summary 1-1

Executive Summary

This document reports on the work performed under the first joint research project

between the American Society of Heating, Refrigerating, and Air-conditioning Engineers

(ASHRAE) and the Chartered Institution of Building Services Engineers (CIBSE),Comparison of Load Calculation Procedures.

Calculation of design cooling and heating loads is a key task in the design of HVAC

systems and has long been a subject of strong interest, both to ASHRAE and to its UK

sister organization, CIBSE. Both societies publish methods for calculating design cooling

and heating load calculations in their handbooks. However, each society has taken

somewhat different approaches to the cooling load calculation procedure. Although there

are some differences in climate and construction practices between North America and

Britain, the reason for the different approaches is probably just that each society

developed its methods in isolation. The result is that the methods that have been adopted

by each society have different technical bases. No work has been done to compare the

results obtained with the different methods and to explain the differences in these results in

terms of the underlying assumptions of the methods.

The increasing internationalization of the construction industry has resulted in an

increasing number of US companies working in Europe, and vice versa. US and European

companies are also competing for work in other parts of the world, such as the Far East.

In the longer term, both the efficiency and the reputation of the HVAC industry world-

wide would be improved if common methods of performing key design calculations were

adopted. An essential step in the process of adopting common methods, and a worthwhile

activity in its own right, is the comparison of existing methods and an understanding of the

practical consequences of their differences.

The chief aim of the project is to provide a technical basis for the harmonization of heating

and cooling load calculation methods used by ASHRAE and CIBSE. The objectives of

the project are:

To quantify the differences between the predictions of the various heating and cooling

load calculation procedures used by ASHRAE and CIBSE and emerging from CEN,

and explain these differences in terms of the differences in the methods employed.

To identify the method(s) that appear to offer sufficient accuracy, consistent with the

practical requirements of the user (availability of input data, range of applicability,

computational efficiency, transparency).

To identify any technical or other barriers to the harmonization of loads calculations in

the USA, the UK and Western Europe.

To provide a practical guide for the international practitioner who must deal with the

diverse methodologies in the interim.

7/31/2019 rp-942

4/127


In order to expedite the dissemination of results, the final report has been written as a

series of shorter papers. It is intended that each of the papers in the main body of the

report will be published in the ASHRAE Transactions, International Journal of HVAC&R

Research, or the Building Services Engineering Research and Technology journal. We

also hope to publish a synopsis of the results in the ASHRAE Journal and BuildingServices, the CIBSE Journal.

The first three papers present the core of the project, quantitative and qualitative

comparisons between the different cooling load calculation procedures.

The 1st and 2rd papers, Quantitative Comparison of U.S. and U.K. Cooling Load

Calculation Procedures - Methodology and Quantitative Comparison of U.S. and U.K.

Cooling Load Calculation Procedures - Results present the extensive quantitative

comparisons (load calculations for over 12,000 zones) made between the procedures. The

papers were submitted for publication in the ASHRAE Transactions.

The 3rd

paper, Qualitative Comparison of U.S. and U.K. Cooling Load CalculationProcedures, presents each of the three methods side-by-side in a consistent manner and

delineates the fundamental differences between the different load calculation procedures.

It is nearly complete, containing a fairly thorough description and comparison of the

methods. The Nomenclature section is not yet complete.

The 4th

paper,A Diagnostic Test Method for Cooling Load Calculation Procedures -

BUILDTEST, is not complete to the point of being ready for publication. However, it

contains the descriptions of some test cases that we found very useful as part of this

project. It could possibly be extended to be a very useful standard method of test for

cooling load calculation procedures. Interestingly, we pinpointed some bugs in one

program that been in use for some time. However, the bugs were not previously detected

when the program was validated against other programs using typical zone descriptions,

where the errors were washed out when the zones had a range of different heat gain types.

The first appendix contains an explanation that the PMSC requested be placed in the final

report: On the Equivalence of the Radiant Time Series Method and the Transfer Function

Method. The gist of the paper is that the RTS method can be shown to be mathematically

equivalent to the TFM under certain circumstances.

The second appendix is a review of US and UK load calculation approaches written by

M.G. Davies -- Comparative treatment of wall conduction and room internal heat

transfer. Some of the material is beyond the scope of the project but offers some

interesting insights into the state of load calculation procedures.The third appendix,Notes on Modifications made to the BRE-ADMIT program, specifies

the modifications made to the BRE-ADMIT program for use in the project. These include

both changes so that the program would run with input files, and some fixes to previously

undiscovered bugs.

The 4th

appendix, Comparison of U.S. and U.K. Cooling Load Calculation Procedures to

the Comit Europen de Normalisation Draft Standard, contains some material relevant

7/31/2019 rp-942

5/127


to the CEN standard. When the project was proposed, it was anticipated that the CEN

Standard would be ready for testing prior to the summer of 1997. Unfortunately, it is still

in a rather incomplete state. The draft standard contains some zone descriptions, for

which it is intended that users of the standard will perform load calculations and compare

their results to those published in the standard. While the standard currently contains zone

descriptions, no results are available for comparison. Accordingly, the appendix justcontains a brief description of the test zones, and a letter sent to one of the committee

members presenting our results for the first four test zones.

7/31/2019 rp-942

6/127

Quantitative Comparison of North American and U.K.Cooling Load Calculation Procedures - Methodology

J.D. Spitler, Ph.D., P.E., ASHRAE Member*

S.J. Rees, ASHRAE Student Member

Abstract

This paper describes the methodology used in a quantitative comparison between the

current North American and UK cooling load calculation methods. Three calculation

methods have been tested as part of a joint ASHRAE/CIBSE research project: the

ASHRAE Heat Balance Method and Radiant Time Series Method, and the Admittance

Method, used in the U.K. A companion paper (Rees et al. 1998) describes the results of

the study. The quantitative comparison is primarily organized as a parametric study -

each building zone / weather day combination compared may be thought of as a

combination of various parameters, e.g. exterior wall type, roof type, glazing area, etc.

This paper describes the overall organization of the study, the parameters and parameter

levels that can be varied, the tools developed to create input files, automate the load

calculations, and extract the results. A brief description of the cooling load calculation

procedure implementations is also given.

Keywords

Load calculations, cooling load calculations, building heat transfer.

*

J.D. Spitler is associate professor, School of Mechanical and Aerospace Engineering, Oklahoma State

University, Stillwater; S.J. Rees is Research Associate, Department of Civil and Building Engineering,

Loughborough University, England.

7/31/2019 rp-942

7/127

Quantitative Comparison of North American and U.K.Cooling Load Calculation Procedures - Methodology

Abstract

This paper describes the methodology used in a quantitative comparison between the

current North American and UK cooling load calculation methods. Three calculation

methods have been tested as part of a joint ASHRAE/CIBSE research project: the

ASHRAE Heat Balance Method and Radiant Time Series Method, and the Admittance

Method, used in the U.K. A companion paper (Rees and Spitler 1998) describes the

results of the study. The quantitative comparison is primarily organized as a parametric

study - each building zone / weather day combination compared may be thought of as a

combination of various parameters, e.g. exterior wall type, roof type, glazing area, etc.

Specifically, this paper describes overall organization of the study, the parameters and

parameter levels that can be varied, the tools developed to create input files, automate

the load calculations, and extract the results. A brief description of the cooling load

calculation procedure implementations is also given. The methodology presented, and

the tools described, could equally be used to make comparisons between other

calculation methods.

IntroductionAlthough ASHRAE has developed a number of cooling load calculation procedures over

the last twenty years, relatively little work has been published that makes comparisons

between different North American cooling load calculation procedures. Similarly, no

comparison between North American and other cooling load calculation procedures have

been reported. A brief overview of published inter-method comparisons follows.

Hill and Furlong (1973) gave a qualitative comparison of the Total Equivalent

Temperature Difference / Time Averaging method (TETD/TA) and the Transfer Function

Method (TFM) to the heat balance method as implemented in NBSLD, an early energy

calculation program. For a single example zone, they presented a quantitative

comparison.

Rudoy and Robins (1977) reported a comparison between cooling load calculations

performed with the ASHRAE RP-138 method and the NBSLD program. A single zone

of fixed size, but with two different roof types, three different wall types, two different

glazing percentages, two different internal load densities, four different locations, and four

different zone orientations, for a total of 384 combinations, was used for comparison.

7/31/2019 rp-942

8/127

942-RP Final Report Paper 1 2-2

Shah (1983) qualitatively compared the TETD/TA methods and the Cooling Load

Temperature Difference / Cooling Load Factor (CLTD/CLF) method. He attempted to

explain possible reasons why the TETD/TA method results vary from those of the

CLTD/CLF method. However, no head-to-head comparisons were reported.

Sowell (1988) reported on a comparison between four different computer programs, eachof which could be used to compute the steady-periodic response to a sinusoidal heat gain.

A total of 20 different zone types covering a range of zone response were used for the

comparison.

Spitler, McQuiston and Lindsey (1993) reported on the additional error caused by using

printed tables of Cooling Load Temperature Differences (CLTD), Cooling Load Factors

(CLF) and Solar Cooling Loads (SCL) as opposed to using the custom-generated

coefficients or the Transfer Function Method. Deviations from the Heat Balance

procedure were not estimated.

Only two of the inter-method comparisons described above compared results fromsimplified methods to the Heat Balance procedure. Only Rudoy and Robbins (1977)

attempted to make comparisons for a range of zone types.

In order to perform a detailed quantitative comparison between the ASHRAE Heat

Balance procedure (Pedersen, Fisher and Liesen 1997), the ASHRAE Radiant Time Series

procedure (Spitler, Fisher and Pedersen 1997), and the CIBSE Admittance procedure

(CIBSE, 1986), a parametric study was undertaken. This involved:

collecting data for a range of wall, roof, ceiling, floor, and window constructions;

defining eighteen different parameters that describe the different constructions, internalheat gains, zone dimensions, and weather data;

developing a scheme for keeping track of all the possible combinations;

developing software tools that generate input files automatically for variouscombinations of parameters, run the load calculation procedures, and extract and

summarize the results.

This paper describes the above steps. A companion paper (Rees, Spitler and Haves 1998)

describes the analysis of the results and the conclusions.

Organization of Parametric Study

In order to study the differences between the different load calculation methods, results

for a wide range of sample zones, with different construction types, internal heat gain

densities and schedules, weather days, etc. were compared. Each design cooling load test

case in the study was defined by a combination of 18 parameters, each of which has an

associated value or parameter level. In this particular study the parameters were things like

wall constructions, % window glazing, ACH infiltration, etc. Some of the parameters are

discrete, like wall constructionlevel 1 is one specific construction, level 2 is another.

7/31/2019 rp-942

9/127


Other parameters are continuous and can have any range of values, like percentage

window glazing or East-West dimension of the zone. However, in this study, we only

allow certain values, e.g. the percentage of exterior wall that is glazed can only be integer

percentages between 0% and 99%. Infiltration can only have 10 different values of ACH:

0, 0.2, 0.4, 1.8.

Since literally thousands of design load calculations were performed, it was necessary to

have a systematic procedure for keeping track of all the different parameter combinations.

Accordingly, a coding scheme was developed so that by looking at the alpha-numeric code

- which was used to name the input and output files - the corresponding combination of

parameter values that were used can be determined. This code, with the addition of a

suitable file extension, was used to name the files associated with each test case. For

example when input files for the heat balance and RTS procedures are created, they have

the form code.inp and output files are named code.out". Similarly, the admittance

method uses code.adi for input files and code.ado for output files. The code is a 23-

character string that has the form: RlzwwpplseqixprcfwtARsw, where each of the letters

has a meaning, as given in Table 1. This scheme can easily be extended to allow theinclusion of other calculation methods.

Not every possible parameter can be varied within the scheme, e.g. zone height is kept

fixed at 3 m. However, with the chosen parameters and associated parameter levels, it is

possible to create over 51018 combinations which would take over 325 billion years torun on the 166 MHz 586 machine being used for this project. To make the parametric

study tractable a number of test case sets were defined using a subset of the total possible

variants. These were formed by choosing a particular zone construction and a single

location but varying other parameters. In this way four test case sets of 1248 test cases

were defined and which used zone constructions that could be identified as `lightweight',

`U.S. Mediumweight', `U.K. Mediumweight' and `Heavyweight' (in order of increasingthermal mass). The particular parameter levels used in these test case sets are given in the

companion paper (Rees, Spitler and Haves 1998). Each of these test case sets took

approximately six hours execution time.

Further discussion of the individual parameters is found below, in the Zone Geometry and

Construction, Zone Fabrics, Internal Heat Gains and Schedules, and Design Day sections.

7/31/2019 rp-942

10/127


Table 1: Code definition for code RlzwwpplseqixprcfwtARsw

Code

segment

Meaning

r Room size letter. Indicates the length of the room in the East-West direction, in

meters. (i.e., the length of the North or South walls) A= 6 m, B = 9 m, C= 12 m,D=15 m, E=18 m, F= 21 m, G=24 m, H=27 m, I= 30 m, J=3 m.

l Room level. t is the top floor, meaning zone has an exposed roof.

floor, meaning the zone has a ceiling, with a conditioned zone above.

z z is the zone number, which indicates where the zone is located on a given floor,

and hence which walls are exposed exterior walls, and which are partition walls

with conditioned zones on the other side. See Figure 1 for a diagram. Zone 1 is a

southwest corner zone; zone 4 is an east-facing zone; zone 9 is an interior zone.

ww ww is the two digit percentage of exterior wall area that is glazed. It can range

from 00 to 99.

pp pp is the two digit people per 100 /m2of floor area. This represents the peak

number of people in the zone. Each hour it is multiplied by the corresponding

fraction from the people schedule.

ls ls is the two digit peak internal heat gain due to lighting in W/m2 of floor area.

Each hour it is multiplied by the corresponding fraction from the lighting schedule.

eq eq is the two digit peak internal heat gain due to equipment in W/m2 of floor area.

Each hour it is multiplied by the corresponding fraction from the equipment

schedule.

i i represents the infiltration in ACH, multiplied by 5,

e.g. i=9 represents 1.8 ACH

x x represents the exterior wall type (0-9), defined below.

p p represents the partition type (0-9), defined below.

r r represents the roof type (0-9), defined below. Roof type is only meaningful for

top floor zonesc c represents the ceiling type (0-9), defined below. Ceiling type is only meaningful

for mid-floor zones

f f represents the floor type(0-9) , defined below

w w represents the window type, defined below.

t t represents the internal thermal mass, defined below.

AR AR is the two digit representation of the aspect ratio of the zone multiplied by 10.

The aspect ratio is the North-South dimension of the zone divided by the East-West

dimension. AR can have a value between 01 and 99 - creating an aspect ratio

between 0.1 and 9.9

s s (0-9) defines the lights, equipment, and people schedules. Each schedule can be

set differently, but only a total of 10 sets of schedules are allowed.d d (0-9) defines the weather day, location, day of the year.

7/31/2019 rp-942

11/127


1 2 3

4

567

8 9

N

Figure 1: Zone numbering scheme

Parametric Levels

The following sections describe the different parametric levels (values) of each parameter.

Zone Geometry and Construction

The zone geometry is controlled by the room size and aspect ratio parameters. A wide

range of zone sizes can be created from 3 m x 0.3 m to 30 m x 297 m using different

combinations of the room size and aspect ratio parametersall zones being 3 m high.

The construction of the zone from fabric types (described in the next section) is controlled

by the room level, zone number and window percentage parameters. The room level

parameter determines whether the zone has an exposed roof (and hence uses a roof type)

or has another conditioned zone above it (and hence uses a ceiling type). Since groundcoupling is not included in the study, ground level floors are not used. All zones use the

floor type to describe the floor, and it is assumed a conditioned zone is below the zone

being modeled.

The zone number parameter determines the zone location, as shown in Figure 1, and hence

which walls are exterior walls (constructed from the exterior wall type and exposed to the

outside) and which walls are partitions (constructed from the partition type and have

another conditioned zone opposite.) Note that although Figure 1 represents the individual

zones as having an aspect ratio of 1, it can vary between 0.1 and 9.9. Hence, the zones

can also be long rectangles oriented in either the N-S direction or the E-W direction.

The exterior walls have windows of the specified window type that cover a percentage of

the wall area specified by the window percentage parameter.

7/31/2019 rp-942

12/127


Zone Fabric Types

A range of fabric types have been selected following consultation with practitioners to try

and establish some specific constructions from real projects that cover a wide range in

thermal mass. Information has been received in the form of architectural drawings andnumerical data from Hallam Associates Inc. (Wilkins 1996), Arup R&D, part of the Ove

Arup Partnership (Holmes 1996) and Troup Bywaters and Anders (Arnold 1996). From

this information a number of wall fabric types have been selected.

Table 2 summarizes the general characteristics of each fabric type. In addition to the U-

value, a measurement of the fabrics overall thermal mass, the thermal capacity per unit

area, dC p has been calculated, where d is the layer thickness, is the material density,and Cp is its specific heat capacity. The thermal mass of the wall fabrics selected cover the

range 32-550 kJ/m2K (1.7-27 Btu/ft

2F). To indicate the position and relative size of the

insulation layer in the wall fabric the thicknesses are normalized and presented as outside

thickness (%) : insulation thickness (%) : inside thickness (%) in Table 2.

Table 2: Fabric Characteristics -- Summary

Fabric

Thermal Capacity per

Unit Area

kJ/m2.K Btu/ft

2.F

U-Value

W/m2.K Btu/h.ft

2.F

Insulation

Distribution (%)

Outside : Ins. : Inside

ext. wall type 1 32 1.6 0.23 0.041 21 : 72 : 7

ext. wall type 2 148 7.3 0.23 0.041 48 : 47 : 5

ext. wall type 3 268 13.1 0.495 0.087 34 : 21 : 45

ext. wall type 4 361 17.7 0.45 0.079 29 : 23 : 48

ext. wall type 5 521 25.5 1.98 0.35 100 : 0 : 0

ext. wall type 6 551 27.0 0.52 0.092 42 : 10 : 48part. wall type 1 25 1.2 0.35 0.062 10 : 80 : 10

part. wall type 2 209 10.2 1.06 0.19 0 : 0 : 100

floor/ceil. type 1 32 1.6 1.97 0.35 0:100:0

floor/ceil. type 2 223 10.9 1.77 0.31 100 : 0 : 0

floor/ceil type 3 537 26.3 2.83 0.49 100 : 0 : 0

floor/ceil type 4 586 28.7 1.43 0.25 100 : 0 : 0

roof type 1 34 1.7 0.22 0.039 1 : 13 : 86

roof type 2 527 25.8 0.63 0.11 10 : 23 : 67

The wall types are specified in detail in Table 3, with the layers specified from the outside

to the inside. The source of information for each type is noted in brackets in the title ofeach table (where none is given the information has come from the researchers). Exterior

wall constructions selected for the project are specified in Table 3. They span the range

from very lightweight (type 1) to extremely heavy (type 6).

7/31/2019 rp-942

13/127


Table 3: Exterior wall constructions

Layer

Material

Thickness

(mm)

Thickness

(inches)

Kg/m3

CpkJ/kg.K

k

W/m.K

EXTERIOR WALL TYPE 1: LIGHTWEIGHT TIMBER CLAD (Holmes 1996)

cedar wood planks 15 0.59 400 1.63 0.11

air gap 19 0.79 1.2 1.005 *

ply wood 9 0.35 540 1.21 0.12

insulation 150 6 32 0.71 0.04

vapor barrier 1 0.04 1860 0.84 0.35

plaster board & skim 13 0.5 800 1.09 0.16

EXTERIOR WALL TYPE 2: BRICK & STUD INNER LEAF (Wilkins 1996)

facing brick 92 3.625 1600 0.79 0.84

air gap 48 1.875 1.2 1.005 *

gypsum sheathing 16 0.625 800 1.09 0.16

insulation (R-19) 150 6.0 32 0.71 0.04

gypsum wall board 16 0.625 800 1.09 0.16

EXTERIOR WALL TYPE 3: BRICK-BLOCKWORK CAVITY WALL WITH

INSULATION (Arnold 1996)

facing brick 75 3 1700 0.92 0.55

air gap 50 2 1.2 1.005 *

insulation 75 3 300 1.0 0.067

concrete block 150 6 950 1.06 0.2

plaster 15 0.55 1570 0.84 0.53

EXTERIOR WALL TYPE 4: GRANITE FACED CONCRETE (Holmes 1996)

granite panel 40 1.5 1600 0.79 1.1

air gap 50 1.9 1.2 1.005 *

insulation 70 2.7 32 0.71 0.04

cast concrete 150 5.9 2300 0.9 2.15

EXTERIOR WALL TYPE 5: SOLID BRICK/BLOCK UNINSULATED (Wilkins 1996)

facing brick 100 4 1600 0.79 0.84solid concrete block 200 8 2100 0.92 1.63

plaster 13 0.5 720 0.84 0.16

EXTERIOR WALL TYPE 6: HEAVYWEIGHT BLOCKWORK & CAVITY INSULATION

(Arup R&D)

facing brick 100 4 1600 0.79 0.84

air gap 100 4 1.2 1.005 *

insulation 50 2 32 0.71 0.04

solid concrete block 215 8.5 2100 0.92 1.63

plaster 13 0.5 720 0.84 0.16

* All air gaps have been given a constant resistance of 0.18 m2.K/W

7/31/2019 rp-942

14/127


Two partition wall types were selected for the project, as shown in Table 4, a lightweight

partition typical of commercial construction (type 1) and a heavyweight construction (type

2).

Table 4: Partition wall types

LayerMaterial

Thickness(mm)

Thickness(inches)

Kg/m

3 CpkJ/kg.Kk

W/m.K

PARTITION WALL TYPE 1: STUD WALL INTERNAL PARTITION


insulation 100 4 32 0.71 0.04


PARTITION WALL TYPE 2: BLOCKWORK INTERNAL PARTITION

plaster 13 0.5 720 0.84 0.16

concrete block 100 4 2100 0.92 1.63

plaster 13 0.5 720 0.84 0.16

Floors and ceilings are described in Table 5, from the lower layer to the upper layer. It is

not expected that very many commercial buildings will have low-mass wood floors.However, it is desirable to have a low-mass floor as an option, so it is included as

floor/ceiling type 1. The other floor/ceiling combinations all have thermally massive

concrete layers and represent the three combinations of mass exposed to the zone below,

mass exposed to both zones, and mass exposed to the zone above.

Table 5: Floor and Ceiling Types

Layer

Material

Thickness

(mm)

Thickness

(inches)

Kg/m3

CpkJ/kg.K

k

W/m.K

FLOOR/CEILING TYPE 1: WOOD FLOOR WITH GYPSUM BOARD CEILING

gypsum wall board 13 0.5 800 1.09 0.16air gap 190.5 7.5 1.2 1.005 *

pine 20 0.79 640 1.63 0.15

FLOOR/CEILING TYPE 2: METAL DECKING AND RAISED FLOOR (Hallam Inc.)

steel pan 2 0.08 7689 0.42 45

cast concrete 100 4 2300 0.9 1.73

air gap 150 6 1.2 1.005 *

insulated floor tile 40 1.6 100 1.2 0.6

carpet tile 8 0.32 400 1.38 0.1

FLOOR/CEILING TYPE 3: IN-SITU CONCRETE SLAB & TILE FINISH

cast concrete 200 8 2300 0.9 1.73

screed 70 2.75 1920 0.88 1.4

vinyl tiles 5 0.2 800 1.26 0.6FLOOR/CEILING TYPE 4: IN-SITU CONCRETE SLAB, SUSPENDED CEILING, TILE

FINISH FLOOR

ceiling tile 10 0.4 370 0.59 0.06

ceiling air space 1000 39 1.2 1.005 *

cast concrete 200 8 2300 0.9 1.73

screed 70 2.75 1920 0.88 1.4

vinyl tiles 5 0.2 800 1.26 0.6

7/31/2019 rp-942

15/127


Two roof types were selected for the project, as shown in Table 6. Roof type 1 is a very

lightweight roof with almost no thermal mass. Roof type 2 has a substantial amount of

thermal mass 150 mm (6) of concrete.

Table 6: Roof Types

Layer

Material

Thickness

(mm)

Thickness

(inches)

Kg/m3

CpkJ/kg.K

k

W/m.K

ROOF TYPE 1: STEEL DECKING INSULATED (Hallam Inc.)

membrane 10 0.4 1121 1.67 0.19

insulation 150 6 32 1.21 0.04

steel pan 2 0.08 7689 0.42 45

ceiling air space 1000 39 1.2 1.005 *

ceiling tile 10 0.4 370 0.59 0.06

ROOF TYPE 2: CONCRETE SLAB INSULATED

stone chippings 13 0.5 881 1.67 1.436

felt & membrane 10 0.4 1121 1.67 0.19

insulation 50 2 40 0.92 0.025cast concrete 150 6 2300 0.9 1.73

7/31/2019 rp-942

16/127


Two glazing types have been selected to represent minimum and maximum cases of solar

heat gain. Window properties were computed using the Window 4 (LBNL 1994)

program from LBL. Standard library components were used to create the windows.

Table 7: Window Types

GLAZING TYPE 1: SINGLE PANE CLEAR GLASS,

ALUMINUM FRAME

Layer

Material

Thickness

(mm)

Coating

Clear glass 6 none

U-Value 6.0

Shading coefficient 0.90

Solar heat gain coefficient 0.78

Normal solar transmittance 0.74

Normal solar absorptance 0.154

Inside emissivity 0.84Outside emissivity 0.84

Surface-to-surface thermal

conductance

150.

GLAZING TYPE 2: DOUBLE GLAZED, TINTED,

LOW-E COATING, ALUMINUM FRAME WITH

THERMAL BREAK

Layer

Material

Thickness

(mm)

Coating

Grey glass 6 none

Argon filled

gap

12.7

Clear glass 6 Low-EU-Value 2.7

Shading coefficient 0.43

Solar heat gain coefficient 0.37

Normal solar transmittance 0.30

Normal solar absorptance 0.636

Inside emissivity 0.10

Outside emissivity 0.84

Surface-to-surface thermal

conductance

4.23

7/31/2019 rp-942

17/127


Presumably, most zones of interest contain internal furnishings such as chairs, tables, filing

cabinets, office equipment, retail goods, etc. Even in sophisticated building simulation

programs, these are seldom modeled in any detail, as it is too much to expect the user to

input and it is likely to change many times over the building life. Accordingly, interior

thermal mass is usually modeled as one or more surfaces that store energy and exchange

radiation and convection with the rest of the zone. For purposes of this study, thermalmass types have been defined in Table 8 as a given material with an area proportional to

the zone floor area. The material properties are given in Table 9.

Table 8: Thermal Mass Types

Thermal mass

type

Material Ratio of

thermal mass

area to zone

floor area

0 - 0

1 25mm (1) pine 25%

2 25mm (1) pine 50%

3 25mm (1) pine 100%

4 25mm (1) pine 200%

5 25mm (1) pine 400%

6 50mm (2) face brick 25%




Table 9: Thermal Mass Materials

Layer

Material

Thickness

(mm)

Thickness

(inches)

Kg/m3

CpkJ/kg.K

k

W/m.K

THERMAL MASS TYPES 1-5: 25MM (1) PINE

pine 25 1 640 1.63 0.15

THERMAL MASS TYPES 6-9: 50MM (2) Brick

brick 50 2 1700.00 0.92 0.55

7/31/2019 rp-942

18/127


Internal Heat Gains and Schedules

Three different types of internal heat gains are implemented people, equipment, and

lighting. Each heat gain is specified with a peak heat gain rate and a schedule of 24 hourly

fractions that multiply the peak heat gain rate. Different schedules can be specified foreach heat gain type, but only up to 10 combinations of schedules may be specified. For

purposes of this study, 7 different schedule types have been defined, as described in Table

10. Although different schedule fractions may be specified for every hour, currently only

two different values are used for each schedule - one which applies from 8 a.m. - 5 p.m.,

and one which applies the rest of the day.

Table 10: Schedule types

Schedule

type

People Lights Equipment

8 a.m.-5 p.m. 5 p.m.-8 a.m. 8 a.m.-5 p.m. 5 p.m.-8 a.m. 8 a.m.-5 p.m. 5 p.m.-8 a.m.0 0 0 0 0 0 0

1 1 0 1 0 1 0

2 1 0 1 0.1 1 0.1

3 1 0 1 0.35 1 0.35

4 1 0 1 0.7 1 0.7

5 1 0 1 1 1 1

6 1 1 1 1 1 1

The hourly heat gain for people is set using the pp parameter. The hourly sensible heat

gain due to people is:

qpeople= pp*Azone/100* 75.4 W /person * hourly people schedule fraction (1)

where Azone is the area of the zone in square meters.

The hourly heat gain for lights is set using the ls parameter. The hourly sensible heat

gain due to lights is:

qlights=ls*Azone* hourly lights schedule fraction (2)


The hourly heat gain for equipment is set using the eq parameter. The hourly sensible

heat gain due to equipment is:

qequipment=eq*Azone* hourly equipment schedule fraction (3)


7/31/2019 rp-942

19/127


Design Days

Although it was not anticipated that the choice of different design days (with different

temperatures, latitudes, and solar radiation) will cause a substantial difference in the

relative results between different methods, it is desirable to be able to change these and seewhat happens. Design day weather data is chosen from the ASHRAE Handbook of

Fundamentals (1993). The locations utilized are London, Chicago, Denver, Miami, and

Phoenix. Each location uses the 1% peak design temperature listed in the ASHRAE

Handbook, along with the mean daily range. In order to give two sets of solar data, days

0-4 occur on June 21, while days 5-9 occur on September 21. The latitude and longitude

are as given in the ASHRAE Handbook.

Table 11: Design days

Weather /Location

#

Location Month Peak designtemperature

(C/F)

Mean dailyrange

(C/F)

0 London 6 23 (73.4) 9 (16.2)

1 Chicago OHare AP 6 33 (91.4) 11 (19.8)

2 Denver AP 6 34 (93.2) 16 (28.8)

3 Miami AP 6 33 (91.4) 8 (14.4)

4 Phoenix AP 6 43 (109.4) 15 (27)

5 London 9 23 (73.4) 9 (16.2)

6 Chicago OHare AP 9 33 (91.4) 11 (19.8)

7 Denver AP 9 34 (93.2) 16 (28.8)8 Miami AP 9 33 (91.4) 8 (14.4)

9 Phoenix AP 9 43 (109.4) 15 (27)

Load Calculation Procedure Implementations

The load calculation procedures being studied, the Heat Balance method, the Radiant

Time Series method, and the Admittance method will be discussed in a companion paper

(Spitler, Davies and Rees 1998). However, it is necessary here to at least discuss the

implementations in a general sense. In the case of the heat balance method and the radianttime series method, a stand-alone computer program (Pedersen and Associates, 1997)

developed as part of ASHRAE 875-RP was used to perform both procedures. The

computer program, as previously developed, operated with input and output from files.

File input/output was a prerequisite for this project, as it would have been impossible to

reliably enter data for thousands of different cases into a graphical user interface.

7/31/2019 rp-942

20/127

7/31/2019 rp-942

21/127


Figure 2: Preprocessor User Interface

Parametric run generator

The parametric run generator program is a substantially modified version of a program

written by Strand (1996). It takes as its input a data file created by the preprocessor

specifying which parameter levels are to be used for each parameter and the type of

parametric study to be done. Three different types of parametric studies can be

performed:

1. A fully-populated parametric study where every possible combination of the

parameter levels is used to create an input file. For example, for a case with 6 exterior

wall types, 2 partition types, and 9 window percentages, 6x2x9=84 input files would becreated for each method.

2. A sparsely-populated parametric study where a base case is specified and only one

parameter is varied at a time. This is very convenient when looking for trends caused

by individual parameters.

3. A min/max parametric study is a fully-populated parametric study where there are

only two levels for each parameter, a high and low level.

7/31/2019 rp-942

22/127


The parametric run generator creates both a complete set of input files and a batch file to

run each case with all three programs.

The parametric run generator program is written in Fortran 90. Briefly described, the

program uses the information from the input file to determine all of the cases that are to be

created. For each case, it puts together a series of small files containing parts of the inputfiles. Some of the small files are used for every input file; others are specific to a certain

parameter level, e.g. exterior wall type. In addition, some lines in each input file have

values that must be computed, e.g. wall and window dimensions, and in these cases entire

lines of the input file are written by the program directly.

Postprocessors

A fairly simple program, written in Fortran 90, is used to extract the results from each

method, for each case. It writes the run code and the 24 hourly cooling loads for each of

the three methods to a comma-delimited file. A spreadsheet program reads the comma-

delimited file.

The spreadsheet program allows functions written in a high level language to be integrated

into the spreadsheet. Functions that find the peak load and time of peak load for each

method are used to summarize the results. Functions that can decode the run code are

used to show what the individual parameter levels are for each case.

Conclusions

This paper has presented the methodology used in a large parametric comparative study

between cooling load calculation methods published in North America and the United

Kingdom. The results and conclusions of the study are published in another paper, by

Rees, et al. (1998). However, the project methodology has several unique features:

With the existing parameter levels, it is theoretically possible to create over 5x1018

unique zone types.

While the theoretical limit may not be reached due to computer speed and storagelimitations, it is still quite possible to create and run three different load calculation

methodologies on several thousand zone types over-night.

Using the existing postprocessor, it is possible to summarize the results for severalthousand zone types, and quickly determine which parameter combination was

responsible for any single result. The methodology used in the study can easily be extended to include comparisons of

other calculation methods.

7/31/2019 rp-942

23/127


Acknowledgments

This work was carried out under a joint ASHRAE and CIBSE funded project

Comparison of Load Calculation Procedures (ASHRAE 942-RP, CIBSE 22/95). The

authors would like to thank C. Wilkins (Hallam Inc.), D. Arnold (Troup Bywaters &Anders) and M.J. Holmes (Ove Arup R&D) for their help in providing details of building

fabrics. Thanks are also given to M.J. Holmes and M.G. Davies (Liverpool University) for

their technical advice given during the project.

References

Arnold, 1996. Personal Communication.

Bloomfield, D.P. Undated.BRE-ADMIT: Thermal design of buildings. Watford: BRE

Publishing, Building Research Establishment.

CIBSE. 1986. Guide Book A Design Data, London: Chartered Institution of Building

Services Engineers

Hill, J.E., R.R. Furlong. 1973. ASHRAE Cooling Load Calculations. ASHRAE Journal.

15(5):61-66.

Holmes, 1996. Personal Communication.

LBNL 1994. WINDOW 4.1 Computer Program, Windows & Daylighting Group, Building

Technologies Program, Energy & Environment Division, Lawrence Berkeley NationalLaboratory.

Pedersen, C.O., Fisher, D.E., and Liesen, R.J. 1997. Development of a heat balance

procedure for calculating cooling loads.ASHRAE Transactions 103(2):

Rees, S.J., J.D. Spitler, P. Haves. 1998. Quantitative Comparison of North American and

U.K. Cooling Load Calculation Procedures Results. Submitted for publication to

ASHRAE Transactions.

Rudoy, W., L.M. Robins. 1977. Comparison of Cooling Load Calculations. ASHRAE

Transactions. 83(1):38-50.

Shah, D.J. 1983. ASHRAE Cooling Load Calculation Methods.ASHRAE Journal.

25(11): 50-56.

Sowell, E.F. 1988. Cross-check and Modification of the DOE-2 Program for Calculation

of Zone Weighting Factors. ASHRAE Transactions 94(2):737-753.

7/31/2019 rp-942

24/127


Spitler, J.D., Davies M.G., and Rees, S.J. 1998. Qualitative comparison of North

American and UK cooling load calculation procedures. Submitted for publication to

ASHRAE Transactions.

Spitler, J.D., Fisher, D.E. and Pedersen, C.O., 1997. The radiant time series cooling load

calculation procedure. ASHRAE Transactions 103(2):

Spitler, J. D., F. C. McQuiston and K. L. Lindsey. 1993. The CLTD/SCL/CLF Cooling

Load Calculation Method. ASHRAE Transactions, 99(1):183-192.

Strand, R. 1996. Parahbss.f90 computer program. Personal Communication.

Wilkins, 1996. Personal Communication.

7/31/2019 rp-942

25/127

Quantitative Comparison of North American and U.K.Cooling Load Calculation Procedures Results

S.J.Rees, ASHRAE Student Member*

J.D.Spitler, Ph.D., P.E., ASHRAE Member

P.Haves, Ph.D., C.E., ASHRAE Member

Abstract

Calculation of design cooling loads is of critical concern to designers of HVAC systems.

The work reported here has been carried out under a joint ASHRAE/CIBSE research

project to compare design cooling calculation methods. Three calculation methods have

been tested, the ASHRAE Heat Balance Method and Radiant Time Series Method, and

the Admittance Method, used in the U.K. The results presented show the general trends in

over/under prediction of peak load in the simplified methods compared to the Heat

Balance Method. The performance of the simplified methods is explained in terms ofsome of the underlying assumptions in the methods, and by reference to specific

examples.

Keywords

Load calculations, cooling load calculations, building heat transfer.

*

S.J. Rees is Research Associate, P. Haves is senior lecturer, Department of Civil and Building

Engineering, Loughborough University, England; J.D. Spitler is associate professor, School of Mechanical

and Aerospace Engineering, Oklahoma State University, Stillwater.

7/31/2019 rp-942

26/127

Quantitative Comparison of North American and U.K.Cooling Load Calculation Procedures Results

AbstractCalculation of design cooling loads is of critical concern to designers of HVAC

systems. The work reported here has been carried out under a joint ASHRAE/CIBSE

research project to compare design cooling calculation methods. Three calculation

methods have been tested, the ASHRAE Heat Balance Method and Radiant Time

Series Method, and the Admittance Method, used in the U.K. The results presented

show the general trends in over/under prediction of peak load in the simplified

methods compared to the Heat Balance Method. The performance of the simplified

methods is explained in terms of some of the underlying assumptions in the methods,

and by reference to specific examples.

Introduction

ASHRAE and its UK sister organization, the Chartered Institution of Building

Services Engineers (CIBSE) have published methods for calculating design coolingand heating load calculations in their handbooks for many years. Each organisation,

working largely independently, has developed a somewhat different approach to

design cooling load calculation procedures. There is increasing internationalization ofthe construction industry and in the longer term both the efficiency and the reputation

of the HVAC industry world-wide would be improved if common methods of

performing key design calculations were adopted. The aim of the work reported herehas been to compare the results obtained with the different methods and to explain the

differences in these results in terms of the underlying assumptions of the methods.

At present, both ASHRAE and CIBSE are re-evaluating and revising their loadcalculation procedures. In addition, the Comit Europen de Normalisation (CEN),

the standards-making organization that includes all the major countries of Western

Europe, including the UK, is in the process of developing a standard approach to loadcalculations. The draft CEN standard takes the form of a specification consisting of a

set of heat balance equations and a set of qualification tests against which particular

computer codes can be evaluated.

ASHRAE has a long history of developing and revising load calculation methods.

Romine (1992) gives a good summary through to 1992. At present, ASHRAErecommends three methods in the Handbook of Fundamentals (ASHRAE 1997) andthe Cooling and Heating Load Calculation Manual (McQuiston and Spitler 1992): the

Transfer Function Method (TFM), the Cooling Load Temperature Difference/ Solar

Cooling Load /Cooling Load Factor (CLTD/SCL/CLF) method and the TotalEquivalent Temperature Difference /Time Averaging (TETD/TA) method. Each

method attempts to approximate the results of the Heat Balance Method, either

directly, or indirectly.

7/31/2019 rp-942

27/127


More recently, ASHRAE has funded a research project entitled Advanced Methodsfor Calculating Peak Cooling Loads (875-RP). The goal of this project has been to

replace the existing methods with two newmethods: the Heat Balance Method

(Pedersen, Fisher and Liesen 1997) and the Radiant Time Series Method(Spitler,

Fisher and Pedersen 1997). The Heat Balance Method is the most fundamental of alldesign load calculation methods and may be the method most understandable by

practising engineers, as it closely follows physical processes and has a minimum of

mathematical abstraction. However, it does require the solution of severalsimultaneous equations. The second method, the Radiant Time Series Method, is

intended to be simpler from a calculation standpoint and builds on the concepts of the

TFM and TETD/TA method.

CIBSE is currently in the process of revising the sections of its Guide (CIBSE 1986)

that relate to load calculation procedures (Holmes and Wilson 1996). In the current

CIBSE Guide (1986) two load calculation methods are described. These are theEnvironmental Temperature Nodal method, which deals with steady state loads, and

the Admittance Method, which deals with fluctuating loads. A number of models of

differing complexities are proposed in the draft revision. Two dynamic methods are

proposed, one based on a detailed reference model and another based on a simplifiedmodel (which is, in fact, the Admittance Method). The reference model consists of a

performance specification, along with a list of features that must be included.Particular model equations or calculation methods are not specified, although it is

difficult to see how the requirements could be met other than by a method based on

explicit heat balances. To date, no public domain design cooling load calculationcomputer codes have been developed that comply with the CIBSE draft standard. It

appears then that ASHRAE and CIBSE are adopting similar approaches.

This paper describes the work carried out under a project jointly funded by ASHRAEand CIBSE: Comparison of Load Calculation Procedures(ASHRAE 942-RP). In

this project comparisons have been made between the peak cooling load predictionsof the new ASHRAE Heat Balance Method (Pedersen, Fisher and Liesen 1997) andtwo simplified methods, the ASHRAE Radiant Time Series (RTS) Method (Spitler,

Fisher and Pedersen 1997), and one implementation of the Admittance Method

(Danter 1983). The calculation procedure along with full descriptions of theparameters, e.g. construction properties, is described in a companion paper (Spitler

and Rees 1998). A systematic comparison of the predictions of the methods, along

with an analysis of the sensitivity of the various parameters is given here. The causes

of the variations in predictions between the different methods are identified, andexplained in terms of the approximations made in the simplified methods.

The Calculation MethodsThree design cooling load calculation methods have been compared in this work.

These are the 'Heat Balance' (HB) method, the 'Radiant Time Series' (RTS) method

and the 'Admittance Method'. The main features of the three methods can be aresummarised below. (A systematic comparison of the methods is given in Spitler,

Davies and Rees 1998.)

The Heat Balance Method involves the solution of heat balance equations for each ofthe outside and inside zone surfaces, along with the zone air. This approach is similar

7/31/2019 rp-942

28/127


to that of existing load and energy calculation codes such as TARP (Walton 1983) andBLAST (1986). Radiant and convective heat exchange are treated separately at both

inside and outside surfaces, with interior radiant exchange being calculated using the

Mean Radiant Temperature/Balance algorithm of Walton (1980). Transient

conduction through the zone fabric is dealt with using conduction transfer functions.The two simpler methods combine radiation and convection heat transfer into a single

equivalent resistance.

The Radiant Time Series Method uses a two-stage calculation procedure. First,

convective and radiant heat gains are calculated for each hour assuming a constant

zone air temperature. Second, the resulting cooling loads are calculated. The methodmodels exterior convection, long-wave radiation and absorbed solar radiation using a

sol-air temperature and combined, constant, radiant/convective surface conductances.

Transient conduction is calculated using a series of response factors that are used with

the hourly outside sol-air temperatures and a fixed zone air temperature as theirboundary conditions. The radiant heat gains are converted to cooling loads using a set

of zone response factors (the so called Radiant Time Series) which define how much

of the radiant load at a particular hour becomes a cooling load on the zone air at future

hours.

In contrast to the U.S. cooling load calculation procedures, the Admittance Methodrelies in its derivation on analytical techniques that assume the boundary conditions

(outdoor temperature, solar radiation etc.) fluctuate sinusoidally with a period of 24

hours. Accordingly, the Admittance Method is a two-stage calculation procedure, inwhich the mean and fluctuating components of the loads and temperatures are

calculated separately. The mean components are calculated using the CIBSE

simplified steady state model that is defined by a three node thermal network. The

Admittance Proceduredefines how the fluctuating components of the loads andtemperature differences are calculated.

Whereas the U.S. methods generally use the zone air temperature node as the point atwhich internal surfaces are convectively coupled, the Admittance Method relies on

the concept of environmental temperature which is used to calculate the combined

radiant and convective heat exchange with the room surfaces. The concept ofenvironmental temperature is similar to that of sol-air temperature used to define

external surface heat transfer in that a combined radiant and convective conductance

is used. All the zone surfaces are linked to a common environmental temperature node

at which a heat balance is calculated. The derivation of the environmental temperaturemodel has in fact been criticised since its introduction (e.g. Davies 1992). In the

Admittance Method, transient conduction heat transfer through the wall is modelled

with a frequency-response derived decrement factor and time lag. Although thederivation is based on frequency-response and a sinusoidal driving function with a 24

hour period, the decrement factor and time lag are used in the same way as the

decrement factor and time lag were used in the Total Equivalent TemperatureDifference / Time Averaging (TETD/TA) Method (ASHRAE 1993).

The Heat Balance Method, being the most detailed of the three methods, and beingbased, to a greater extent, on fundamental physical principles, has been used as a

reference model in this work, against which the two simplified models have been

compared.

7/31/2019 rp-942

29/127


Implementations of the Calculation Methods

In choosing the computer implementations to use in this comparison exercise onlyimplementations that were readily available and where there was access to the source

code were considered. Carrying out the large number of runs necessary for the study

also made it a requirement that the implementation could be easily adapted to run

from standard input files. Fortran 90 implementations of the Heat Balance and RTSMethods produced for the 875-RP (Pedersen, Fisher and Liesen 1997; Spitler, Fisher

and Pedersen 1997) project were used. One executable in fact runs both the Heat

Balance and RTS Methods in turn from the same input data file.

There are a number of commercial implementations of the Admittance Method but

there is only one well-known code in the public domain: BRE-ADMIT (Bloomfield1983). This is a set of programs written in BASIC that provide a user interface,

generate solar data, calculate admittances, etc. from fabric thermal properties and

perform the actual load/temperature calculation. The latter program (BRECALC) hasbeen adapted to run from file input and send output to a file. The calculation of

cooling loads in BRE-ADMIT is set out by Danter (1983). It should be noted that theBRE-ADMIT code deviates from the Admittance Method as defined in the Guide,

particularly with respect to calculation of solar gains (this is discussed further inSpitler, Davies and Rees 1998).

The CIBSE Guide (1986) prescribes two methods of calculating solar gains using theAdmittance Method:

1. If an overheating calculation1 is required, the total incident radiation is dividedinto its mean and fluctuating components. A Solar Gain Factor and an Alternating

Solar Gain Factor then multiply the mean and fluctuating components. Thesefactors are constant and are defined for energy transfer to both the air and

environmental temperature points. The solar gain is then given by multiplying theglazing area by the incident irradiation by the appropriate Solar Gain Factor. Thealternating component is shifted in time by a lag associated with the Alternating

Solar Gain Factor. These solar gain factors are tabulated in the Guide for various

window/blind types in heavy and lightweight buildings located in London.

2. If a peak cooling load is required then tabulated loads due to solar gains in either atypical heavyweight or lightweight zone are given in the Section A9 of the Guide.These tabulated loads have been calculated using what is otherwise the admittance

model, but with a detailed model of solar transmission through and absorption by

glazing. This is done for various latitudes and window/shading combinations. The

exact calculation method is described in (Holmes and Wilson 1996).

The BRE-ADMIT implementation of the Admittance Method deviates from the

procedure defined in the guide in that it uses the Solar Gain Factors in the calculation

of zone cooling loads, as well as floating internal temperatures.

In the Admittance Method, internal gains are normally added as loads at the

environmental temperature node. This implies that the internal gain is 2/3 radiant and

1 The Admittance Method was originally developed for calculation of internal temperatures in

unconditioned buildings. In this case, zone temperatures may float freely.

7/31/2019 rp-942

30/127


1/3 convective. The BRE-ADMIT code however, attempts to deal with this in a moreexact way by allowing the user to specify the radiant-convective split of all internal

gains. These gains are then apportioned to the air and environmental temperature

points accordingly. It can be argued however, that the radiant portion of the internal

gains should interact with the thermal mass of the zone as determined by the zonesurface factors and the associated time lags.

Preparation of the Test Data

The computer implementations of the three load calculation procedures inevitably

have different input data requirements. Simplified methods generally require a lessdetailed description of the building and loads in their input data. One possible source

of difference in the results of the load calculations could be different interpretations of

the building zone definitions in the input data to each code. Some variables in the

building model may be under the control of the user in one method while being out ofthe users control in another implementation. In preparing the test data for each of the

calculation method implementations, the aim has been to ensure that the data

normally under the control of the user is consistent between the methods.

In the case of the fundamental properties of the building fabric and the zone

geometry, the automatic input file generation methodology set out in the companionpaper (Spitler and Rees 1998) ensures a high degree of quality control over the input

data. Input data quality control is also helped by the fact that the Heat Balance and

RTS implementations use the same input file. The principle adopted in the case ofother model variables is that where the variable is intended to be an input to the

model, we have chosen values equivalent to those in the other methods. Where no

user control is intended (although some may be possible by customising the input

files) the default value built into the implementation has been accepted. Specificdetails of how different input data were treated are given in Appendix A.

Parametric Test Results

The aim of the parametric study was to make comparisons between the peak loadpredictions of the calculation methods over a wide range of building zone types,

firstly to identify any general trends in the load predictions and also to identify

sensitivity to particular parameters. Each test zone in the study was characterised by a

set of twenty-three parameters. These define the zone size, notional orientation/position in a building, fabric construction, windows, internal loads, internal thermal

mass and weather day. Further details of the parameters, zone construction, along

with the methods used to compile the input data and run the calculations is discussedthe companion paper (Spitler and Rees 1998).

It is not feasible to run a set of calculations where every possible parameter levelpermutation is used - that would require many millions of calculations. To identify

general trends in the calculation comparisons we have however, defined limited sets

of test cases (1296 test cases in each set) in which most parameters are varied over thefull range for a particular zone construction. These test case sets are referred to below

7/31/2019 rp-942

31/127


as fully populated2. The extremes in building thermal mass are represented by two

sets of test cases based on the lightweight and heavyweight fabric types, as shown in

Table 1.

To evaluate the performance of the calculation methods with constructions in betweenthe extremes of thermal mass, two further sets of test cases have been used. The zone

construction in these two sets of test cases is intended to be typical of U.S. and U.K.

mediumweight commercial office buildings. The U.K. mediumweight constructionthat has been defined is slightly heavier than for the U.S..

The zone construction parameters for each of the four test sets are shown in Table 1.The fabric properties corresponding to the particular construction type and parameter

level are given in detail in the companion paper (Spitler and Rees 1998). Each of the

other parameters defining the test case sets are either set at one level, or varied over a

range as defined in Table 2.

Fabricelement

Lightweight USMediumweight

UKMediumweight

Heavyweight

External

Wall

cedar wood planks

air gap, plywood

insulation

vapour barrier

plaster board &

skim

facing brick

(92 mm / 3.625")

air gap

gypsum sheathing

insulation (R-19)

gypsum wall board

facing brick

(75 mm / 3")

air gap

insulation

concrete block

plaster

facing brick

(100 mm / 4")

air gap

insulation

solid concrete block

plaster

Internal

Wall

gypsum wall board

insulation

gypsum wall board

gypsum wall board

insulation

gypsum wall board

gypsum wall board

insulation

gypsum wall board

Plaster

concrete block

(100 mm / 4")

plaster

Floor &

Ceiling

gypsum wall board

air gappine

steel pan

cast concrete(100 mm / 4")

air gap

insulated floor tile

carpet tile

ceiling tile

ceiling air spacecast concrete

(200 mm / 8")

screed

vinyl tiles

ceiling tile

ceiling air spacecast concrete

(200 mm / 8")

screed

vinyl tiles

Roof membrane

insulation

steel pan

ceiling air space

ceiling tile

membrane

insulation

steel pan

ceiling air space

ceiling tile

stone chippings

felt & membrane

insulation

cast concrete

(150 mm /6")

stone chippings

felt & membrane

insulation

cast concrete

(150 mm /6")

Table 1: A summary of the zone constructions (listed by layer from outside to inside)

for the four test cases of the main parametric study.

2

The parametric studies exist in a multi-dimensional space, with each dimension corresponding to a

single parameter. The parameters may be varied in a number of different ways. For purposes of

describing the ways in which the parameters might be varied for any particular set of test cases, an

analogy to a two-dimensional matrix is utilized. A fully populatedset would represent all

combinations of the levels of each parameter. A sparsely populatedset would represent a test set

where only a single parameter is varied at a time. A min-maxtest set would include all combinations

of the extreme values of each parameter in two dimensional form, the analogous matrix would be

populated only at the corners.

7/31/2019 rp-942

32/127


Parameter No. of

Levels

Parameter Levels

Room Level 2 m, t (middle and top floors)

Zone Number 9 1 9 (all zone orientations)

Window type 2 Types 1 & 2 (single & double glazed)

Aspect Ratio 3 0.5, 1.0, 2.0

Weather Day 2 0, 4 (London and Phoenix, June 21)

Load Schedule 2 1, 5 (on all day, stepped schedule)

% Glazing 3 10, 50, 90 %

Thermal Mass 1 Type 1 (25% floor area, pine)

Infiltration (ACH) 1 1.0 Air Change per Hour (ACH)

People (per 100m2) 1 10 ( 1 person per 10m

2)

Equipment (W/m2) 1 30 (W/m

2)

Lighting (W/m2) 1 20 (W/m

2)

Table 2: The parameter ranges used in the main parametric study.

To illustrate the trends in the peak cooling load predictions for each of the test sets,peak cooling loads given by the RTS and Admittance Methods have been shown

plotted against the Heat Balance Method prediction for each case. In this way

predictions by the simplified methods that are in exact agreement with the HeatBalance Method are shown by a point lying on the diagonal line in the graphs points

above the line represent overprediction of the peak load. The results for the

Lightweight, Heavyweight and both Mediumweight test case sets are shown inFigures 1-4.

Figure 1: Peak load comparisons for the lightweight fully populatedtest case set.RTS Method (left) and Admittance Method (right) vs. the Heat Balance Method.

0

10 0

20 0

30 0

40 0

50 0

60 0

70 0

0 100 200 300 400 500 600 700

Heat Balance Peak Load (W/m2)

RTSPea

kLoa

d(W/m

2)

0

100

200

300

400

500

600

700

0 100 200 300 400 500 600 700


Adm

ittance

Pea

kLoa

d(W/m

2)

7/31/2019 rp-942

33/127


Figure 2: Peak load comparisons for the U.S. Mediumweight fully populatedtest

case set. RTS Method (left) and Admittance Method (right) vs. the Heat Balance

Method.

Figure 3: Peak Load comparisons for the U.K. Mediumweight fully populatedtest

case set. RTS Method (left) and Admittance Method (right) vs. the Heat BalanceMethod.

0

10 0

20 0

30 0

40 0

50 0

60 0

70 0

0 100 200 300 400 500 600 700


RTSPea

kLoa

d(W/m

2)

0

50

10 0

15 0

20 0

25 0

30 0

35 0

40 0

45 0

50 0

0 100 200 300 400 500


RTSPea

kLoa

d(W/m

2)

0

50

100

150

200

250

300

350

400

450

500

0 100 200 300 400 500


Adm

ittance

Pea

kLoa

d(W/m

2)

0

100

200

300

400

500

600

700

0 100 200 300 400 500 600 700


AdmittancePeakLoad

(W/m2)

7/31/2019 rp-942

34/127


Figure 4: Peak Load comparisons for the Heavyweight fully populatedtest case set.

RTS Method (left) and Admittance Method (right) vs. the Heat Balance Method.

The errors for the peak load predictions for the simplified methods relative to the Heat

Balance Method for these cases have also been analysed numerically. The mean

errors in peak load along with the minimum and maximum errors (i.e. worst cases ofunder and over prediction) have been listed in Table 3.

RTS Method BRE-ADMITParametric Test

Case Set Mean Min. Max. Mean Min. Max.

Lightweight 3.55 -0.79 28.32 -8.03 -31.51 8.13

US Mediumweight 2.60 -2.83 29.32 -7.59 -30.49 9.36

UK Mediumweight 6.61 -0.08 43.36 3.51 -12.48 35.96

Heavyweight 5.06 -0.78 37.56 1.38 -13.76 35.58

Min-Max 6.94 -6.45 43.94 -4.96 -39.73 28.76

Table 3: Summary of the percent errors in peak load prediction for the simplifiedmethods for the parametric test case sets.

The results for the RTS Method for the four test case sets show similar trends. Thevast bulk of the RTS predicted peak loads are greater than the corresponding Heat

Balance Method predictions. A small amount of peak load underprediction is shown

in the densely populated moderate load regions of the graphs for a limited number oftest cases. The RTS Method could be said to perform better for the lightweight and

US Mediumweight test cases (see Table 4) as the mean error and the maximum errorsare lower for these test case sets. There is more of a tendency for the RTS Method to

over predict the peak load in the cases with high cooling loads.

There is a greater variation in results for the Admittance Method calculations over the

four test case sets. For the lightweight and the U.S. mediumweight cases theAdmittance Method results show general underprediction of the loads, and are spread

in a wider band than the corresponding RTS Method results. Higher load cases for the

lightweight zones show correspondingly high degrees of underprediction. The

0

10 0

20 0

30 0

40 0

50 0

0 100 200 300 400 500


RTSPea

kLoa

d(W/m

2)

0

100

200

300

400

500

0 100 200 300 400 500


Adm

ittance

Pea

kL

oa

d(W/m

2)

7/31/2019 rp-942

35/127


Admittance Method results for the U.K. mediumweight and Heavyweight test casesets show a rather different trend with many of the results showing overprediction of

the peak load. These heavier weight test cases show a mean overprediction of peak

load.

One way of illustrating the extremities of the performance envelope of the calculation

methods is to use a series of tests using only the maximum and minimum values of

the parameters. This has been done for a test series with 2048 cases and the results areplotted for the RTS and Admittance Method together in Figure 5. At higher loads

certain clusters of results can be seen, representing similar parameter combinations. In

many cases it can be said that the Admittance Method underpredicts the peak load insituations where the RTS Method overpredicts the load.

Figure 5: Peak Load comparisons for 2048 min-maxcases where only minimumand maximum values of the parameters are used to define the test cases.

Parameter Sensitivity

In order to analyse the trends in peak load prediction of the calculation methods inmore detail and to check the sensitivity to particular parameters a second series of

parametric tests were made. These test cases have been termed Sparsetests in that

the parameter matrix is sparsely populated. In this type of test zone base casesare

defined that are typical of the lightweight, U.S. mediumweight, U.K. mediumweightand Heavyweight classifications used previously. Further test cases were generated by

changing one parameter of the base case at a time through a number of levels (giving

95 tests in each set). The parameter levels of the base cases are given in Table 4.

0

500

1000

1500

2000

2500

3000

0 500 1000 1500 2000 2500 3000


PeakLoad(W

/m2)

RTS Peak

Admittance Peak

7/31/2019 rp-942

36/127


Parameter Lightweight US Mediumweight UK Mediumweight Heavyweight

Zone size 6 x 6 x 3m 6 x 6 x 3m 6 x 6 x 3m 6 x 6 x 3m

Zone Level Top Top Top Top

Zone

orientation

South facing South facing South facing South facing

Glazed area 10 % 10 % 10 % 10 %People 10 per 100m

210 per 100m

210 per 100m

210 per 100m

2

Lighting 20 W/m2

20 W/m2

20 W/m2

20 W/m2

Equipment 30 W/m2

30 W/m2

30 W/m2

30 W/m2

Infiltration 1.0 ach 1.0 ach 1.0 ach 1.0 ach

External Wall

type

cedar wood

planks

air gap, plywood

insulation

vapour barrier

plaster board &

skim

facing brick

(92 mm / 3.625")

air gap

gypsum sheathing

insulation (R-19)

gypsum wall board

facing brick

(75 mm / 3")

air gap

insulation

concrete block

plaster

facing brick

(100 mm / 4")

air gap

insulation

solid concrete

block

plaster

Internal Wall

type

gypsum wall

boardinsulation

gypsum wall

board

gypsum wall board

insulationgypsum wall board

gypsum wall board

insulationgypsum wall board

plaster

concrete block(100 mm / 4")

plaster

Floor &

Ceiling type

gypsum wall

board

air gap

pine

steel pan

cast concrete

(100 mm / 4")

air gap

insulated floor tile

carpet tile

ceiling tile

ceiling air space

cast concrete

(200 mm / 8")

screed

vinyl tiles

ceiling tile

ceiling air space

cast concrete

(200 mm / 8")

screed

vinyl tiles

Roof type membrane

insulation

steel pan

ceiling air spaceceiling tile

membrane

insulation

steel pan

ceiling air spaceceiling tile

stone chippings

felt & membrane

insulation

cast concrete(150 mm /6")

stone chippings

felt & membrane

insulation

cast concrete(150 mm /6")

Window type double glazed double glazed double glazed double glazed

Thermal mass

type

25% floor area,

pine (25mm /1")

25% floor area,

pine (25mm /1")

25% floor area,

pine (25mm /1")

25% floor area,

pine (25mm /1")

Load schedule On 8 am - 5 pm On 8 am - 5 pm On 8 am - 5 pm On 8 am - 5 pm

Weather day London Phoenix London London

Aspect ratio 1.0 1.0 1.0 1.0

Table 4: The parameters for the base cases in the parameter sensitivity study.

The results of this parameter sensitivity study have been presented graphically and by

tabulating the range in the peak load percentage error when a particular parameter isvaried over its full range. For example, if the RTS Method over predicts the peak load

by between 5% and 23% when a single parameter is varied, the range in percentageerror is then 23%-5%, or 18%. The range of the percentage error is given for each

parameter, calculation method, and sparse test case series in Table 5.

7/31/2019 rp-942

37/127


Lightweight US

Mediumweight

UK

Mediumweight

HeavyweightParameter

RTS Admit. RTS Admit. RTS Admit. RTS Admit.

Zone size 1.44 1.30 1.24 2.25 5.14 3.56 5.37 2.05

Aspect ratio 0.35 0.71 0.61 0.75 3.38 5.12 2.83 0.81

Zone orientation 0.76 3.19 0.71 4.04 1.90 2.42 1.40 1.35

% Glazing 0.93 6.46 0.2 1.68 3.83 0.82 2.39 0.25

Persons 0.28 0.24 0.3 1.25 0.11 0.64 0.25 0.66

Lighting load 0.57 0.89 0.61 3.27 0.61 0.72 0.58 1.47Equipment load 0.41 3.76 0.36 1.04 1.19 1.91 0.11 2.74

Infiltration 0.01 1.20 0.45 3.35 1.03 1.13 0.43 0.87

External wall type 0.83 1.69 0.56 1.23 1.16 2.06 1.03 1.41

Internal wall type 1.08 0.72 1.08 0.72 1.08 0.72 1.08 0.72

Roof type 0.96 0.25 0.96 0.25 0.96 0.25 0.96 0.25

Floor type 1.95 5.11 1.13 2.63 1.43 3.88 1.44 2.63

Window type 1.50 2.70 0.08 0.94 1.81 3.73 1.64 4.94

Thermal mass 0.17 1.44 0.12 1.25 0.25 4.54 0.64 0.94

Load schedule 2.42 10.98 1.21 6.88 0.81 6.34 2.43 7.49

Weather day 0.75 10.11 0.55 7.81 1.47 8.6 0.72 9.37

Table 5: The percentage error range for each parameter and sparse test case set.

Results from the sparse test case series show the same general trends as the main

parametric studies. The RTS Method results always show an overprediction of peakload. The Admittance Method results show underprediction in the lighter weight cases

and better agreement with the Heat Balance Method (but still underpredicting) in the

heavier weight cases.

Figure 6: The effects of changing zone size (left) and changing zone aspect ratio

(right) for both simplified methods in the sparse U.K. mediumweight tests. (Note

suppressed zeroes.)

Nearly all the parameters show a range in the percentage error of less than 2% in the

RTS Method results (Table 5). The parameters that cause the error to change by more

than 2% for the RTS Method are the zone size, aspect ratio, percentage glazing andload schedule. Zone size and aspect ratio are inter-related parameters in that both have

40

45

50

55

60

65

70

40 45 50 55 60 65 70


Pea

kLoa

d(W/m

2)

RTS Peak

Admittance Peak

40

45

50

55

60

65

70

40 45 50 55 60 65 70


Pea

kLoa

d(W/m

2)

RTS Peak

Admittance Peak

7/31/2019 rp-942

38/127


the effect of changing the wall and window area in relation to the floor area. Theeffects of changing zone size and aspect ratio are shown in Figure 6.

Figure 7: The effects of changing Load Schedule in the lightweight (left) andheavyweight (right) sparse test case sets. (Note suppressed zeroes.)

Figure 8: The effects of changing weather day (left) and percent glazing (right) for

the lightweight sparse test cases.

The results from the parameter sensitivity tests for the BRE-ADMIT code show more

sensitivity than the RTS Method for nearly all parameters. Sensitivity to the load

schedule and weather day is quite marked, as illustrated in Figures 7 and 8respectively. The lightweight cases show particular sensitivity to increases in

percentage glazing. This is illustrated in Figure 8. Some sensitivity to floor type is

also evident in Table 5 (as in the RTS Method results, but to a lesser extent).

60

65

70

75

80

85

90

95

100

105

60 70 80 90 100 110


PeakLoad(W/m

2)

RTS Peak

Admittance Peak

0

20

40

60

80

100

120

140

0 20 40 60 80 100 120 140


Pea

kLoa

d(W/m

2)

RTS Peak

Admittance Peak

60

62

64

66

68

70

72

74

76

78

80

60 65 70 75 80


PeakLoad(W/m2)

RTS Peak

Admittance Peak

50

52

54

56

58

60

62

64

66

68

70

50 55 60 65 70


PeakLoad(W/m2)

RTS Peak

Admittance Peak

7/31/2019 rp-942

39/127


Analysis of the Admittance Method Performance

In order to study the ability of the simplified models to model particular aspects of

building heat transfer, a series of special test cases were constructed. In each test case,only one heat transfer path or type of heat gain is exercised to facilitate the diagnosis

of particular weaknesses of the models. For the BRE-ADMIT implementation of the

Admittance Method this showed that two mechanisms appear to be responsible for the

general underprediction of peak loads, the simplified way in which solar gainsthrough glazing are calculated, and the treatment of the radiant components of internal

gains.

The sky models of the Heat Balance/RTS implementationas noted in Appendix

A give slightly higher incident solar irradiances than that of the BRE-ADMIT code.

This may account for approximately 2% of the difference between the results for theparticular weather days used. However, there are several reasons why the modelling

of solar gains in the BRE-ADMIT code may cause greater discrepancies in the peak

load predictions when compared to the Heat Balance Method. The significance ofsolar gains in the errors of the Admittance Method calculations can be illustrated by

considering the zones with the greatest errors in load prediction from the heavyweightand lightweight test sets. Figure 9 shows calculations for these zones with different

amounts of glazing.

Figure 9: The response to increasing amounts of glazing for the Admittance Method

worst case lightweight (left) and heavyweight (right) zones.

The Solar Gain Factorsused in the BRE-ADMIT implementation of the Admittance

Method define the proportion of the mean and fluctuating part of the incident solarirradiation that becomes a load at the environmental point of the model. Solar gain

factors are tabulated in the CIBSE Guide (1986) and have been derived for South-

west facing light and heavyweight rooms in a London location for variouswindow/shade combinations. This method of calculating the load due to transmission

and absorption of solar radiation through the glazing can be seen to be simplistic for

the following reasons:

75

100

125

150

175

200

75 100 125 150 175 200


Pea

kL

oa

d(W/m

2)

RTS Peak

Admittance

50

75

100

125

150

175

200

225

250

275

50 75 100 125 150 175 200 225 250 275


Pea

kL

oa

d(W/m

2)

RTS Peak

Admittance

7/31/2019 rp-942

40/127


The proportion of solar irradiation that becomes a load in the room is assumed tobe a fixed and so no account is taken of the effect of different angles of incidenceof the solar beam either at each hour of the day or at different latitudes. (The latter

problem could be addressed by CIBSE publishing solar gain factor data for a

range of latitudes other than London).

The response of a building zone to transmitted and absorbed solar irradiationdepends heavily on its thermal mass. Relying on only two sets of factors derivedfor a typical lightweight and a typical heavy weight zone means that zones of

different construction can not be adequately represented.

The Solar Gain Factors in the CIBSE Guide (1986) have been calculated forSouthwest facing windows. At the peak hour the factors define the proportions ofthe mean and fluctuating parts of the solar gains. The incident solar irradiation on

south facing surfaces differs somewhat from that on east and west facing surfaces

when averaged over 24 hours (see Figure A-1). The same factors applied to

windows exposed to solar irradiation from other directions will not thereforereproduce the correct peak load.

The second reason identified for the general underprediction of peak loads by theAdmittance Method is due to the treatment of the radiant component of internal gains.

The convective component of internal gains appears instantaneously as a system load

when the system load is calculated at the air temperature node. (Tests have confirmedthat all the implementations model this feature correctly.) The radiant component of

internal loads however, interacts with the zone thermal mass. The radiation is

absorbed at the internal surfaces and subsequently is released to the room air byconvection. In the implementation of the Admittance Method tested here, the radiant

energy is redistributed in time but only by changing the average value of the internal

gain. The interaction between the radiant energy from internal sources and the fabric

is modelled simplistically. This is illustrated in Figure 10, which shows the response

of typical light and heavyweight zones to an internal, 50% radiant, heat gain. TheAdmittance Method results show no interaction between the radiant heat gain and the

zone thermal mass, resulting in an underprediction of the peak load for bothlightweight and heavyweight cases.

Figure 10: The response to stepped internal loads with 50% radiant component for

lightweight (left) and hea

Date post:	05-Apr-2018
Category:	Documents
Upload:	asmaa-ramadan
View:	219 times
Download:	0 times

rp-942

Documents