This paper has been downloaded from the Building and Environmental Thermal Systems

Research Group at Oklahoma State University (www.hvac.okstate.edu)

The correct citation for the paper is:
Chantrasrisalai, C., D.E. Fisher, I.S. Iu, and D. Eldridge. 2003. "Experimental Validation
of Design Cooling Load Procedures: The Heat Balance Method", ASHRAE Transactions.
109(2):160-173.
Reprinted by permission from ASHRAE Transactions (Vol. #109, Part 2, pp. 160-173).
2003 American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.

Copyright 2003, American Society of Heating, Refrigerating and Air-Conditioning

Engineers, Inc. (www .ashrae.org). Reprinted by permission from ASHRAE
Transactions 2003, Volume 109, Part 2. This paper may not be copied nor
distributed in either paper or digital form without ASHRAE's permission .

4646 (RP-1117)

Experimental Validation of Design Cooling

Load Procedures : The Heat Balance Method
Chanvit Chantrasrisalai
Student Member ASHRAE

Ipseng 1u

Student Member ASHRAE

ABSTRACT
Under an ASHRAE-sponsored research project (RP1117), two office-sized test cells having identical geometry but
different thermal mass were constructed in order to experimentally validate the performance of two new procedures for
design cooling load calculations: the heat balance method
(HBM) and radiant time series method (RTSM) . In this paper,
the HBM is compared with measured cooling loads . The validation of the RTSM is presented in a companion paper (Iu et
al. 2003). The performance of the experimental facility is
presented in a third paper (Eldridge et al. 2003) .

The heat balance method coded in the ASHRAE Loads

Toolkit was used in this study. Cooling loads predicted by the
HBM deviatedfrom measured data by less than 4%for base
case and drop ceiling configurations . The HBMoverpredicted
the cooling by more than 10% for the carpeted zone, but the
addition of thermal mass and furniture did not significantly
affect the cooling load. Although the effect of blinds could be
estimated by varying the thermal parameters, the ASHRAE
Loads Toolkit does not support the modeling of blinds . The
resultsfrom this study demonstrate that the HBMisfundamentally reliable, but that additional studies are required in order
to determine the significance of input parameters for the
carpeted zone and additional models are required in order to
estimate the effect of blinds on the cooling load.

INTRODUCTION
Building thermal load calculations underpin the design of
all indoor environmental systems. The accuracy of the thermal
load calculation procedure is, therefore, of concern to heating,
ventilating, and air-conditioning (HVAC) engineers, who
design HVAC systems and/or perform building energy analy-

D.E . Fisher, P.E., Ph .D.

Member ASHRAE

David S. Eldridge
Student Member ASHRAE

sis. Recently, two new procedures for design cooling load

calculations were developed by ASHRAE (Pedersen et al .
1997, 1998). They are the heat balance method (HBM) and the
radiant time series method (RTSM) .
Although the HBMhas been used as a thermal load model
in building energy simulation programs for over two decades
(BSO 1991), it has rarely been used as a design cooling load
calculation method. This was primarily due to the fact that the
procedure had not been codified and documented in a format
that supported the basic assumptions required for cooling load
calculations . The ASHRAE Loads Toolkit (Pedersen et al .
2001) now provides a flexible and well-documented version of
the heat balance method that is tailored specifically for cooling
load calculations . Although the HBM requires more computation rime than other existing load calculation methods, such
as the total equivalent temperature difference/time averaging
(TETD/TA) method, cooling load temperature difference/
cooling load factor (CLTD/CLF) methods, and transfer function method (TFM), the time constraint has become less of a
concern due to the low cost of computing power.
The RTSM (Spitler et al . 1997) is a simplified, heat
balance-based design cooling load method intended for
spreadsheet applications . The RTSM is similar in concept to
the TFM in that both methods are based on the principle of
superposition. This means that these methods do not solve the
system of heat balance equations simultaneously but calculate
the individual heat gains independently of one another. The
overall thermal response of the building element at a given
time is equal to the sum of the responses caused by several
input heat gain/temperature excitations during previous times.
Although less accurate than the HBM, the RTSM has the
advantage of allowing the user to directly compare the magni-

Chanvit Chantrasrisalai and Ipseng Iu are research assistants and D .E. Fisher is an assistant professor at Oklahoma State University, Stillwater, Okla . David S. Eldridge is a project engineer at Grumman/Butkus Associates, Chicago, Ill .

160

02003 ASHRAE .

tude of individual surface, solar, and internal heat gains. It thus

has value as a forensic tool that can be used to determine the
relative magnitude of component heat gains. In addition, as
demonstrated by Rees (Rees et al. 1998), the method matches
both the heat balance and measured cooling loads quite well
for a wide range of cases. Since the RTSM is explicitly based
on the heat balance method, validation of the HBM also serves
to validate the RTSM. In particular, the experimental validation of the HBM provides a reference point for the parametric
study presented by Rees (Rees et al. 1998), in which he
compares the HBM and the RTSM over a wide range of building configurations.
Although ASHRAE has developed a number of cooling
load procedures in the past, it has never experimentally validated those procedures. Earlier cooling load procedures were
adopted with only comparative numerical studies to support
them. The ASHRAE-sponsored research project entitled
Experimental Validation of Heat Balance/RTS Cooling Load
Calculation Procedure (RP-1117), upon which this and
companion papers are based, undertook for the first time in
ASHRAE history to experimentally validate a cooling load
procedure. Eldridge et al. (2003) describe the development of
experimental facilities at Oklahoma State University in detail.
In this paper, the validation of the HBM is emphasized. Overall, comparisons between predicted cooling loads by the HBM
and measured data are illustrated for various building configurations. A companion paper (Iu et al. 2003) focuses on the
validation of the RTSM using the same sets of observed data.

DESCRIPTION OF TEST BUILDINGS

To experimentally validate the cooling load calculation
procedures, two geometrically identical test cells were
constructed in an open field at Oklahoma State University. The
two test cellsone thermally massive, designated the heavy
building, and one less thermally massive, designated the
light buildingwere oriented directly north and south of
each other. Development of the test cells and their construction
and instrumentation are described in detail in a companion
paper (Eldridge et al. 2003). To estimate the cooling loads for
each construction, the test cells were modeled as shown in
Table 1. The material layers are listed from outside to inside.
The thermal and physical properties of the building materials
were obtained from the available literature (ASHRAE 1997,
2001) and are presented in Table 2.
OVERVIEW OF THE HEAT BALANCE METHOD
The HBM is based on the law of conservation of energy.
A set of energy balance equations for an enclosed space is
solved simultaneously for unknown surface and air temperatures. As a result, significant assumptions required by other
methods, such as the linearity of heat transfer processes and
time-invariant system characteristics, are not necessary in the
heat balance method.
In current implementations of the HBM (BSO 1991;
Crawley et al. 2000; Pedersen et al. 2003; Walton 1983), a
uniform zone air temperature is commonly assumed.

TABLE 1

Details of Building Construction for Base Case Configuration

Type of
Construction

Heavy Building

Light Building

Walls

4 in. face brick

1/4 in. stucco

1 in. plastic foam (expanded polystyrene, extruded)

1 in. plastic foam (expanded polystyrene, extruded)

8 in. HW concrete block filled w/ HW concrete

1/2 in. plywood

3 in. fiberglass insulation
1/2 in. gypsum board

Roof

Shingle roll

Shingle roll

Tarpaper

Tarpaper

1/2 in. plywood

1/2 in. plywood

2 in. plastic foam (expanded polystyrene, extruded)

2 in. plastic foam (expanded polystyrene, extruded)

1/2 in. plywood

1/2 in. plywood

5 in. concrete

Ceiling air space

Metal decking

Metal decking

Metal decking

Metal decking

5 in. concrete

3.5 in. concrete

Clear 1/8 in. glass

Clear 1/8 in. glass

Floor

Window

ASHRAE Transactions: Research

161

TABLE 2

Thermal and Physical Properties of Building Materials

Thickness

Thermal
Conductivity

Density

Specific Heat

Resistance

m
(in.)

W/mK
(Btuin./hftF)

kg/m
(lb/ft)

kJ/kgK
(Btu /lbF)

mK/W
(hftF/Btu)

Acoustic tile

0.019
(3/4)

0.061
(0.423)

481
(30.0)

0.84
(0.20)

B3-insulation

0.051
(2)

0.043
(0.298)

32.0
(2.0)

0.84
(0.20)

carpet

0.306
(1.740)

Ceiling air space

0.176
(1.00)

3.5 in. concrete

0.0889
(3 )

0.1730
(1.20)

641
(40.0)

0.84
(0.20)

5 in. concrete

0.1270
(5)

0.1730
(1.20)

641
(40.0)

0.84
(0.20)

4 in. face brick

0.1016
(4)

1.333
(9.24)

2002
(125)

0.92
(0.22)

3 in. fiberglass insulation

0.0889
(3 )

0.036
(0.025)

96
(6.0)

0.96
(0.23)

Clear 1/8 in. glass

0.00667
(0.038)

1/2 in. gypsum board

0.0127
(1/2)

0.727
(5.00)

1602
(100)

0.84
(0.20)

8 in. HW concrete block filled w/ HW

concrete

0.2032
(8)

1.731
(1.20)

2243
(140)

0.84
(0.20)

Metal decking

0.002
(0.08)

45
(31.20)

7689
(480)

0.42
(0.10)

1/2 in. plywood

0.0127
(1/2)

0.116
(0.080)

540
(33.7)

1.21
(0.29)

Shingle roll

0.001
(0.04)

0.0369
(0.026)

1100
(68.7)

1.51
(0.36)

1/4 in. stucco

0.0064
(1/4)

0.692
(0.480)

1858
(116)

0.84
(0.20)

1 in. plastic foam

(expanded polystyrene, extruded)

0.0254
(1)

0.0280
(0.019)

32
(2.0)

1.21
(0.29)

2 in. plastic foam

(expanded polystyrene, extruded)

0.0508
(2)

0.0280
(0.019)

32
(2.0)

1.21
(0.29)

Tarpaper

0.0010
(0.04)

0.0129
(0.009)

1100
(68.7)

1.26
(0.30)

Wood

0.10
(4)

0.121
(0.084)

593.0
(36.6)

2.51
(0.60)

Material Layer

162

ASHRAE Transactions: Research

instance, the outside surface heat balance includes sky radiation, solar position, exterior convection coefficient, and sky
temperature models, while the inside surface heat balance
consists of longwave radiant exchange, interior convection
coefficient, and radiative distribution models. Since the HBM
requires implementing a series of various component models,
additional assumptions included in the various component
models are also included in each formulation of the HBM.
However, these assumptions are not inherent in the method.
This means that its accuracy can be improved by replacing a
coarse component model with a refined one without modifying the overall procedure. In the next section, validation
approaches and modifications of an ASHRAE Loads Toolkitbased computer program used in this investigation of the HBM
are described in detail.
TECHNICAL APPROACH

Figure 1 Heat balance processes in a space (Pedersen et

al. 1998).

Although the HBM is extensible to multinode or computational fluid dynamic (CFD) models, previous investigations
(Fisher and Pedersen 1997) have shown that the well-stirred
zone model is adequate for the calculation of cooling loads in
office-sized zones with radial ceiling diffusers. Additional
assumptions in the ASHRAE Loads Toolkit version of the heat
balance include:
1.

All exposed surfaces of the room are diffuse radiators and

have uniform longwave and shortwave emissivities.

2.

All exposed surfaces have uniform surface temperatures.

3.

Conductive heat transfer through envelopes is one-dimensional.

4.

All airborne moisture entering and generated in the space is

added to the zone air and is not absorbed by the building
materials.

These assumptions are reasonable for building applications. With these assumptions, computational complexity can
be reduced without introducing significant inaccuracies into
the solution.
Pedersen et al. (1998) describe the heat balance procedure
in terms of three different energy balance processes: the
outside surface heat balance, the inside surface heat balance,
and the air heat balance. The relationship between these
energy balance processes is illustrated in Figure 1. The outside
and inside surface heat balances are linked by the conductive
heat transfer process through the envelope, while the inside
surface and air heat balances are linked by the convective heat
transfer process at the inside surface. Each part of the heat
balance procedure consists of various component models. For
ASHRAE Transactions: Research

The ASHRAE Loads Toolkit program (Pedersen et al.

2003), which was used in this investigation, was originally
based on algorithms from the BLAST program (BSO 1991).
The toolkit employs a single-zone heat balance model and is
intended to be a collection of component models rather than a
full-featured program for energy analysis or load calculations.
Its modular nature conveniently allows incorporation of submodels available in other programs with little restructuring.
This feature facilitates an investigation of various combinations of different models using the heat balance method.
Using the Loads Toolkit program, the heat balance procedure was validated at two input levels designated baseline
model and tuned model. With the exception of outdoor
temperatures and incident solar radiation, the baseline model
is constructed based on published and estimated values of
input parameters. This model is expected to yield results similar to those calculated under realistic field conditions. Typical
design data obtained from the literature (ASHRAE 1997) were
used as input data for the baseline model, while measured data
collected at the OSU test facilities and the Mesonet weather
station were used to improve input parameters for the tuned
model. The measured data included outside and inside air
temperatures, global horizontal solar radiation, wind speed,
wind direction, system air temperature, system airflow rate,
and surface shortwave absorptances. These observed data1
were averaged and utilized as hourly input data.
In order to field validate the load calculation method
properly, hourly measured outside air temperatures were used
for both the baseline and the tuned models. Since the Loads
Toolkit program predicted the amount of global horizontal
solar radiation quite well on a clear day and since all field
experimental results were selected for clear or almost clear
days, design solar radiation was used for the baseline model.
To use measured data, two changes in the Loads Toolkit
program were required for the tuned model. First, the
ASHRAE clear sky model was modified to use measured solar
1.

Except for measured shortwave absorptances, which are constant.

163

TABLE 3

Summary of Input Data and Models for Toolkit Test Runs

Input Data / Component Models

Baseline Model*

Tuned Model

Outdoor air temperatures

Measured data

Measured data

Indoor air temperatures

Measured data

Measured data

Sky (solar) radiation

ASHRAE clear sky model

()

Modified ASHRAE clear sky model with

measured global horizontal solar radiation

Sky temperatures

Blast design day model

Blast design day model

Sky (solar) view factors

ASHRAE model

ASHRAE model

Exterior convection coefficients

MoWiTT model with design data

()

MoWiTT model with measured wind speed

and wind direction

SW absorptances

Design data
()

Measured data

Windows

Window4.1 model

Window4.1 model

Internal shortwave radiative

distributions

User-specified model (area-weighted distribution)

with design sw absorptances
()

BLAST model with measured sw absorptances

Interior convection coefficients

ASHRAE model
()

Fishers model with measured system air

temperature and flow rate

The arrow in parenthesis indicates a general trend in predicting cooling loads during the day by each component model as compared to the results predicted by that component
model employed in the Toolkit-tuned model. The up arrow means a higher prediction while the down arrow means a lower prediction. See Chantrasrisalai et al. (2003) for
detailed explanations for model comparisons.

The up and down arrows indicate that the trend depends primarily on how design data compare to measured data. Since wind speed and wind direction are unpredictable,
no general trend can be readily identified.

radiation instead of using apparent solar irradiation tabulated

in the literature (ASHRAE 1997). The other change in the
tuned model was to replace the constant wind speed and wind
direction parameters in the baseline model with measured
hourly data. Table 3 summarizes input data and component
models used for the Toolkit baseline and tuned model tests.
Both baseline and tuned models employed almost the
same component models. However, radiative distribution and
interior convection algorithms differed between the baseline
and tuned models for the following reasons:
1. Design practitioners typically specify the radiative distribution.2
2. It is difficult to properly estimate hourly system air temperatures and flow rates3 at the design stage as required by
Fishers model (Fisher and Pedersen 1997).
Therefore, the baseline model, which represents a design
HB model, utilized the user-specified radiative distribution
and constant interior convection models. However, the tuned
model, which represents a field-validated HB model,
employed the BLAST radiative distribution and Fishers interior convection algorithm.
For the Toolkit baseline model, design shortwave absorptances were 0.9 for exterior roofs and 0.5 for all other
2.
3.

This is quite important, especially for passive solar design when

designers need to specify thermal mass location.
These are system air conditions that enter a zone of interest, not
the conditions that leave the cooling coil.

164

surfaces.4 However, measured surface absorptances used for

the Toolkit-tuned model were 0.9 for exterior roofs, 0.7 for a
heavy buildings exterior walls, 0.6 for a light buildings exterior walls, and 0.3 for interior surfaces. In this study, no internal heat gains were considered since the test buildings were
unoccupied. For infiltration, the rate of 0.25 air changes per
hour (ACH) was included for all test runs. Eldridge et al.
(2003) present an analysis of infiltration for the test buildings.
TEST CELL CONFIGURATIONS
The field experiments collected hourly cooling loads for
various test cell configurations, as shown in Table 4. The
experiments were designed to vary one parameter at a time
with a single test (office case) showing the combined effect of
all parameters. Default configurations for heavy and light
buildings, extensively described by Eldridge et al. (2003),
were used for base case comparison. Each test ran for at least
24 consecutive hours after the room had reached steady-periodic conditions. As previously mentioned, the experimental
tests selected for this study were on clear or almost clear days.
The experimentally measured cooling loads were compared to
cooling loads calculated by the HBM under comparable environmental conditions. The results are presented and discussed
thoroughly for each configuration in the next section.
4.

Exterior roofs were black while other surfaces were mediumcolor.

ASHRAE Transactions: Research

TABLE 4

Field Experiments for Heavy and Light Buildings

Base
Configuration

Suspended
Ceiling

Carpet

Blinds

Furniture

Office Configuration

Suspended ceiling:
2 ft 4 ft lay-in ceiling
below bar joists

No

Yes

No

No

No

Yes

White venetian
blinds

No

No

No

Yes
(at 45)

No

Yes
(at 0)

Carpet with in. pad

No

No

Yes

No

Yes

Yes

Furniture: office furniture including desks,

tables, chairs, and
bookshelves filled with
books

No

No

No

No

Yes

Yes

Test

TABLE 5

Peak Errors and Sources of Errors in Baseline Models Relative to Tuned Models (%)
Base Case

Suspended Ceiling

Carpet

Furniture

Model/Input
Parameter

Heavy
Bldg

Light
Bldg

Heavy
Bldg

Light
Bldg

Heavy
Bldg

Light
Bldg

Light
Bldg

Sky radiation

3.8%

2.1%

0.6%

0.6%

2.4%

2.4%

1.8%

Design solar radiation (clear sky)

with clearness number of 0.98

Exterior convection
coefficients

1.7%

1.1%

1.2%

0.3%

4.1%

3.5%

0.1%

Design wind parameters (constant

wind speed and direction)

SW absorptances

7.5%

7.1%

6.2%

6.6%

1.5%

1.5%

1.6%

Design surface absorptances

Internal shortwave
radiative distributions

12.5%

12.0%

11.6%

11.5%

3.5%

6.6%

5.7%

Area-weighted shortwave
radiative distribution with design
surface absorptances

Interior convection
coefficients

17.1%

6.5%

18.6%

8.4%

7.7%

2.2%

4.1%

Constant natural convection

coefficients

Peak error (+)

15.1%

14.1%

13.2%

13.3%

3.5%

6.7%

5.9%

Peak error (-)

17.2%

6.6%

18.6%

8.4%

9.2%

4.7%

4.5%

RESULTS AND DISCUSSION

Overview of Results
To validate and illustrate the performance of the heat
balance method, hourly cooling load profiles calculated by
both baseline and tuned models are plotted against measured
results. The results are plotted as a percentage of the total
measured peak load in order to facilitate comparative analysis
of the methods. That is, the measured peak cooling load is
always shown as 1.0 in the graphical comparisons. Iu et al.
(2003) describe how cooling loads are normalized. The
discussion focuses on the ability of the HBM to accurately
predict the 24-hour cooling load profile for each configuration
using the two different input models described in the previous
section. As shown in Table 3, the input data and components
that form the models can have offsetting effects on the overall
cooling load calculation. The purpose of the baseline model is
ASHRAE Transactions: Research

Source of Error

to assess the uncertainty associated with commonly made

modeling assumptions. Although the baseline model cooling
load profiles shown in the following sections give some idea
of the suitability of the baseline model, they do not give an
accurate picture of the uncertainty associated with the baseline
modeling assumptions. This uncertainty, which is calculated
as a deviation from the tuned model peak cooling load, and its
sources of errors are shown in Table 5.
The peak error is based on the assumption that the individual errors are independent and random and can therefore be
calculated as the RMS error. The peak error shows the maximum expected deviation of the baseline model from the tuned
model. It does not address deviation of the results from the
measured data but serves to illustrate the sensitivity of the
predicted cooling load to various input parameters.
Table 5 shows that the ASHRAE interior convection coefficients, which are used in the baseline models, could predict
165

the peak cooling loads about 19% lower than the tuned model
coefficients (for the suspended ceiling configuration). The
experimental tests were run at relatively high air circulation
rates (about 19.5 ACH for both test buildings). Consequently,
the natural convection coefficients specified by the ASHRAE
model introduced large errors in the cooling loads for some
cases, as shown in Table 5. The radiative distribution algorithm and surface absorptances utilized by the baseline model
tended to give higher cooling loads than those used by the
tuned model. Hence, the effects of these parameters would
cancel each other out depending on thermal mass of the buildings and the relative interior surface temperatures.
The cooling load for each of the buildings was dominated
by the solar heat gain with some contribution from conductive
heat gains and a small contribution from infiltration heat gains
for all test configurations. Internal heat gains, which often
dominate a cooling load calculation in a typical office building, were eliminated for the purpose of testing the procedures.
By eliminating instantaneous internal heat gains, the experiments were able to test the procedures ability to predict the
thermal mass effect. It is worth noting that an inclusion of
internal loads would certainly reduce the percentage error in
the baseline model relative to the tuned model because both
models would utilize the same procedure in converting internal heat gains into cooling loads. Also, internal heat gains
could either dampen or enhance the thermal mass effect
depending on their type5 and magnitude. Of particular interest
is the fact that the cooling load profiles peak at the same time
for all zone configurations for both the heavy and the light
buildings. Significant load absorption is accomplished by the
thermally massive heavy building without a shift in peak hour.
This somewhat counterintuitive result, which is correctly
predicted by the heat balance for all cases, is primarily due to
5.

Different types of internal heat gains have different radiative and

convective fractions that contribute to cooling loads. High
convective fraction would dampen the thermal mass effect, while
high radiative fraction would enhance the thermal mass effect.

the high convective heat transfer rates in the zones and the
similarity of the floor constructions in the two buildings, as
discussed in a companion paper (Iu et al. 2003).
Base Configuration
Both the heavy and light buildings were modeled, as
shown in Tables 1 and 2. Data for this test were collected in
mid-September. Figure 2 shows predicted hourly cooling load
profiles and measured cooling loads for the heavy and light
buildings. As expected, the tuned model, which represents a
field-validated HB model, shows better agreement with
measured results than the baseline model. The tuned model
overpredicts the peak cooling loads by less than 2% for the
heavy building and about 4% for the light building. The baseline model overpredicts the measured peak cooling loads by
4% for the heavy building and 14% for the light building. The
excellent peak load agreement achieved by the baseline model
for the heavy building is a serendipitous result of canceling
errors, as shown in Table 5.
Partially cloudy skies during the morning hours caused
the baseline model (which assumes a clear sky) to overpredict
the cooling load during the three early morning hours. Consequently, design solar radiation predictions were higher than
the measured solar radiation during the morning. These results
indicate the significance of the solar heat gain for the test
configurations. The results of the base configuration show that
the thermal storage capacity of the heavy building significantly damped the peak cooling load during the day, resulting
in a heavy building peak load that was over 25% lower than the
light building peak.
The differences between the baseline and tuned model
results, as shown in Figure 2, also illustrate the thermal mass
effect. For the heavy building, the overprediction of the solar
radiation during the early morning hours did not result in an
on-peak error in the baseline model cooling load results.
Rather, it resulted in an off-peak (nighttime) error. That is,
the solar heat gain of the early morning hours showed up as

Figure 2 Cooling loads for base configuration.

166

ASHRAE Transactions: Research

Figure 3 Cooling loads for suspended ceiling configuration.

cooling loads during the night. For the light building, the overprediction of the solar heat gain during the early morning
hours by the baseline model showed up as an on-peak
(daytime) error in the cooling load. For this building, there was
no noticeable nighttime error in the baseline model results.
The tuned model, which used measured solar data to correctly
predict the early morning solar heat gains, resulted in significantly improved calculated cooling loads for the heavy building during the night and for the light building during the day.
Together, the two models (baseline and tuned) illustrate the
sensitivity of the heat balance procedure to thermal mass
effects.
Suspended Ceiling Configuration
Suspended ceilings, which were installed in both buildings, were modeled by adding a ceiling air space layer and an
acoustic tile layer to the roof constructions shown in Table 1.
Cooling load data for the suspended ceiling configuration
were recorded in mid-August. Figure 3 shows predicted
hourly cooling load profiles and measured data for both buildings. The tuned model underpredicts the peak cooling load by
less than 2% for the heavy building and overpredicts the peak
load about 2% for the light building. The baseline model
underpredicts the peak result by 6% for the heavy building and
overpredicts the peak loads by 9% for the light building. This
is well within the uncertainty intervals predicted in Table 5 for
this configuration.
The suspended ceiling case produced the best agreement
between measured and predicted cooling loads of all configurations tested. This is likely due to the fact that the Fisher
inside convection correlations, employed by the tuned model,
were developed for an attached radial ceiling jet (Fisher and
Pedersen 1997). This convection model closely matches the
suspended-ceiling configuration. The base case models, as
illustrated by Iu et al. (2003), calculate the ceiling convection
as if the ceiling jet is attached, when, in fact, the diffuser is
located at the suspended ceiling level, resulting in a free jet in
ASHRAE Transactions: Research

the space. Consequently, the base case tuned models overpredict the ceiling heat transfer rate. The Fisher correlations,
which were valid for the relatively high air circulation rates
experienced by the test cells, correctly predict the cooling rate
of the thermal mass.
Carpet Configuration
To investigate the effect of floor coverings on the cooling
load profile, carpet was installed in each test building. Pedersen et al. (1998) tabulate R-values of typical carpet samples
and suggests that carpet can be modeled as a purely resistive
layer because of its lightweight construction. When the Rvalue of the carpet is not available, it can be estimated by
multiplying 18.64 mK/W (2.69 hft2F/Btuin.) by the total
carpet thickness in meters (inches) (CRT 2001). The test cell
carpet was modeled by adding a resistive layer of R-value of
0.30 m2K/W (1.74 hft2F/Btu) to the floor constructions
shown in Table 1.
Test cell cooling loads with the carpet installed were
measured in mid-October. Figure 4 shows predicted hourly
cooling load profiles plotted against measured data for both
buildings. Surprisingly, the tuned model overpredicted the
peak cooling loads by more than 10% for both buildings, while
the baseline model underpredicted the peak load by 4% for the
heavy building and overpredicted the peak load by 10% for the
light building.
The off-peak (nighttime) error for both the tuned and baseline models for both buildings indicates that a significant fraction of the error is due to one or more model input parameters
(in addition to those shown in Table 5) that were incorrectly
specified in the load calculation. The most likely candidates are
an incorrectly reported outdoor dry-bulb temperature or an
underspecified infiltration rate. Both would have a significant
effect at night and a negligible effect during the day when the
cooling load is dominated by solar heat gains. The measured
temperature data show rapid 2C (3.6F) temperature variations during the night that were lost in the translation to hourly
167

Figure 4 Cooling loads for carpeted test cells.

Figure 5 Off-peak error corrected by varying infiltration rate.

model input data. In addition, a higher than expected infiltration rate likely occurred due to a change in wind direction and
wind speed that resulted in an increased pressure differential
across the door. For the measured temperature difference, an
infiltration rate of 1 ACH would completely account for the
off-peak error, as shown in Figure 5.
Four factors may contribute to the on-peak error: the
carpet thermal resistance, the convective heat transfer coefficient associated with the carpeted surface, the radiation distribution, and the carpet solar absorptivity. To assess the effect of
the carpet thermal resistance, R-values of the carpet were
investigated over an expected range (0.0 0.8 m2K/W or 0.0
4.5 hft2F/Btu) for the tuned model. The carpet was also
modeled as a thermal mass layer (see thermal properties in the
literature [ASHRAE 2001]). Figure 6 shows hourly cooling
load profiles predicted by the tuned model with the carpet
having different thermal properties.
As shown, the maximum differences between peak loads
predicted by the tuned model using different thermal proper168

ties are less than 8% over the expected range for both buildings. The original tuned model (with R-value of 0.3 m2K/W
or 1.74 hft2F/Btu) predicted an increase in the peak cooling
load of only 3% and 4% for the heavy and light buildings,
respectively, with no carpet layer. The uncertainty in the thermal resistance of the carpet layer, therefore, contributes little
to the total on-peak error.
The convective heat transfer coefficient has already been
shown to have a significant effect on the cooling load for the
test cell configuration. The carpet test was run without the
suspended ceiling in place. This alone can result in a significant error in the predicted cooling load, as shown in the base
case analysis. In addition, the Fisher model for inside convection coefficients was developed for smooth surfaces. Approximating the carpet as a smooth surface could introduce
significant error in the predicted floor convective heat flux, as
shown in Figure 7.
The cooling load was most sensitive to the internal solar
distribution algorithm and the solar absorptivity of the carpet.
ASHRAE Transactions: Research

Figure 6 Effect of carpet thermal resistance and mass on cooling load.

Figure 7 Impact of floor heat transfer coefficient on cooling load.

Using the detailed beam solar algorithm results in a decrease

in predicted peak loads of 6% for the heavy building and 3%
for the light building. Changing the solar absorptivity of the
carpet from 0.3 to 0.7 results in an increase in the peak cooling
load of 10% for the heavy building and 8% for the light building, as shown in Figure 8. The uncertainty in the measured
solar absorptivity is at least 0.1, and the uncertainty associated with an estimate of solar absorptivity in the absence of
measured data is even higher. For carpeted zones with high
solar heat gains, the solar absorptivity, not the thermal resistance, is the significant parameter. To facilitate accurate calculation of design cooling loads, radiative properties, such as
those given by Mabinton and Goswami (1980), should be
published in the ASHRAE literature.
Furniture Configuration
Several desks, chairs, tables, and bookshelves filled with
340 kg (750 pounds) of books were moved into the carpeted
ASHRAE Transactions: Research

lightweight test cell to investigate the effect of furnishings on

the cooling load. As recommended by Pedersen et al. (1998),
furnishings can be modeled as an additional surface with a
surface area and thermal mass roughly equivalent to the
furnishings in the zone. For this investigation, the thermal
mass was modeled as three material layers: wood, B3-insulation, and sheet metal (see properties in Table 2). These materials approximate the mass and surface area (15.0 m or 161.5
ft) of the furniture used in this study.
Figure 9 shows cooling loads predicted by both models
compared with the measured results for the light building. The
tuned model overpredicted the peak load by about 12%, while
the baseline model overpredicted the peak result by about
10%. These results do not vary significantly from the carpet
only test (Figure 4) for the light building. This suggests that it
is reasonable to model the furniture as an additional surface in
the zone.
169

Figure 8 Influence of carpet surface absorptance on cooling load.

Window Blind Configuration

Horizontal-slat blinds set at a 45-degree slat angle to
mostly block direct beam radiation were added to the base case
configuration for both heavy and light buildings to study the
effect of an interior-shading device on the cooling load. Presently, the Window 4.1 model employed in the official version
of the ASHRAE Loads Toolkit does not directly support the
addition of blinds to windows for determining the influence of
the blinds on transmitted solar radiation and cooling loads.
Three blind models (EnergyPlus 2002; Parmelee and Aubele
1952; Pfrommer et al. 1996) were implemented in a research
version of the Loads Toolkit to evaluate model capabilities in
predicting optical characteristics of fenestration having slattype blinds. All three models show good agreements with
measured solar radiation transmitted through the fenestration
system with blinds. The Parmelee model6 was used with the
Window 4.1 model to study the effect of the blinds on the cooling loads for this case. Note that the blind model can provide
only optical properties of the fenestration having blinds. In
order to predict the thermal effects of the blinds, the interior
solar attenuation coefficient (IAC) recommended by
ASHRAE (2001) was used. For a clear single glazing system
with light-colored blinds at 45 slat angle, the recommended
value is 0.68. Using the IAC along with the solar heat gain
coefficients (SHGC) for an unshaded glazing and optical
properties predicted by the blind model, the inward and
outward flowing fractions can be calculated for fenestration
with blinds.
Figure 10 shows the results predicted by the baseline and
tuned models using the research version of the Loads Toolkit.
As shown, only the tuned model with R-value of 0.100 m2K/
W (0.568 hft2F/Btu) had good agreements with experimental cooling loads for both buildings. It should be noted that the

R-value of 0.00667 m2K/W (0.038 hft2F/Btu) is the thermal

resistance value for a single-glass layer. The presence of the
blind layer and air layer between the blinds and the windows
would definitely increase the total thermal resistance7 of the
fenestration system. However, the value of thermal resistance
is generally unknown and can vary considerably depending on
a number of parameters including the type of glazing systems,
spacing between the glazing and the blinds, as well as the blind
geometry (slat spacing and slat angle). Also, the presence of
the blinds can alter the convective heat transfer rate at the
fenestration system, resulting in different convection coefficients from those for the fenestration without blinds. Although
the results of the tuned model shown in Figure 10 demonstrate
the ability of the heat balance method to support interior shading devices, more general thermal models and/or thermal

6.

7.

Differences between the three models have a trivial effect on

cooling load results. The maximum difference in peak loads
predicted by the three models is less than 2%.

170

Figure 9 Furniture and carpet configuration for the light

building.

The total thermal resistance was used in the calculation of

conductive heat transfer through the fenestration system because
all layers (glass, air, and blinds) have very low thermal mass.
ASHRAE Transactions: Research

Figure 10 Cooling loads for blind configuration obtained by using a blind model, IAC of 0.68, SHGC of unshaded glass, and
different R-values of the fenestration system.

parameters (including inward/outward flowing fractions, thermal resistance, and convection coefficients for various situations) are needed to predict thermal characteristics of the blind
systems so that they can be used in conjunction with the optical
blind models in design calculations.
It is worth noting that the presence of blinds caused reductions of peak measured cooling loads of 26% for the heavy
building (from 2573 to 1905 W) and 35% for the light building
(from 3413 to 2231 W) as compared to the peak measured
loads for the base-case configuration. The differences between
blind and base-case measured results are more significant than
differences between base-case measured results for other
configurations.8 This indicates the importance of using shading devices to reduce building heat gains and cooling requirements for highly glazed spaces. The effect of blinds on the
peak cooling load is related to the fractional contribution of the
solar heat gain. For the test buildings, solar radiation
accounted for approximately 75% of the total heat gain. Thus,
the impact on the peak cooling load was significant. The
results also illustrate the importance of the thermal interactions between the zone and the blind. Although the blind
settings were the same in the two buildings, the resulting
reduction in cooling load was 9% greater for the lightweight
building. A thermal blind model that accurately predicts the
convective and radiative heat transfer rates from the fenestration system is required to account for these effects.
Office Case Comparison
A general office configuration combined the effects of the
four previous tests by configuring each building with a
suspended ceiling, blinds, carpet, and furnishings, as shown in
Table 4. It should be noted that the blinds were set at 0 degrees
(opened blinds) instead of 45 degrees. Similar to the window

blind configuration, the Parmelee model was used to predict

solar radiation transmitted through fenestration with blinds at
0 slat angle. In the literature (ASHRAE 2001), information of
IAC values is only available for blinds at 45 slat angle. Since
the blinds were set at the fully opened position, it could be
expected that the IAC for this case would be higher than that
for the blinds at 45 slat angle. Therefore, the IAC of 0.85 was
used in this simulation.
Again, the R-values for a fenestration system having
opened blinds are unknown. It was assumed that the total thermal resistances for the fenestration system with blinds set at
different slat angles were not significantly different;9 thus, the
R-value of 0.100 m2K/W (0.568 hft2F/Btu), which gave
good agreement for the window blind configuration, was also
used for this case for both buildings. Figure 11 shows cooling
load profiles predicted by the baseline and tuned models using
the research version of the Loads Toolkit program compared
with measured data for both heavy and light buildings. Once
again, the results illustrate the ability of the heat balance
method to support interior shading devices but cannot be
generalized in the absence of thermal blind models in the toolkit.
Summary of Results
Figure 12 shows the deviation of the predicted peak cooling load from the measured peak cooling load for each building configuration. The baseline model underpredicts the peak
load for the heavy building and overpredicts the peak loads for
the light building. This suggests that for the thermally massive
structure, the inside convection coefficient has a relatively
greater influence on the instantaneous cooling load than the
solar heat gain. On the other hand, solar radiation becomes
9.

8.

Different environmental conditions had an important influence on

cooling loads as well as different building configurations.

ASHRAE Transactions: Research

The most substantial difference of thermal resistance could be

expected between the fully closed (90 or 90) and opened (0)
blind cases.
171

Figure 11 Cooling loads for office configuration obtained by using a blind model, IAC of 0.85, SHGC of unshaded glass, and
R-value of 0.100 for the fenestration system.

Figure 12 Measured peak-load error for all configurations.

relatively more important than inside convection for the light
building. Additionally, it was observed that the BLAST solar
distribution algorithm tends to overpredict the peak loads
when the inside absorptance of the floor is significantly different from the absorptances of other surfaces. For these cases,
accurate distribution of the beam radiation significantly
improves predicted results. Finally, the Fisher convection
model tends to overpredict the peak loads when the air diffuser
results in a free jet instead of a radial wall jet. Peak cooling
load errors for the drop-ceiling case were minimized by the
inlet jet configuration.
CONCLUSIONS AND RECOMMENDATIONS
To evaluate the performance of the heat balance procedure, cooling loads predicted by the ASHRAE Loads Toolkit
program were compared with measured results for various
building configurations. Since the heat balance procedure
consists of a series of component models describing various
172

heat transfer processes, the accuracy of the procedure is

dependent on which component models are used. In addition,
the cooling load can be quite sensitive to various input parameters.
Both the tuned and baseline model results deviated from
measured cooling loads by less than 15% for all cases. This
excellent agreement was obtained for cooling load profiles
that were entirely dominated by solar and conductive heat
gains. It is expected, that the models will accurately predict
hourly building thermal loads for all cases where reasonable
input data and appropriate sets of component models are used.
It should also be noted that the tuned model underpredicted the
cooling for only one case, and that by less than 2%.
The investigation showed that the cooling loads were
quite sensitive to interior convection coefficients and solarrelated parameters. This suggests that more research on solar
radiation transmitted through complex fenestration system,
internal solar distribution, and inside convection for various
ASHRAE Transactions: Research

flow regimes and zone configurations are needed. For spaces

with high solar heat gains, the cooling load is very sensitive to
the physical properties of the floor covering. Convection
correlations and optical properties should be developed for the
ASHRAE literature. Finally, the ASHRAE Loads Toolkit
should be extended to include interior shading devices.
Although the ASHRAE window model is adequate for singlepane windows, commercial glazing systems require the accuracy of the Window 4.1 models. Blind models should be developed specifically to support these models.
ACKNOWLEDGMENTS
This research project (RP-1117) was sponsored by
ASHRAE TC 4.1. The authors wish to extend special thanks
to Tom Romine and Steve Bruning for their thoughtful and
careful oversight of the project. Additional thanks are due to
Jeff Spitler, Simon Rees, and Curt Pedersen for their significant contributions.
REFERENCES
ASHRAE. 1997. 1997 ASHRAE HandbookFundamentals. Atlanta, Ga.: American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.
ASHRAE. 2001. 2001 ASHRAE HandbookFundamentals. Atlanta, Ga.: American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.
BSO. 1991. BLAST Technical Reference Manual, BLAST
Supporting Office, University of Illinois, Urbana-Champaign, Ill.
CRT. 2001. Carpet and Rug Institute, http://www.carpetrug.com.
Crawley, D.B., L.K. Lawrie, C.O. Pedersen, and F.C.
Winkelmann. 2000. EnergyPlus: Energy simulation program. ASHRAE Journal 42(4): 49-56. Atlanta, Ga.:
American Society of Heating, Refrigerating and AirConditioning Engineers, Inc.

Iu, I.S., D.E. Fisher, C. Chantrasrisalai, and D. Eldridge.

2003. Experimental validation of design cooling load
procedures: The radiant time series method. ASHRAE
Transactions 109(2). Atlanta, Ga.: American Society of
Heating, Refrigerating and Air-Conditioning Engineers,
Inc.
Mabinton, R.C., and D.Y. Goswami. 1980. Compilation of
thermal radiative properties of building and furnishing
materials and other interiors for use in energy conserving designs. Proceedings of the 1980 Annual Meeting of
American Section of the International Solar Energy
Society, Inc., Phoenix, Ariz., June, 1980, Newark, Del.
Parmelee, G.V., and W.W. Aubele. 1952. The shading of sunlit glass: An analysis of the effect of uniformly spaced
flat opaque slats. ASHRAE Transactions, Vol. 58, pp.
377-398.
Pedersen, C.O., D.E. Fisher, J.D. Spitler, and R.J. Liesen.
1997. Development of a heat balance procedure for calculating cooling loads. ASHRAE Transactions 103(2):
459-468. Atlanta, Ga.: American Society of Heating,
Refrigerating and Air-Conditioning Engineers, Inc.
Pedersen, C.O., D.E. Fisher, J.D. Spitler, and R.J. Liesen.
1998. Cooling and Heating Load Calculation Principles. Atlanta, Ga.: American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.
Pedersen, C.O, D.E. Fisher., R.J. Liesen, and R.K. Strand.
2001. ASHRAE toolkit for building load calculations.
Atlanta, Ga.: American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.
Pedersen, C.O., D.E. Fisher, R.J. Liesen, and R.K. Strand.
2003. ASHRAE Toolkit for building load calculations.
ASHRAE Transactions 109(1).
Pfrommer, P., K.J. Loma, and Chr. Kupke. 1996. Solar radiation transport through slat-type blinds: A new model and
its application for thermal simulation of buildings. Solar
Energy 57(2): 77-91.

Eldridge, D., D.E. Fisher, I.S. Iu, and C. Chantrasrisalai.

2003. Experimental validation of design cooling load
procedures: Facility design. To be published in ASHRAE
Transactions. Atlanta, Ga.: American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.

Rees, S.J., J.D. Spitler, and P. Haves. 1998. Quantitative

comparison of North American and U.K. cooling load
calculation procedures-results. ASHRAE Transactions
104(2): 47-61. Atlanta, Ga.: American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.

EnergyPlus. 2002. EnergyPlus engineering document: The

reference to EnergyPlus calculations. U.S. Department
of Energy.

Spitler, J.D., D.E. Fisher, and C.O. Pedersen. 1997. The radiant time series cooling load calculation procedure.
ASHRAE Transactions 103(2): 503-515. Atlanta, Ga.:
American Society of Heating, Refrigerating and AirConditioning Engineers, Inc.

Fisher, D.E., and C.O. Pedersen. 1997. Convective heat

transfer in building energy and thermal load calculations. ASHRAE Transactions 103(2): 137-148. Atlanta,
Ga.: American Society of Heating, Refrigerating and
Air-Conditioning Engineers, Inc.
ASHRAE Transactions: Research

Walton, G.N. 1983. Thermal Analysis Research Program

Reference Manual, NBSIR 83-2655. Washington, D.C.:
National Bureau of Standards.
173