Plea2004 - The 21th Conference on Passive and Low Energy Architecture.

Eindhoven, The Netherlands, 19 - 22

September 2004 Page 1 of 6

Building thermal loads: validation of a dynamic

model for the thermal performance of buildings
based on the ASHRAE Standard 140-2001.

Luigi Marletta, Gianpiero Evola and Fabio Sicurella 1

1
Department of Industrial Engineering, University of Catania, Catania, Italy

ABSTRACT: The ASHRAE has recently published a standard (ASHRAE 140-2001) for the validation
of computer programs for the thermal analysis of buildings. This standard presents a large number of
test-cases, characterized by different values of building thermo-physical properties and
environmental conditions. The validation of a program is based on the comparison of the yearly
heating and cooling loads for each case with the values obtained from eight reference validated
programs. The standard also proposes a sensitivity test to find out the weak points of the tested
algorithm.
This paper discusses the application of such a validation tool to a building energy simulation program
based on the Admittance Procedure (AP). The paper presents a detailed description of the validation
procedure and the assessment of the model.

Conference Topic: 2 Design strategies and tools.

Keywords: admittance, thermal load, validation.

INTRODUCTION 2. THE AP CODE

In recent years, several building energy simulation 2.1 Definitions

programs have been developed; these are generally The program tested in this paper is a home-made
based on different approaches and algorithms to program for the thermal analysis of building based on
model the thermal phenomena occurring in the the Admittance Procedure, and named AP code.
building. The admittance procedure is a mathematical
The validation of each new code is necessary, in procedure in which a matrix technique is used to
order to verify its validity and eliminate bugs or solve a system of partial differential equations.
algorithm errors. Three ways can be followed to This procedure can only be adopted when all the
evaluate the accuracy of a building energy simulation forces working on the system are sinusoidal
program [1]. functions. In the energy simulation of a building the
The first one is the empirical validation: the results forces (i.e. indoor and outdoor temperature, solar
obtained from the simulation are compared to data radiation) can be described as periodical functions
measured in a real building. This validation procedure with a period equal to one day (T = 86400 s). This
is affected by uncertainties due to the use of means that the Fourier Transform can be applied to
instrumentation and to the determination of the actual decompose them as a sequence of harmonics, the
value of the building materials’ properties. frequency f = k/86400 (s-1) pertaining to the k-th one.
The second way to validate a code is the If the system is linear and time-invariant, its response
analytical verification, which is based on the to each harmonic will be a sinusoidal function, too.
comparison between the output of the software and a According to this approach, the thermal flux
known analytical solution. This approach is possible through the walls can be defined as the sum of a
only for very simple cases whose solution can be steady component (i.e. the mean value) and a series
easily determined, and great attention must be paid of sinusoidal components (i.e. the fluctuation around
on the boundary conditions definition. the mean value).
The validation can finally be obtained through the The first one can be easily determined by applying
comparison between the results of the tested program the heat transfer laws in steady–state conditions, and
and those obtained from several other programs considering the mean values of the forces acting on
recognized as correct. the building.
The standard method of test presented in the To evaluate the sinusoidal components of the
ASHRAE Standard 140-2001 is based on this last thermal fluxes, the Admittance Procedure introduces
approach. three basic factors: the admittance (Y), the decrement
factor (f) and the surface factor (F).
Plea2004 - The 21th Conference on Passive and Low Energy Architecture. Eindhoven, The Netherlands, 19 - 22
September 2004 Page 2 of 6

The admittance and the decrement factor are Eq. (5) is a time-dependent function; it can be
defined as follows [2]: solved through a matrix approach, by splitting it into
24 equations, each one relative to each hour of the
~ ~ ~
qe day. Then the solution equation can be written as:
q qi
Y = ~i f = ~ = ~ (1)
Ti ~
Te =0
U ⋅ Te ~
Ti =0
U ⋅ Ti ~
Te =0 [Q 0 ] = M ⋅ [t i ] (6)

where U is the thermal transmittance of the wall, Here M is the transfer matrix, while Q0 is a vector
while the subscripts i and e refer to indoors and whose components are the hourly heat loads. The
~
outdoors, respectively. The terms T and ~q are the solution equation can be solved in order to determine
the hourly indoor air temperature ti; if ti is imposed,
sinusoidal components of temperature and heat flux. the hourly thermal load can be calculated.
If the sinusoidal functions are represented as vectors
rotating in the imaginary plane, we can write:

~ 3. THE ASHRAE STANDARD 140 – 2001

T = T ⋅ e jωτ q = q ⋅ e jωτ
~ (2)
3.1 The BESTEST procedure
where ω = 2πf, while T and q are the amplitudes of In 1986 the International Energy Agency (IEA)
the sinusoidal components. produced a procedure for the testing of building
Considering Eq. (1) and Eq. (2), it is evident that Y energy analysis computer programs, called “IEA
and f are imaginary numbers, whose phase Building Energy Simulation Test and Diagnostic
represents the time shift between a temperature Method” (IEA BESTEST). The “Standard Method of
variation and the consequent heat flux produced. Test for the Evaluation of Building Energy Analysis
The phase of the decrement factor f is defined Computer Programs”, published by ASHRAE in 2001,
“time lag”, as the thermal flux emerging from the is based on the IEA BESTEST [6].
interior surface is always delayed compared with the The BESTEST procedure consists of a series of
outdoor temperature variation which produced it. On test cases, applied to a rectangular single-zone
the contrary, the phase of the admittance Y can be building without interior partitions. The cases differ
named “time lead”, as the fluctuation of the indoor from each other for one or more parameters, so
temperature always follows the emerging thermal flux. covering a wide range of situations, from the
From the definitions of Eq. (1) it is possible to extremely simple to the relatively realistic one. A
express Y and f as functions of the geometric and detailed description of each case is given, and all the
thermo-physical properties of the wall (see [3],[4],[5]). geometric and thermo-physical variables are defined.
As regards the surface factor F, it accounts for the As an output, annual sensible cooling and heating
fraction of the solar radiation absorbed by the internal loads, peak loads, annual minimum and maximum
surfaces and given back to the indoor air. Its definition temperature can be determined.
and the operative expression are the following: The results of each simulation are then compared
to those obtained by eight programs, which at the
~
q present time are considered the most reliable and
F = ~i F = 1 − Y hi (3) precise. If the results of the tested program don’t fall
qa ~ ~
Ts =Ti =0 into the range defined by the reference programs, the
program is not necessarily incorrect, but some
algorithm may contain a bug.
where ~ qa is the thermal flux absorbed by the
The Standard also contains a number of
internal surface, Ts is the internal surface temperature sensitivity tests, based on the comparison between
and hi is the interior combined surface coefficient. the results of two similar test cases, differing from
each other for only one parameter.
2.2 The solution equation
The AP code allows the determination of either 3.2 Test cases considered in the paper
the indoor air temperature or, for a constant The validation of the code based on the
temperature room, the thermal load. In both cases it Admittance Procedure has been carried out only for
is necessary to state the energy balance equation for some test cases provided by the Standard. Those
the indoor air. Due to the low thermal mass of the air, cases including shading devices or a “deadband”
the balance gives: thermostat control strategy have been excluded, as
the code is not able to consider these effects. When a
~
∑ Qk =0 (4) “deadband” thermostat control strategy is adopted,
k the thermostat does not drive the HVAC system if the
indoor air temperature lies between 20°C and 27°C;
In Eq. (4), the terms concerning the heat fluxes within this range the thermal load must then be
through the wall can be written as a function of Y, f considered zero. On the contrary, Authors considered
and F. For a wall with a surface extension Sk, the net only the test cases including “non-deadband”
heat flux to the indoor air is: thermostat control strategy; here the HVAC system is
always running, and it supplies cold or heat
~
( ~ ~
Q k = S k ⋅ U ⋅ f ⋅ te − Y ⋅ ti + F ⋅ ~
qa ) (5)
depending on whether the indoor air temperature is
above or below 20°C.
Plea2004 - The 21th Conference on Passive and Low Energy Architecture. Eindhoven, The Netherlands, 19 - 22
September 2004 Page 3 of 6

Figure 1 shows the test cases considered in this 4. VALIDATION OF THE AP CODE
paper; two cases connected in series on the block
diagram differ from each other for only one variable. 4.1 Preliminary remarks.
The main data about the test cases are reported in In Fig. 2 the base building with two windows on
Table I, where only the variation with respect to the the south wall is shown. The size of the rectangular
case coming just before in the block diagram is building are the same for all the test cases, while the
reported. windows can change their position and size. In some
All the cases of Table I are classified as “low- cases an opaque window (high-conductance wall)
mass test cases” as the floor and the exterior walls replaces them: it has the same transmittance of the
are lightweight structures; their composition is shown transparent windows but it is opaque to the solar
in Table II (see [6] for more details). radiation.
The floor insulation is always considered very
thick - as high as the program allows – in order to
220 make heat losses negligible; in fact even the most
reliable energy simulation program can’t correctly
210 215 230 240 250 270 describe heat transmission to the ground.

200 280 300 8.0 m

6.0 m
195

Figure 1: Connection between the test cases. F. F. 2.7 m

2.0 m 2.0 m
3.0 m 3.0 m
Table I: Characterization of the test cases.

Case Nr Main data Figure 2: Base building adopted in the Standard.

Opaque window; αe = 0.1 When a transparent window is given, the Standard
220
ACH = 0 Qi = 0 εi = 0.9 εe = 0.9 provides the interior solar distribution on the different
210 εi = 0.1 surfaces. This information can be used to determine
200 εe = 0.1 the fraction Fw of the solar radiation lost through the
195 No windows window itself. For a south window, Fw = 0.06 or 0.304
215 εe = 0.1 if the interior shortwave absorptance αi is respectively
230 ACH = 1 0.9 or 0.1. In the code developed by the Authors the
240 Qi = 200 W solar radiation qa,k absorbed by each surface is
250 αe = 0.9 determined by means of the Ulbricht sphere model:
270 Windows on the south wall, αi = 0.9
αk w
280 αi = 0.1 q a,k = ⋅ ∑ (1 − Fw ) ⋅ (τ ⋅ I ⋅ S )w (7)
n
i=1
300 Windows on the east and west walls
∑ α k ⋅ Sk
α : Short-wave absorptance ε : Infrared emittance k =1
ACH : Air Changes per Hour Qi : Internal gains (W)
Subscripts : i : internal e : external where n and w are the number of walls and
windows, respectively, I (W/m2) is the solar radiation
incident on the window, S the window surface and τ
Table II: Lightweight structures for low-mass cases. its shortwave transmittance. This last optical property
depends on the angle of incidence (AOI) of the solar
Exterior walls (inside to outside) radiation; the Standard provides its value for AOI
Plasterboard Fibreglass Wood Siding varying from 0° to 90°, with a 10° interval.
0.012 (m) 0.066 (m) 0.14 (m) As regards the combined surface coefficients hi
and he, the Standard provides their values if the
Floor (inside to outside) program to be tested is not able to determine them
Timber flooring Insulation from the weather data (temperature, wind speed). As
0.14 (m) 0.04 (m) an example, hi = 3.73 W/(m2K) or 8.29 W/(m2K) for
internal infrared emittance εi = 0.1 or 0.9, respectively.
Roof (inside to outside)
Hourly weather data to be adopted in the
Plasterboard Fibreglass Roofdeck
simulations are provided by the Standard; they
0.16 (m) 0.04 (m) 0.14 (m) correspond to the Typical Meteorological Year (TMY),
defined through the measurements carried out
between 1953 and 1975 in a reference weather
station sited 1609 m above the sea. In the definition
of the Typical Meteorological Year, 8760 hourly
values are provided for each weather variable.
Plea2004 - The 21th Conference on Passive and Low Energy Architecture. Eindhoven, The Netherlands, 19 - 22
September 2004 Page 4 of 6

In this work, average monthly weather data have the external combined surface coefficients he of the
been defined, by averaging hourly data amongst all walls and of the opaque windows (Table VI).
the days of the same month. The determination of the This means that the AP code is not sensible
annual heating and sensible cooling load is thus enough to the effects of this variation. However, the
simplified, only 12 simulations being now necessary. error is not too big, as the annual sensible cooling
As far as solar radiation is concerned, only the load falls slightly outside the “optimal“ range for both
hourly total solar radiation on a horizontal surface is the cases. Besides, εe = 0.1 is a very unusual
provided, so the computation of the radiation on the condition, as almost all the building materials –
vertical surfaces is left to the program to be tested. At plaster, concrete, bricks - show a high infrared
this purpose the Liu-Jordan model was used [7]. emittance (εe = 0.85 ÷ 0.95).
In Table III, a comparison is made between the
annual solar radiation on the vertical surfaces Table IV: Annual heating load for each test case
determined through the Liu-Jordan model, and the (MWh/year).
range of the values used by the eight reference
programs of the Standard. This comparison is useful Case Nr Min Aver. Max AP
to check possible errors in this early phase of the 195 4,17 4,99 5,87 4,76 OK
work, which can reflect on the validation procedure. 200 5,23 6,39 6,88 6,51 OK
As it can be seen, the Liu-Jordan solar radiation to 210 6,46 6,68 6,97 6,56 OK
surfaces due west is out of the range resulting from 215 5,55 7,07 7,94 7,71 OK
the Standard, while results prove to be quite good for 220 6,94 7,67 8,79 7,88 OK
the other exposures. On the contrary, Liu-Jordan
230 10,38 11,18 12,24 11,51 OK
solar radiation transmitted by transparent windows
240 5,65 6,40 7,45 6,49 OK
does not match the values used by the reference
programs: such an outcome will be accounted for 250 4,74 6,03 7,02 6,23 OK
when discussing the results of the validation. 270 4,51 5,21 5,92 5,43 OK
280 4,67 5,48 6,15 5,68 OK
Table III: Annual solar radiation on vertical surfaces 300 4,76 5,44 5,96 5,69 OK
(kWh/m2).
Table V: Annual sensible cooling load for each test
Incident solar radiation (kWh/m ) 2 case (MWh/year).
Min Average Max Liu-Jordan
Case Nr Min Aver. Max AP
South 1456 1490 1566 1512
195 0,26 0,44 0,51 0,37 OK
East 959 1080 1217 1208
200 0,57 0,65 0,72 0,52 NO
West 857 1019 1090 1152
210 0,16 0,57 0,68 0,50 OK
North 367 429 457 457
215 0,64 0,73 0,85 0,62 NO
Transmitted solar radiation (kWh/m2) 220 0,19 0,64 0,84 0,61 OK
Min Average Max Liu-Jordan 230 0,45 0,93 1,14 0,82 OK
South 914 962 1051 1082 240 0,42 0,99 1,25 0,98 OK
West 563 675 735 818 250 2,18 2,79 3,38 2,26 OK
270 7,53 8,92 10,35 9,30 OK
280 4,87 6,03 7,11 6,05 OK
4.2 Results: absolute tests. 300 4,30 5,87 7,10 7,06 OK
Table IV and Table V show the annual heating
and sensible cooling loads, respectively, obtained by
the AP code, as well as the comparison with the Table VI: External combined surface coefficients for
range of the values provided by the eight reference different infrared emittance (W/m2K).
programs.
It is important to underline that the goal of the Walls Opaque windows
Standard is not to classify as wrong those building εe = 0.1 εe = 0.9 εe = 0.1 εe = 0.9
energy simulation programs which fall outside the he 25.2 29.3 16.9 21.0
range, but to help developers test their program’s
reliability and check incidental errors or bugs. As a
consequence, even when the tested program’s Case number 300, characterized by a window on
results hardly fall inside the range it would be the east and the west wall, shows an annual cooling
worthwhile to check the program or the specific load quite higher than the reference programs’ mean
algorithm. value and very close to the maximum value; so AP
As regards the annual heating load, the AP code apparently slightly overestimates the cooling load due
proves to give good results (Table V). All the values to west and east solar radiation. This deviation is
fall inside the respective range, frequently getting attributable to the high input values of the Liu-Jordan
close to the mean value. On the contrary, the AP incident and transmitted west solar radiation (Table
shows a little disagreement in the calculation of the III). In fact, if the average solar radiation of Table III is
annual cooling load for the cases number 200 and considered as an input, the annual sensible cooling
215. These cases are both characterized by a low load reduces, and it approaches to the average value
external infrared emittance (εe = 0.1), which modifies obtained by the reference programs (Fig. 3).
Plea2004 - The 21th Conference on Passive and Low Energy Architecture. Eindhoven, The Netherlands, 19 - 22
September 2004 Page 5 of 6

However, αe = 0.9 is a very unusual value for the

8 short-wave absorptance of the most common building
7 materials.
6 As far as the two last sensitivity tests are
concerned, (case No. 280-270 and 300-270 in Table
(MWh / year)

5
4 VIII), the deviation from the “optimal” range can be
3 mostly imputed to the difference in solar radiation, as
2 mentioned in sect. 3.1. Such an outcome is evident
1
from Figure 4 and Figure 5, where the sensitivity of
the AP code is correct if the adopted solar radiation is
0
the average taken from the reference programs.
Min Average AP (Aver. Max AP (Liu-
sol. rad.) Jordan)
Table VII: Sensitivity tests: difference between annual
heating loads (MWh/year).
Figure 3: Annual sensible cooling load for case
number 300. Case Nr Min Aver. Max AP
200 – 195 1,09 1,60 1,96 1,75 OK
4.3 Results: sensitivity tests. 210 – 200 0 0,28 1,20 0,06 OK
Sensitivity tests are carried out to study the 220 – 215 0 0,34 1,4 0,17 OK
accuracy of every single algorithm. The standard 215 – 200 0,29 0,67 1,26 1,2 OK
procedure considers the case of both heating and 220 – 210 0,47 0,73 1,33 1,31 OK
cooling loads (Tab. VII and VIIII); for each one a 230 – 220 3,43 3,52 3,61 3,63 NO
number of comparisons are made. Each comparison 240 – 220 -1,34 -1,27 -1,20 -1,38 NO
refers to two building configurations, which differ from 250 – 220 -2,19 -1,64 -1,45 -1,65 OK
each other by only one parameter; for instance, the 270 – 220 -2,62 -2,3 -1,95 -2,45 OK
difference between cases 210 and 200 is the value of 280 – 270 0,17 0,27 0,46 0,26 OK
the external infrared emittance εe (0.9 and 0.1, 300 – 270 0,044 0,23 0,59 0,27 OK
respectively). The deviation in the thermal load is a
measure of the sensitivity of the simulation program
to that parameter. Table VIII: Sensitivity tests: difference between
Table VII shows a little deviation from the annual sensible cooling loads (MWh/year).
“optimal” range for test number 230-220 and 240-220,
and this means that AP overestimates the increase of Case Nr Min Aver. Max AP
the heating load determined by outdoor air infiltration 200 – 195 0,16 0,2 0,24 0,15 NO
(case 230) and its reduction due to internal heat gains 210 – 200 -0,41 -0,08 0,008 -0,015 OK
(case 240). As regards air infiltration, its effect on the 220 – 215 -0,45 -0,08 0,012 -0,015 OK
annual sensible cooling load is undervalued, too (see 215 – 200 0,04 0,08 0,15 0,1 OK
Table VIII, test nr 230-220). Then, the algorithm used 220 – 210 0,024 0,075 0,15 0,1 OK
to determine the thermal load produced by natural 230 – 220 0,26 0,29 0,30 0,21 NO
ventilation probably needs a review. The relation used
240 – 220 0,23 0,35 0,41 0,37 OK
by the Authors is:
250 – 220 1,75 2,16 3,03 1,65 NO
~ ~ ~
(
Qv = ACH ⋅ V ⋅ ρ ⋅ c p ⋅ te − ti ) (7)
270 – 220
280 – 270
7,34
-3,24
8,25
-2,89
9,51
-2,46
8,59
-3,25
OK
NO
300 – 270 -3,25 -3,05 -2,83 -2,24 NO
where ρ and cp are the air density and the specific
heat of the air at the reference site height (1609 m
above the sea level), V is the volume of the room, AP (Liu- AP (Aver.
~ ~ ~ Jordan) Min sol. rad.) Average Max
while Qv (W), te and ti are the time dependent
sinusoidal functions describing the ventilation thermal 0
load, the outdoor and the indoor air temperature, -0,5
(MWh / year)

respectively. Eq. (7) is true if no “time lag” is -1

considered between outdoor air temperature and
ventilation thermal load; this hypothesis is well -1,5
founded if Qv is transferred to the indoor air directly by -2
convection. -2,5
The AP code also shows a little deviation for test
number 250-220 (see Table VIII), regarding the -3
variation of the external short-wave absorptance from -3,5
αe = 0.1 to αe = 0.9. The effect on the building thermal
load is described through the sun-air temperature:
Figure 4: Sensitivity test number 280-270.
αe ⋅ I
t sa = t e + (8)
he
Plea2004 - The 21th Conference on Passive and Low Energy Architecture. Eindhoven, The Netherlands, 19 - 22
September 2004 Page 6 of 6

AP (Aver. AP (Liu-
Min sol. rad.) Average Max Jordan)
0
-0,5
(MWh / year)

-1
-1,5

-2
-2,5
-3

-3,5

Figure 5: Sensitivity test number 300-270.

CONCLUSION

The validation procedure proposed by the

ASHRAE Standard is a very useful tool, but some
remarks can be made.
The determination of the solar radiation incident
on the vertical surfaces is left to the tested program,
so problems may occur to understand whether a
faulty result has to be attributed to the algorithm for
the solar radiation or to the main routine of the code.
Besides, hourly weather data are assigned for
each day of the year, so the transient thermal
behaviour of the building can be simulated. But the
Admittance Procedure works if a steady-periodic
regime is given, and the Standard do not pass any
information on how to act in such an instance. The
Authors have defined average weather data for each
month, but this procedure could introduce a deviation
from the expected results.
As regards the AP code, the outcome of the
validation is quite satisfactory; some improvements
will be produced to make the code able to simulate
shading devices and a “deadband” thermostat control
strategy. Then the validation will be carried out again,
considering peak loads, annual minimum and
maximum temperature, too.

REFERENCES

[1] J. Neymark, R. Judkoff et al., Energy and

Buildings, 34 (2002), 917.
[2] N.O. Millbank, J. Harrington-Lynn, Build. Services
Eng. 42 (1974), 38.
[3] M.Lj. Bojić, D.L. Loveday, Energy and Buildings
26 (1997), 153.
[4] A. Degiovanni, International Journal of Heat and
Mass Transfer, 31 (1988), 553.
[5] K. Ulgen, Energy and Buildings, 34 (2002),
273.
[6] ASHRAE, ANSI/ASHRAE Standard 140-2001,
Atlanta (USA), 2001.
[7] J.A.Duffie and W.A. Beckman, Solar Engineering
of Thermal Processes, John Wiley & Sons, NY, 1980.