A multizone building model for matlab/simulink environment

Abstract

Matlab/Simulink is known in a large number of fields as a powerful and modern simulation tool. In the field of building and HVAC simulation its use is also increasing. However, it is still believed to be a tool for small applications due to its graphical structure and not to fit well for the simulation of multizone buildings. This paper presents the development of a new multizone building model for Matlab/Simulink environment, implemented into the SIMBAD Building and HVAC Toolbox. It's general enough to model a variety of useful cases. Conforming to the Simulink philosophy, the model is modular and structured into blocks to represent the modelled phenomena. To simplify the description of the simulated building, a graphical user interface SIMBDI is developed in parallel, generating an xml building description file. This file can be read directly by the SIMBAD multizone building model. Finally, a simulation case is presented in order to compare the new model with the Trnsys multizone building model.

Full text

A MULTIZONE BUILDING MODEL FOR MATLAB/SIMULINK ENVIRONMENT
Zaki El Khoury*, Peter Riederer*, Nicolas Couillaud*, Julie Simon**, Marina Raguin**
*Centre Scientifique et Technique du Bâtiment, 84, Avenue Jean Jaurès, 77421 Marne la
Vallée Cedex 2, France
** Gaz de France - GDF, 361 Avenue Président Wilson, 93211 Saint Denis la Plaine, France
ABSTRACT
Matlab/Simulink is known in a large number of fields
as a powerful and modern simulation tool. In the field
of building and HVAC simulation its use is also
increasing. However, it is still believed to be a tool
for small applications due to its graphical structure
and not to fit well for the simulation of multizone
buildings. This paper presents the development of a
new multizone building model for Matlab/Simulink
environment, implemented into the SIMBAD
Building and HVAC Toolbox. It’s general enough to
model a variety of useful cases. Conforming to the
Simulink philosophy, the model is modular and
structured into blocks to represent the modelled
phenomena. To simplify the description of the
simulated building, a graphical user interface
SIMBDI is developed in parallel, generating an xml
building description file. This file can be read directly
by the SIMBAD multizone building model. Finally, a
simulation case is presented in order to compare the
new model with the Trnsys multizone building model.
INTRODUCTION
The present work is intended to complete the model
library Simbad Building and HVAC toolbox
(SIMBAD, 2003). This toolbox provides ready to use
HVAC models and related utilities to perform
dynamic simulation for buildings and HVAC plants.
The development of this toolbox was motivated by
the IEA annex 17 concerning virtual laboratories
(Annex 17, 1993), where several research groups
developed principles for the realisation virtual
laboratories (Laitila, 91) and (Vaézi, 91).
To date, Simbad has several building models that are
mono-zone models. When the user needs to simulate
a multizone building, he had to break it down to
several monozone blocks and to couple them
manually, that can be a source of mistakes. Therefore,
CSTB, in cooperation with GDF, started to develop a
multizone building model that facilitates the
modelling of multizone buildings. This new model is
developed with a graphical interface ‘SIMBDI’ to
draw the building and to enter data interactively.
Among other features, the Simbad multizone model
allows the use of small timesteps (in the order of
seconds) that is crucial for control purposes, has
modular structure (so it’s easy to replace a block or
add a new one) and is transparent to the user (an
expert Simulink user can access internal variables of
the model). This transparency is due to the exclusive
use of Simulink blocks and Matlab language (no S-
functions written with C…).
MATLAB AND SIMULINK
Matlab
Matlab is a high level language dedicated to technical
computing (Matlab, 2004). It is based on matrix
operations: the matrix is the basic datatype for
Matlab. Beside its built-in and main functionalities,
Matlab has a wide variety of toolboxes developed for
specialized technologies such as control systems,
neural network and several other domains.
Simulink
Simulink is a software package for dynamic systems
(Simulink, 2004), used in parallel with Matlab. It can
model linear and nonlinear systems using continuous
time, sampled time or a combination of both.
Simulink is very suitable for problems that have a
known configuration. On the other hand, it is more
difficult to deal with general purpose models, such as
a complete building model. For example, it is simple
to model with Simulink a time independent state
space problem. It is given by:
BUAXX+=
& (1)
When A and B are time independent matrices, this
system can be easily modeled with Simulink (cf.
figure 1). The matrices A and B can either be defined
directly (if known) or calculated in separate Matlab
functions or scripts. Simulink will initialise those
matrices before the simulation start.
Figure 1 A state space problem with Simulink;
Ninth International IBPSA Conference
Montréal, Canada
August 15-18, 2005
- 525 -

The problem is more complicated when these
matrices are time dependent. In this case, the
modeller has two possibilities:
1. Split down the problem to a set of time
independent blocks;
2. Use the “S-function” block that allows to
use other programming languages such as C
and Fortran.
For this building model, we have chosen the first
approach. Its implementation is described hereafter.
MODEL COMPONENTS
A complete description of the mathematical and
numerical treatments is not possible in this paper. So
only a brief description of the assumptions and the
mathematical formulation is given for the model
components as well as a general description of the
different modules of the model.
Air zones model
Each air zone is assumed to be homogeneous in
temperature. Every zone has two set points: one for
cooling and one for heating. The zone air temperature
can vary between these two set points.
When the air temperature is allowed to fluctuate
(inbetween the two setpoints), the zone heat balance
is used to predict the air temperature variation:
cplvcgceqcwzaaa PPPPPTCpV++++=⋅⋅⋅ ,,,,
&
ρ
(2)
where
∑−⋅⋅= izaisiiccw TTAhP )( ,,,, (3)
∑−⋅⋅= izaivaivv TTCpmP )( ,,,
& (4)
∑−⋅⋅= izaiazaaicplcpl TTCpmP )( ,,,,
& (5)
In the case where the air temperature tends to be
outside both set points, the heat balance is used to
calculate the heating or the cooling loads related to
the corresponding set point:
cplvcgcw PPPPP +++= ,, (6)
To be used into the multizone model, the above
equations must be defined in a matricial format (for
all zones of the building). For example, the final
format of the equation (2), used in the model is (now
equivalent to equation (1)):
)( ,,,, vcplcgceqpassaaaa PPPPBTBTAT +++++=
& (7)
The bar above a variable indicates a vector (not a
scalar). Aa, Ba,s and Ba,p are matrices that are
defined by the building zones. This way of
description is also used in the following sections.
Walls model
Multilayer walls are modelled using constant thermo-
physical properties for each layer. The heat transfer is
assumed to be one-dimensional. The wall surfaces
can be of two kinds: those in contact with air and
those with imposed surface temperature (called
“boundary” surfaces).
The governing equations for a wall are thus:
The heat conduction equation within each layer (j) of
the wall:
2
2
x
T
t
Tw
j
w∂
∂
=
∂
∂
α
(8)
where
jj
j
jCp
k
×
=
ρ
α
(9)
The boundaries conditions are:
• If the first surface of the wall is on contact
with air:
1
0
1,1,
0
1)(
ϕ
+−=
∂
∂
−=
=x
wac
x
wTTh
x
T
k (10)
• If it has an imposed temperature:
)(
1,
0tTT b
x
w=
= (11)
• If the second surface of the wall is on
contact with air:
22,2, )(
ϕ
+−=
∂
∂
−=
=Lx
wac
Lx
w
nTTh
x
T
k (12)
• If it has an imposed temperature:
)(
2, tTT b
Lx
w=
= (13)
For each interface between two layers ( j and j+1):
+− =
+
=∂
∂
=
∂
∂
jj xx
w
j
xx
w
jx
T
k
x
T
k1 (14)
This set of partial differential equations has been
discretisized using a combined finite difference and
finite volume schemes. The discretization step for
each layer is calculated by:
tx jj ∆⋅⋅=∆
αζ
2 (15)
The resulting final differential equation for all walls
in matrix form is given by:
- 526 -

ϕ
ϕ
,, waawwww BTBTAT ++=
& (16)
where
ϕ
is a vector that contains absorbed incident
fluxes (IR and solar radiation) on the walls surfaces:
solarIR
ϕϕϕ
+= (17)
Windows model
The window model is a relatively simplified model
for external windows with sun blinds. The model
assumes that the window solar transmittance and
absorbtivity are independent from the incidence angle
of the solar radiation. Hence the same solar
transmittance and absorbtivity are used for both beam
and diffuse component of the solar radiation.
The window is divided into a clear part and a shaded
part. Each part is modelled by two temperature
nodes: interior node and exterior node. The window
thermal inertia was neglected.
The used window parameters are:
1. Its total area;
2. Sunblind position (0 = completely open, 1 =
completely closed)
3. U-values for both shaded and clear parts;
4. Solar transmittance and absorptivity for
both parts.
The surface temperatures of the clear part of the
window can be explicitly calculated from the air
temperatures and the total radiative absorbed fluxes
(IR + solar) on the two window surfaces. This can be
formulated in the following matricial equation:
ϕ
ϕ
,,,,, clwnaaclwnclwn BTBT += (18)
A similar equation is used for the window shaded
part:
ϕ
ϕ
,,,,, shwnaashwnshwn BTBT += (19)
The window mean surface temperatures are
calculated using the sunblind position as weighting
factor. So
shwnclwnswn TfTfT ,,, ).1(. −+= (20)
As before, elementary equations for individual
windows are assembled into one unique matricial
equation (for the following only elementary equations
will be given).
Infrared heat exchange
The infrared heat exchange model is based on the
assumption that all surfaces behave as black bodies.
The black body equation has been linearised. This is
justified by relatively small differences between
surface temperatures in buildings.
The resulting linearised radiative heat transfer
coefficient is:
3
,)273(4 +××= msrad Th
σ
(21)
This coefficient was modified to include a first
correction for the black surface assumption using a
mean surface emissivity. The used coefficient is:
3
,)273(4 +×××= msrad Th
σε
(22)
The model distinguishes between internal infrared
heat exchange (internal surfaces) and the external one
occurring between the external surfaces and the
surrounding (composed of the sky and the
surrounding objects):
The internal IR flux calculation is based on the zone
mean radiant temperature concept. This is equivalent
to use the following view factors:
t
j
ij A
A
f= (23)
If the indexes i and j represent surfaces within the
same air zone;
0=
ij
f otherwise. (24)
Radiant powers from equipments and internal heat
gains are treated within the internal IR exchange and
are distributed to the zone surfaces according to their
areas.
The external IR fluxes are calculated using the fictive
sky temperature that can be estimated using
correlations available in the literature (Matrin &
Berdahl, 1984). The net IR flux between an external
surface and the sky is:
)( sskyskyrad TTfh −××=
ϕ
(25)
An inclined planar surface with slope θ beyond the
horizontal plane has a sky view factor given by:
2
)cos(1
θ
+
=
sky
f (26)
The external IR fluxes due to surrounding objects are
calculated assuming the surrounding temperature to
be equal to the ambiant temperature.
)()1( sextskyrad TTfh −×−×=
ϕ
(27)
- 527 -

Solar radiation
The solar radiation is composed of two parts: beam
and diffuse. Solar radiation measurements can have
several formats. This model uses normal solar beam
radiation and solar diffuse radiation.
The position of the sun in the sky is defined by the
zenith β and the azimuth. γ. Both are necessary to
model the beam solar radiation. It can be calculated
using trigonometric formulas available in the
literature for example from ASHRAE Handbook of
Fundamentals (ASHRAE, 1972). From sun position,
slope and azimuth of each of the building facets, the
solar radiation fluxes on the building facets can be
calculated.
The beam solar radiation is divided into two parts:
1. Beam solar radiation absorbed on the
external surfaces of the building;
2. Beam solar radiation transmitted through
windows into the building zones.
The transmitted beam radiation is distributed on the
internal surfaces of the building. The building
description for the model is not based on geometrical
data. The user has thus to specify coefficients that
determine the distribution of solar radiation to the
particular internal surfaces. For example the whole
beam solar radiation can be injected on the floor. The
reflected part of the beam radiation is assumed to be
diffuse so it is treated by the diffuse solar radiation
model.
The diffuse solar radiation on an external surface of
slope θ is composed of two parts:
The first part coming from the sky:
dskyd
ϕ
θ
ϕ
×
+
=
2
)cos(1
, (28)
The second part is the reflected solar radiation from
surrounding objects:
ggrd
ϕρ
θ
ϕ
××
−
=
2
)cos(1
, (29)
The global radiation on the horizontal plane is then
calculated by:
)sin(
βϕϕϕ
ndg += (30)
The total diffuse solar radiation (transmitted through
windows to a zone + the reflect part from the beam
solar radiation) is distributed to the internal surfaces
using the following distribution factors (Trnsys,
2000):
∑−
=
jjj
ii
id A
A
f)1(
,
ρ
α
(31)
MODEL STRUCTURE
In this section, the link between the thermal
phenomena is described. The graphical and modular
methodology of Simulink is used to structure the
model using layers (subsystems).A full presentation
of this structure will be cumbersome so we’ve
restricted it to the most important features.
Structure of the building model
Figure 2 shows a simplified representation to the
model first layer in a top to down approach. This
layer has two main blocks: Building envelope and air
zones.
The block “building envelope” calculates surface
temperatures for all the building surfaces (walls and
windows) depending on weather data, blind positions
of windows, radiant fluxes due to equipment and
internal gains as well as the air temperatures for the
building zones.
The block “air zones” calculates the air temperature
in the zones depending on the building surface
temperatures and all convective powers in each of the
zones.
Figure 2 Simplified representation of the model
structure’s first layer;
Ventilations and interzone airflows are integrated in a
separate block (“Zones ventilations & Zones
airflows”. This allows variable values for all flow
rates: a direct integration into the air zones matricial
equation would make the matrices A in equation (1)
time dependent which would need the calculation of
A at each time step. The approach to calculate the
heat fluxes due to ventilation separately allows to
consider these fluxes in the matrix B for the input
vector U (equation (1)).
Structure of the block “building envelope”
The building envelope is shown in figure 3. The solar
processor block calculates the absorbed solar fluxes
for all surfaces (windows and walls). These fluxes are
input to the windows and walls blocks. The other
inputs, external and internal air temperatures and IR
fluxes for all surfaces are fed back from a separate
block “IR exchange” and the air zones block
- 528 -

Figure 3 Simplified representation of the building
envelop layer;
Link between the different blocks
When modelling with Simulink it is convenient to
avoid algebraic loops because they slow down the
model and can sometimes provoke non-convergence.
An algebraic loop occurs in a model when the inputs
of a block depend on its own outputs through a direct
feedback. The problem can thus not be solved in a
straightforward way and an iteration process is
needed. In the case of the windows model which is
based on a steady state heat balance, an algebraic
loop is created (the window surface temperatures are
function of the IR fluxes on the windows surfaces
which at the same depend on the windows surface
temperatures). A transient model would break this
loop but would lead to a model of higher order. The
problem has been solved by decoupling both models
by a so called “memory block” at the surface
temperatures outputs for the windows model. This
feeds back the value of the window surface
temperature to the IR from the previous time step.
Since the model is used mainly for control
applications with a small time step this decoupling
has a negligible impact on simulation results.
Another interesting point concerning the windows
model is the way to deal with the variable geometry
of the window due to the variable positions of sun
blinds. As mentioned previously, variable matrices
are not convenient for Simulink blocks (except for
the S-function block). To avoid this difficulty the
window is split into two representative parts: a clear
part and a shaded part of 1m2 each. The resultant
surface temperature of the whole window is then
determined by equation (20). In this way the variable
configuration problem has been split down into two
invariable configuration problems.
BUILDING DESCRIPTION INTERFACES
The Simulink model as described in the previous
sections demands two kinds of parameters:
- General building description data:
An independent graphical user interface
“Simbdi” (Simbad Building Description
Interface) has been developed in Visual Basic
environment. This interface allows the user to
draw the building floor by floor in 2 dimensions.
The interface automatically generates adjacency
between zones of successive floors. The user can
define its own wall types or uses those from the
interface library. The interface finally generates
an xml file.
This xml building description file could also be
written manually but becomes cumbersome when
the building has several air zones.
This file is read in the Simulink model mask,
translated to a text file by an automatic translator
and finally used to define the matrix parameters
necessary for the Simulink model.
- Project and system related data
Project and system related parameters are to be
specified by the user in a second interface, the
block mask of the Simulink building model
block. The parameters to be specified are eg. The
geographical location of the building, ground
reflectance, choices of model calculation modes
(eg. Walls models) as well as interzonal air flow
parameters.
This double approach for the interface has been
chosen in order to separate general building
characteristics that do not depend on geographical
location or other project related parameters. The
building, once described, can then be transferred to
other locations without the need for modification.
MODEL VALIDATION
In a first step, the model has been confronted to a
series of simple (but important) validation cases: for
example, the simulated building was set in an
ambiance that has a given constant external
temperature without solar radiation (the sky
temperature was set equal to the external one). The
steady state air temperature of the building zones was
found equal to the outside one.
In further steps the new model is planned to be
validated on other cases such as BESTEST.
In this paper, first comparisons with TRNSYS “Type
56” multizone model (Trnsys, 2000) have been
carried out. This series of tests is not completely
finished yet, but there was good agreement for all
tested cases. An exemplary test is shown in the
following section. Validation work is still ongoing.
Some differences between Trnsys type 56 and
Simbad
It’s interesting to mention some main differences
between Trnsys “Type 56” and the Simbad multi-
zone model:
1. Trnsys uses transfer function to simulate
walls heat conduction, while Simbad uses a
combined finite difference/finite volume
scheme.
- 529 -

2. Trnsys uses the so called “star temperature”
(Seem et al., 1987) to simulate IR exchange
for internal surfaces, while Simbad uses the
common mean radiant temperature.
3. Trnsys uses a detailed model for windows
that reads output data from the WINDOW
4.1 program (WINDOW4.1, 1994). Trnsys
takes thus into account the variation of solar
transmittance and absorptance with the
incidence angle and uses hemispherical
transmittance and absorptance for the diffuse
radiation. Simbad windows model is rather
simple and uses the same constant solar
transmittance and absorptance for both beam
and diffuse radiation.
The simulated case
One comparison case is presented here. This case
consists of a building (cf. figure 4) composed of two
zones. Ventilations, airflows between zones,
equipments power, and internal gains are set to zero.
Figure 4 Simulated building - plan and section view;
The building walls are identical to those used in the
BESTEST cases (Judkoff & Neymark, 1995). A
heavyweight construction is used for exterior wall,
floor and roof. The internal wall is taken from the
case 960.
The 3 windows are similar and are composed of a
single glass panel without frame and spacers. The
glass pane has the properties given in table 1 below.
The hemispherical values were used for the Simbad
model.
A weather data file for Madison from Trnsys package
was used. This file contains the following weather
data:
1- Normal direct solar radiation;
2- Total solar radiation on horizontal plane;
3- Outside air temperature;
4- Absolute humidity;
5- Wind velocity (not used);
6- Wind direction (not used).
Table 1 pane properties of the window;
PANE PROPERTY VALUE
Thickness 4.0 mm
Thermal conductance 225 W/m²K
Normal solar transmittace 0.83
Normal solar absorptance 0.095
Hemispherical solar transmittance 0.749
Hemispherical solar absorbtance 0.106
A slight modification was needed on the Simbad
model to accept the solar radiation of the Trnsys
format. The simulation was run for one summer
week. Figures 5 & 6 show air temperature and solar
radiation for this week.
020 40 60 80 100 120 140 160
10
15
20
25
30
35 Madison: Air temperature
Temps [h]
Temper ature [°C]
Figure 5 Air temperature during the simulated week;
020 40 60 80 100 120 140 160
0
200
400
600
800
1000
1200
Temps [h]
Solar flux [W/m²]
Madison: Solar radiation
Global horizontal radiation
Normal beam radiation
Figure 6 Solar radiation for the simulated week;
Simulation results
The air temperature is used to compare the outputs of
the two simulation tools. As the window model in
- 530 -

Trnsys is more detailed than the Simbad one
ragerding radiation phenomena, two cases are
presented here:
1- Case with solar radiation (cf. figures 7 & 8);
2- Case without solar radiation (cf. figures 9 &
10).
Table 2 shows the maximum and the mean value of
the absolute difference between Trnsys and Simbad
results. As we can expect the difference between the
two simulation tools is higher when there’re solar
radiation since the window model is more detailed
into Trnsys but they steel close to each other.
020 40 60 80 100 120 140 160
18
20
22
24
26
28
30 Zone A: air temperature
Temps [h]
Air temperature [°C]
Simbad
Trnsy s
Figure 7 Simbad versus Trnsys – air temperature for
Zone A (case without solar radiation);
020 40 60 80 100 120 140 160
18
20
22
24
26
28
30 Zone B: air temperature
Temps [h]
Air temperature [°C]
Simbad
Trnsy s
Figure 8 Simbad versus Trnsys – air temperature for
Zone B (case without solar radiation);
Table 2 Mean and maximum values of the absolute
difference between Trnsys and Simbad results;
ZONE SOLAR
RADIATION MEAN
VALUE MAX.
VALUE
A off 0.1°C 0.2 °C
B off 0.11 0.22
A on 0.23 0.47
B on 0.26 0.58
020 40 60 80 100 120 140 160
18
20
22
24
26
28
30 Zone A: air temperature
Temps [h]
Air temperature [°C]
Simbad
Trnsy s
Figure 9 Simbad versus Trnsys – air temperature for
Zone A (case with solar radiation);
020 40 60 80 100 120 140 160
18
20
22
24
26
28
30 Zone B: air temperature
Temps [h]
Air temperature [°C]
Simbad
Trnsy s
Figure 10 Simbad versus Trnsys – air temperature
for Zone B (case with solar radiation);
CONCLUSION
This paper has shown a new multizone building
model for Simulink. The originality of this model, the
integration of a multizone building case in the
graphical Simulink environment has been explained
in a first step. The main equations as well as the
structure of the new model in Simulink environment
have been described.
A brief presentation of the related ghraphical
interface for the building description has been given.
In a last step, the paper presents first validation
results, a comparaison with Trnsys simulations. The
results are very promising, but further validations are
in progress in order to ensure the validity of the
model.
Due to the graphical Simulink environment, the
model can be linked to other blocks. A large number
of HVAC systems can thus be simulated (eg.
Addition of air flow pressure networks for VAV or
ventilation systems).
Due to the small time steps that can be used in
Simulink and the developed model, the new model is
very suitable for the study and development of
control algorithms.
- 531 -

NOMENCLATURE
Romaine letters
A area (m²).
A matrix coefficient of the state vector.
B matrix coefficient of the inputs vector.
Cp heat capacity (J/kg.K).
f matrix of blind positions.
f view factor between surfaces.
f beam solar radiation distribution factor (-).
h linearized heat transfer coefficient (W/m²K).
k thermal conductivity (W/mK).
L total thickness of the wall (m).
m
& mass flow rate (kg/s).
P cooling or heating load (W).
P gain, power (W).
t time (s).
T temperature (°C).
U inputs vector.
V zone volume (m3).
x abscise belong the wall thickness (m).
X state vector.
Greek letters
α thermal diffusivity (m²s-1).
α solar absorbtivity (-).
β zenith angle of the sun (rad).
γ azimuth angle of the sun (rad).
∆t Timestep (s).
∆x space step (m).
ε mean IR emissivity for the building surfaces.
θ slope with respect to the horizontal plane
(rad).
ξ stability safety factor for wall discretization.
ρ density (kg/m3).
ρ reflectivity (-).
σ Stefan-Boltzman constant for black body =
5.67×10-8 Wm-2K-4.
ϕ heat flux (W/m²).
Indices
a air
az adjacent zone
b boundary
c convective, convection
cl clear part (for the window)
cpl coupling between zones
d diffuse horizontal solar radiation
eq equipments
ext exterior (air)
g gains
g global horizontal solar radiation
gr ground
i zone, surface, ventilation, coupling number
IR infrared radiation
j layer, surface number
m mean
n total number of layers in the wall
n normal beam solar radiation
p power
s surface
rad radiative
sh shaded part (for the window)
sky relative to the sky
solar solar radiation
t total
v ventilation
w wall
wn window
z zone
ϕ flux
1 first face of the wall
2 second face of the wall
REFERENCES
ASHRAE, 1972. ASHRAE Hundbook of
Fundamentals, American Society of Heating,
Refrigerating, and Air Conditionning Engineers.
Judkoff, R., J. Neymark. 1995. International Energy
Agency Building Energy Simulation Test
(BESTEST) and Diagnostic Method. NREL/TP-
472-6231. Golden, CO: National Renewable
Energy Laboratory, USA.
Laitila PK., Kohonen RO., Katajisto KI. Et al. 1991
An emulator for testing HVAC systems and their
control and energy management systems,
ASHRAE trans.1991, vol.97, paper number NY-
91-9-2, 679-683, 8 figs., 1 tab., 3refs.
Martin, M., Berdahl P. 1984. Characteristics of
Infrared Sky Radiation in the United States,
LBL, University of California – Berkeley, Solar
Energy Vol.33, No.3/4, pp. 321-336.
Matlab, 2004. Matlab. Version 7.0 Mathworks Inc.,
Ma., USA.
Seem, J.E. 1987. Modeling of Heat in Buildings, Ph.
D. thesis, SEL, University of Wisconsin,
Madison USA.
SIMBAD, 2003. SIMBAD Building and HVAC
Toolbox, Version 3.1, CSTB, France
Simulink, 2004. Simulink dynamic System
Simulation for Matlab. Version 6.0, Mathworks
Inc., Ma., USA.
Trnsys, 2000. Trnsys: a transient system simulation
program. SEL, University of Wisconsin,
Madison USA.
Vaézi-Nejad et al, 1991. The use of building
emulators to evaluate the performance of
building energy managment systems , BS1991 ,
91 pp. 209-213.
WINDOW4.1, 1994. WINDOW4.1, PC Program for
Analysing Window Thermal Performance in
Accordance with Standard NFRC Procedures.
Energy and Environnement Division, Lawrence
Berkeley Laboratory, CA 94729 USA.
- 532 -