A SIMPLE BUILDING MODELLING PROCEDURE FOR MATLAB/SIMULINK

G. Hudson & C. P. Underwood

University of Northumbria
Department of the Built Environment
Newcastle upon Tyne
UK

ABSTRACT capacity followed by a slower fall in room

A computationally-efficient building thermal model temperature associated with the structural mass.
is developed for short-timescale investigations
applicable to control system design. A lumped- Levermore (1992) develops a heating system
capacity treatment of the building elements is used optimiser model that considers the building as a first
which can be summarised by a simple analogue order description. In this model all the building
electric circuit. The method is procedurally- fabric initial temperature are assumed to be variant.
transparent and leads to a state-space description of a The various construction elements are assumed to
building space which is implemented using warm up and cool down to their relative steady-state
MATLAB/SIMULINK. The model is tested using temperatures. The internal structure is considered to
experimental data from a building with high thermal be at the same temperature as the internal air and any
capacity. The accuracy of second-order descriptions internal resistance to heat flow is ignored.
of the high thermal capacity building elements Levermore (1988) states that a typical two time
compared with first-order descriptions is also constant model has a short time constant associated
examined. Good agreement with experimental data with the air, windows and doors and a dominant time
over short simulation periods is obtained using first- constant associated with the fabric. The dominant
order element descriptions. time constant was considered to be the controlling
factor. This assumption effectively ignores the short
INTRODUCTION time constant and is considered adequate for
This work details the development of a building temperature prediction for optimiser controls but is
model for coupling to a convective heat emitter clearly limited when investigating high frequency
model for the primary purpose of investigating problems in which the heating system dynamics are
control system design. The building model was required to play a part.
required to adequately reflect the interaction of the
emitter and the space (and include the effects of the McLaughlin et al (1981) in discussing cooling
external climate). Thus the model would investigate response curves state that buildings exhibit two or
typically shorter periods of time than normally three time constants. In contrast to Levermore and
associated with many of the building thermal models Lorenz and Masy the shorter time constant is claimed
available today. A lumped capacitance model was to be due to the cooling of the air and the heat emitter
investigated allowing interaction with the structure and not the air capacity alone.
and external climate without dominating the
modelling process. The model treats radiative and Tindale (1993) refined the basic Lorenz and Masy
convective heat inputs in the same way which is model to address some of the reported inadequacies.
considered acceptable since the associated emitter One of these inadequacies was the poor modelling of
model (not reported here) is fully convective. buildings with very high thermal capacity. As many
modern buildings have these characteristics, or are
“thermally heavyweight” as defined by the CIBSE
SIMPLIFIED BUILDING THERMAL
response factor method (CIBSE, 1986), it was
MODELS considered important that the building model
Lorenz and Masy (1982) describe a simple lumped proposed in the present work adequately modelled
capacity model which has since been frequently used. thermally heavyweight spaces.
This is a two time constant model, one time constant
associated with the air mass and the other associated
LUMPED CAPACITANCE MODEL
with the internal and external structural mass. The
The Lorenz and Masy model achieves its simple two
two time constants describe the traditionally accepted
time constant form by including a proportion of the
form of building cooling profile, a relatively rapid
internal structure (floors, ceilings and partitions) in
fall in room temperature associated with the air
1
the single lumped external structural capacity. This
simplification had benefits at the time of
development, but the advent of cheap computing
power has diluted the case for very simplified
models. Hence the model used in this work is based
on the description first proposed by Lorenz & Masy,
subsequently applied by Crabb et al (1987) and, later,
by Tindale (1993). A similar approach has also been
taken by Bérnard et al (1985). The difference here is
that simplifications regarding the treatment of
internal construction elements which were adopted
by the previous workers were not assumed in the
present work. Our approach is similar to that used by
Achterbosch et al (1992) though their work was
based on housing whereas the present work is relates
to commercial-scale buildings with high thermal
capacity.

The procedure described in the following considers a

room model consisting of external wall elements and
internal floor, ceiling and partition walls. This is the
same as the room from which experimental data
(described later) have been obtained. However, the
method is sufficiently transparent for alternative
room formats to be easily considered. An analogue
circuit form of interpretation is shown in Figure 1.

Whilst the thermal resistances (hence thermal

transmittances) and thermal capacitances can be
calculated trivially, the Lorenz and Masy (1982)
prescriptions for the single capacity equivalent of a
multi-layer construction have been used to calculate
the split between inner and outer region thermal
resistances for the external walls – the “accessibility
factor”.
Equations (1)-(6) can be readily stacked using the
Based upon Figure 1, energy balances about each state-space notation:
internal temperature node give rise the to following.
& = AT + Bi
T

in which T & is a vector of derivatives; A,B are

matrices of coefficients; T is a vector of states and i a
vector of inputs (Equation (7)).

2
Figure 1: Lumped Capacitance Model

3
MODEL IMPLEMENTATION INITIALISATION
The model is implemented using the Simulink Model implementation was initially hampered
structure of Figure 2. Inputs are heat supplied by the choice of initial conditions. Clearly, the
by the plant, Qp , internal “casual” heat gains, initial temperature for the room air can be set
routinely but initial temperatures for structural
Qe , solar radiation through glazing, Qs , and
elements, especially internal structural
external air temperature, Tao . A state-space elements, required further consideration.
block from the Simulink “linear” library These considerations are not important for
represents the building model. The derivatives long term simulations since the inevitable
were integrated using a 4th -order Runge-Kutta initial errors have limited impact on later
scheme and an integration interval of 0.1hr results. However, for short-term dynamics for
was found to work well. which this model was primarily designed the
impact of inappropriately-set initial conditions
would have an unacceptable influence on the
accuracy of results.
Qp x' = Ax+Bu
Mux
Plant y = Cx+Du Mux
The initial conditions for internal structural
load Mux-i State-space
Mux-o Scope elements were especially problematical. The
building
room from which the experimental results
[Tao]
were obtained was heated by a fully convective
External system which would normally imply internal
temperature structure temperatures lower than the room air
temperature. However, this was not found
[Tai_expt] from field monitoring. Over a ten-day period
the air and mean radiant temperatures were
Experimental
internal recorded in the test room. The mean radiant
temperature temperature was measured using a globe
thermometer, which was situated at one of the
occupants working positions, approximately at
Figure 2: Simulink Model Structure the centre of the room. Air velocity results
were obtained and the low mean values
Sample model data were taken from a north- observed (<0.1ms -1) indicated that the globe
facing corner room in a University of temperature would give an accurate indication
Northumbria campus building - the of mean radiant temperature. The
Northumberland Building. Besides the room’s measurements were taken between March –
major northern exposure (with substantial April 1998 and are shown in Figure 3.
glazing), there is a west facing wall with minor
well-shaded glazing. The building itself is the 32
Mean Radiant Air
subject of extensive demonstration research 30
into photovoltaic cladding and, consequently, 28
is extensively monitored. Further details can
26
be found in Hudson & Underwood (1996).
C

24
o

Model parameters were calculated using 22

standard property data for the construction
20
materials involved (CIBSE, 1986). The air
change rate used in the ventilation 18
0 20 40 60 80 100 120 140 160 180 200 220
transmittance was based on an earlier oil-tracer Time (hours)
test which indicated an average air change rate
of 0.5hr-1 under typical mild weather
conditions for the room considered. Figure 3: Air and Mean Radiant Temperatures

4
The variation between mean radiant and air Results are given in Figures 4, 5 and 6
temperature is small under cyclic conditions of respectively.
room heating and cooling. These results
22
indicate that setting the initial temperature of
the internal structure at the room air 20
temperature at the start of the cooling period is
18
not unreasonable. This is the same assumption Tai (model)
as Levermore uses in his whole building single 16
Tai (experiment)

C
time constant model (Levermore (1992)).

o
14 Tao
Note that this comparison was made in a north
12
facing room under typically overcast
conditions. This conclusion may well be 10
invalid for rooms experiencing significant 0 2 4 6 8 10
solar heat gain. Time (hours from plant off)

Over the short timescale considered, the

initialisation of the structural elements’ Figure 4: Cool-down Response
temperatures was found to be less significant.
Fixing this temperature at the mean of the
22
internal air and external air temperatures was
found to work very well in relation to model 20
and experimental results. This would clearly 18
be the correct choice for an homogenous slab 16
at (initial) steady-state conditions. The error in Tai (model)
14
this assumption would be dependent on the
C

Tai (experiment)
o

thermal-capacitance-mix of materials present. 12 Tao

For most external building elements which are 10

dominated by high thermal capacitance 8
materials, the error is likely to be minor.
6
0 1 2 3 4 5
VALIDATION Time (hours from plant on)
The model of Figure 1 was initialised as
described above and applied to the north
facing test room. The room itself was chosen
so that basic model validation could be Figure 5: Preheating Response
investigated by neglecting solar radiation and
monitoring data were chosen from periods in
which the room was not in use (i.e.
Qs = Qe = 0 ) . Monitoring data were
collected in January 1997, a predominantly 24
Tai (model)
overcast period at the test site. 22 Tai (exp't.)
20 Tao
Three model validation trials were carried out: 18
16
Plant on full
• Cool-down (room temperature response 14
C
o

to plant off). 12
• Preheating (room temperature response to 10
plant on). 8
• Cool-down/heating/cool-down (plant 6
fully on for a short period between 4
0 5 10 15 20 25 30 35 40 45 50 55
periods of cool-down) Time (hours)

The plant capacity, Qp , in the test room was

known to be 3.6kW. No account was taken of
the plant dynamics (in all trials, the room Figure 6: Cool-down/Heating/Cool-down
heating was either fully on or fully off). Response
5
 The model gives good agreement with CONCLUSIONS
experimental data for the short-term dynamics A simple thermal model has been developed
considered and for both heat-up and cool-down specifically for the analysis of short-term
conditions. Towards the end of the longer of dynamics in buildings of high (or low) thermal
the three trials (Figure 6), the model and capacity and has been implemented as a state-
experimental results are seen to begin to space block using MATLAB/Simulink. The
diverge slightly (>45hr). It is likely that this model is procedurally-transparent, easy to
divergence would continue as a result of the interpret and can be initialised with reference
compromised accuracy of the lumped capacity to internal and external air temperatures. The
method. Evidently, the lumped capacity model has been applied to a building
method works well for short-term dynamics of exhibiting high thermal capacity and results
special interest to control loop design, but compare very favourably with test data taken
accuracy for long term investigations (certainly from the building for relatively short time
involving high thermal capacity buildings) is periods and under both heating and cooling
uncertain. conditions. A comparison between second-
order and first-order representations of the
SECOND-ORDER ELEMENTS higher thermal capacity building elements
The high thermal capacity paths consisting of present has revealed that there is no advantage
the two external walls and ceiling (Figure 1) to be gained by using the higher-order
were re-modelled using second-order lumped description as far as short-term dynamics are
capacities by splitting the internal resistances concerned.
in two and sharing the capacitances equally.
This was done to establish whether the first- Further work needs to be done on heating
order descriptions for these elements was in system modelling, the treatment of short-wave
any way dubious in terms of accuracy. and long-wave radiation and the validity of the
Results, which are based on the same heating model (if necessary with higher-order
transient for the preheating case above, are elements) for longer term simulations.
given in Figure 7. The results show that there
is no appreciable advantage in using the higher REFERENCES
order description for short-term transient Achterbosch, G. G. J., de Jong, P. P. G., Krist-
analysis. Spit, C. E., van der Meulen, S. F., Verberne, J
(1985). The development of a convenient
22
thermal dynamic building model Energy and
First-order model
Buildings 8, pp183-196.
20 Second-order model
Bérnard, C., Guerrier, B., Rosset-Louërat, M. –
18 M. (1992) Optimal building energy
management: Part I – Modelling American
C

Society of Mechanical Engineers – Journal of

o

16
Solar Energy Engineering, 114, pp2-12.
14 Plant On
CIBSE Guide Section A3: Thermal Response of
12 Building Structures (1985) Chartered
0 2 4 6 8 10 12 14 16 18 20 22 24
Institution of Building Services Engineers,
Time (hours)
London.

Crabb, J. A., Murdoch, N., Penman, J. M.

Figure 7: First and Second-order Model
(1987) A simplified thermal response model
Comparison
Building Services Engineering Research &
Technology, 8 (1), pp13-19.

Hudson, G., Underwood, C. P. (1996) Effect

on energy usage of over-capacity and emission
characteristics in the space heating of buildings
with high thermal capacity Proc.
CIBSE/ASHRAE Joint National Conference,
Harrogate, UK, pp369-378 Chartered
6
Institution of Building Services Engineers,
London.

Levermore, G. J. (1992) Building Energy

Management Systems E & FN Spon, London.

Levermore, G. J. (1988) Simple model for an

optimiser Building Services Engineering
Research & Technology, 9 (3), pp109-116.

Lorenz, F., Masy, G. (1982) Méthode

d’évaluation de l’économie d’énergie apporttée
par l’intermittence de chauffage dans les
bâtiments. Traitment par différences finies
d’un modèle à deux constantes de temps.
Facuite de Sciences Appliquées , Universite dé
Liège, Belgium (in French).

McLaughlin, R. K., McLean, R.C., Bonthron,

W. J. (1981) Heating Services Design
Butterworth, London.

Tindale, A. (1993) Third-order lumped-

parameter simulation method Building
Services Engineering Research & Technology,
14 (3), pp 87-97.

NOMENCLATURE
T temperature (o C)
Q heat flux (W)
p fraction of solar radiation entering floor
A surface area (m-2)
U thermal transmittance (Wm-2K-1)
Ui inner thermal transmittance (Wm-2K-1)
Uo outer thermal transmittance (Wm-2K-1)
Uv ventilation thermal transmittance (WK-1)

Subscripts:
w1 external wall 1
w2 external wall 2
f floor
c ceiling
ip internal partition
ai air – internal
ao air – external
s solar
g glazing
e casual sources (internal heat gains)
p plant