Content uploaded by Mohamed Hamdy
Author content
All content in this area was uploaded by Mohamed Hamdy
Content may be subject to copyright.
..............................................................................................................................................................
.............................................................................................................................................
................................................................................................................................................................................
Optimum design of a house and its HVAC
systems using simulation-based optimisation
Mohamed Hamdy, Ala Hasan and Kai Siren*
School of Technology, Aalto University, PO Box 14400, FIN-00076 Aalto, Finland
*Corresponding author:
kai.siren@tkk.fi
Abstract
This paper deals with a multi-objective optimisation problem where the objective is to minimise CO
2
-
eq emissions and investment cost of a multi-zone single-family house and its heating/cooling systems.
Eight design variables are subjected to study: level of building tightness, insulation thickness of the
external wall, floor and roof, type of window glazing, window shading, heat recovery and heating/
cooling systems. Simulation-based optimisation is implemented to minimise the objective functions by
finding optimum values of the design variables. This is done by combining the dynamic building
performance simulation program IDA-ICE 3.0 with a modified genetic algorithm. The optimisation
results give the optimal solutions for the problem in the form of a Pareto front, showing the trade-off
between the two objectives. The obtained solutions are much better than the initial designs of the
house in terms of lower CO
2
-eq emissions and investment cost. It is noted from the obtained results
that the significance of the heating system is higher than the other design variables so that the optimal
solutions can be classified according to the type of the heating system. It is also noted that there is a
need to include a thermal comfort criterion as a problem constraint to limit overheating hours during
summer.
Keywords: building energy optimisation; space heating; CO
2
emissions; investment cost
Received 15 March 2010; revised 17 March 2010; accepted 19 March 2010
1INTRODUCTION
The building sector accounts for 40% of the world’s total end
energy consumption. Much of this energy is required for heating
and cooling of buildings. Energy for heating and cooling are the
main reasons for the associated CO
2
emissions. To reduce such
emissions, investment has to be made, in terms of better features
of the building envelope and heating, ventilation and air con-
ditioning (HVAC) system type and components. With numerous
options for those parameters, it is a very difficult task to find
cost-effective designs of the house and the HVAC system. The sol-
ution is to make a simulation-based optimisation, which is utilis-
ing optimisation combined with building energy performance
programmes, to minimise the defined objective functions by
finding the optimal values of selected design variables.
To achieve this, a multi-objective optimisation problem has
to be formulated for the design of a house and its heating/
cooling systems. In the reported case, eight design variables
were chosen. The optimisation aims at minimising the
energy-related CO
2
-eq emissions and the investment cost
simultaneously.
2BUILDING DESCRIPTION (INITIAL
DESIGN)
A typical Finnish two-floor semi-detached house, located in
Helsinki, Finland, is considered as a case study. The total floor
area of the house is 143 m
2
. The internal height is 2.5 m. The
two floors are connected by a staircase. In the initial design,
the construction materials are selected in order to achieve heat
transmission coefficient (U)-values (in W/m
2
K) equal to the
maximum values stated in the Finnish building code
(C3-2007). The internal gains due to people, lighting and elec-
tric appliances are assumed according to annual values speci-
fied by the Finnish building code (D5) and inserted in the
calculation as a profile with hourly values.
There are heating and cooling units in the building to cover
the heating and cooling demand. An air-handling unit (AHU)
is supplying fresh air to bedrooms and living room and
exhausting air from bathrooms and kitchen. The AHU heater
keeps the supply air temperature at 188C when the incoming
outdoor air temperature is lower than this temperature. The
average exhaust air flow from the whole house is equal to 0.65
International Journal of Low-Carbon Technologies 2010, 00, 1– 5
#The Author 2010. Published by Oxford University Press. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org
doi:10.1093/ijlct/ctq010 1of 5
International Journal of Low-Carbon Technologies Advance Access published April 21, 2010
at Helsinki University of Technology on April 26, 2010 http://ijlct.oxfordjournals.orgDownloaded from
air change per hour. The heating system keeps the air tempera-
ture at a lower set-point of 218C. There is a heat exchanger for
heat recovery from exhaust air with an yearly efficiency of
70%. Each floor has a total fixed glassing area of 5 m
2
.In
addition, a small openable window (0.9 m
2
) exists, which pro-
vides natural ventilation for summer cooling. It is assumed
that the building tightness at the initial design is n
50
¼4 1/h,
where n
50
is the number of air changes per hour at 50 Pa
pressure difference across the building envelope.
The building energy simulation is carried out using
IDA-ICE 3.0 software [1] and Helsinki-2001 hourly weather
data. To reduce the simulation execution time, the building
model is simplified into three zones (upper floor, lower floor
and staircase). The simplified building model needs about two
and a half minute to execute a 1-year simulation run of the
house and its HVAC system.
3THE OPTIMISATION PROBLEM
3.1 Design variables
In the current study, eight design variables are chosen. Table 1
presents the initial values, lower bounds, upper bounds and
types (discrete or continuous) of the eight design variables. For
the continuous variables, Xrefers to the insulation thickness
(m). Tables 2–6give details of the discrete variables.
3.2 Objective functions
The aim of this study was to achieve a cost-effective low-
emission design of the building. Therefore, CO
2
-eq emissions
of the heating energy and the investment cost were selected as
two objective functions to be minimised. The heating energy
consumption Q(in kW h/year) consists of the energy required
for space heating, system heat losses and domestic hot water.
The first objective, CO
2
-eq emissions (kg/a), is calculated by
the following equation, using the primary greenhouse gas
emission factors, EF (in kg CO
2
-eq/kW h),
CO2-eq emission ¼QEF
h
ð1Þ
where Qis the total heating energy consumption,
h
is effi-
ciency of the heating system (Systems 1–3) or the coefficient
of performance (COP) of the heat pump (Systems 4 and 5).
For sources of emission in Finland, the values of EF for the
studied systems are taken from [2] and shown in Table 6.
Emissions considered are related to three major greenhouse
gases: CO
2
, sulphur and nitrogen emissions.
The second objective, the investment cost, is the total cost
related to the eight design variables. Different types of
Table 1 Design variables.
Variable Type Initial
design
Lower
bound
Upper
bound
X
wall
(m) Continuous 0.124 0.024 0.424
X
roof
(m) Continuous 0.210 0.110 0.510
X
floor
(m) Continuous 0.140 0.040 0.440
Windows type Discrete 1 1 5
Heat recovery type Discrete 2 1 3
Shading type Discrete 1 1 2
Building tightness level Discrete 1 1 5
Heating/cooling system type Discrete 5 1 5
Table 2 Window.
Ty p e U-value (W/m
2
K) SSC factor ( – ) Price (E)
1 1.4 0.656 180
2 1.1 0.656 185
3 1.0 0.530 205
4 0.85 0.482 240
5 1.1 0.437 210
SSC, short-wave shading coefficient.
Table 3 Ventilation unit.
Type Efficiency Specification (%) Price (E)
1 60 Plate heat exchanger 3172
2 70 Rotating wheel 3443
3 80 Rotating wheel 3715
Table 4 Shading [6].
Type MSC MSSC Description Price
1 0.14 0.09 External blind, horizontal laths 200 E/m
2
2 1.0 1.0 No shading 0
MSC, multiplier for shading coefficients of window (– ); MSSC, multiplier
for short-wave shading coefficients (–).
Table 5 Building tightness.
Ty p e n
50
(1/h) Price (E/m
2
)
1 4.0 0
2 3.0 5
3 2.0 12
4 1.0 22
5 0.5 30
Table 6 Heating/cooling system.
Type System Price (E/m
2
) EF (kg/kW h)
h
or COP
1 Direct electric radiator 30 0.459 1.0
2 Oil fire boiler 93.93 0.267 0.9
3 District heating 101.07 0.226 1.0
4 GSHP no free cooling 125.56 0.459 3.0
5 GSHP with free cooling 132.71 0.459 3.0
GSHP, ground-source heat pump; EF, greenhouse gas emission factor;
h
,
efficiency.
M. Hamdy et al.
2of 5International Journal of Low-Carbon Technologies 2010, 00, 1 – 5
at Helsinki University of Technology on April 26, 2010 http://ijlct.oxfordjournals.orgDownloaded from
insulation are used in the external wall, roof and ground floor,
which have prices of 56.3, 32.5 and 100 E/m
3
, respectively. The
prices of the other design variables are presented in Tables 2–
6, and are mainly from [3].
4OPTIMISATION AND SIMULATION
A large number of design variables usually require a large
number of simulation-runs to get feasible solutions in a multi-
objective optimisation problem. Thus, in order to perform the
optimisation process in shorter times, a modified multi-
objective genetic algorithm (PR_GA) is implemented. It is a
two-phase, multi-objective optimisation solver that works under
MATLAB environment and was developed by the authors.
Details about PR_GA can be found in [4]. Briefly, in PR_GA,
the genetic algorithm (GA) from MATLAB 2008a Genetic and
Direct Search Toolbox [5] was modified to be able to deal with
discrete and continuous variables. Then, it was combined with a
deterministic optimisation algorithm in order to supply GA
with a good collection of individuals as an initial population.
This process is called preparation phase (PR). The major advan-
tage of PR_GA is that it tries to reduce the random behaviour
of GA in an attempt to obtain good solutions with lower
number of simulation iterations. The current study uses PR_GA
combined with IDA-ICE 3.0, whole-building dynamic simu-
lation program, to perform the optimisation process. In the case
under study, 1010 simulation iterations are used: 290 for the
preparation phase and 720 for GA phase (using 40 population
individuals and 18 generations). The execution time was about
44 h on a computer with Windows Vista system (Intel
w
core
TM2 Quad CUP 2.40 GHz processor, 3061 MB RAM).
5RESULTS AND DISCUSSIONS
Altogether 41 Pareto optimal solutions were obtained in
terms of lower CO
2-
eq emissions and investment cost for
the studied house, shown in Figure 1.Thesesolutionsare
classified on the Pareto front according to the heating
system type. There is a remarkable influence of the type of
heating system on the results. The main reason for this is
that EF in equation (1) is directly dependent on the type
of the heating system. Besides, a part of the heating energy
Qis almost constant (those related to the domestic hot
water production and the system loss) and is not changing
by variations in the envelope-related design variables.
Anotherreasonisthattheinvestmentcostoftheheating/
cooling systems is often higher than the cost related to
other design variables.
For comparison with the obtained solutions, four initial
designs are also presented on Figure 1. All the initial
designs have the same values for the design variables
(Table 1) except for the type of the heating system. It is
evident from Figure 1that the optimisation solver suc-
ceeded in providing a set of solutions which have much
lower CO
2
-eq emissions and investment costs compared
with the initial designs.
Figure 2shows the house annual space heating energy, for
the room units and AHU, for the optimal solutions indicated
by Figure 1. From right to left, the solver has found solutions
inside one heating system where the average U-value of the
building is decreasing and the investment cost is increasing,
until it is more economical to switch to another heating
system than continue reducing the U-value. By changing the
heating system, the emission factor is decreasing and more
energy can be used by reducing the insulation level of the
building envelop. This cycle is repeated from right to left as
shown in Figure 2.
The average U-value for the building is defined by the fol-
lowing equation:
Ubldg ¼ðUwall Awall þUroof Aroof þUfloor Afloor
þUwindow AwindowÞ=Atotal ð2Þ
Figure 1. Pareto front and the initial designs.
Figure 2. Annual space heating energy of the optimal designs as a per cent of
the space heating energy of the corresponding initial design.
Optimum design of a house and its HVAC systems
International Journal of Low-Carbon Technologies 2010, 00, 1 – 5 3of 5
at Helsinki University of Technology on April 26, 2010 http://ijlct.oxfordjournals.orgDownloaded from
where Ais the area (m
2
), A
total
¼A
wall
þA
roof
þA
floor
þ
A
window
and U
bldg
is the average U-value for the whole building
envelop (W/m
2
K).
The average U-values of the optimal solutions are shown in
Figure 3. From Figures 2and 3it is obvious that for a lower
U
bldg
lower space heating energy is required. Variations in the
type of heat recovery and building tightness explain the small
deviation in the relation between the results of the two figures.
It is worthwhile mentioning that the average U
bldg
of the
optimal solutions is 0.238 W/m
2
K, where 80.5% of the solutions
have U
bldg
less than that for the initial design (U
bldg
¼0.3 W/
m
2
K). However, reducing U
bldg
,0.17 W/m
2
K seems not to be
the optimal solution whatever the type of heating system.
Figure 3indicates that the optimisation solver preferred to
change the heating system from System 1 (Solution 23, U
bldg
¼
0.17 W/m
2
K) to System 2 (Solution 22, U
bldg
¼0.34 W/m
2
K)
instead of continuing in reducing U
bldg
with System 1 as what
occurred from solutions 41 to 23, knowing that the minimum
bound for U
bldg
in the space solution is 0.131 W/m
2
K.
In addition, at solution (23), building tightness (n
50
¼
1.0 1/h) and heat recovery efficiency (
h
¼70%) were selected,
while the solution space includes higher building tightness
(n
50
¼0.5 1/h) as well as better heat recovery (
h
¼80%),
which provide higher savings in the space heating energy. This
illustrates that changing the type of the heating system from
System (1) to System (2) is a better solution for the two objec-
tive functions than reducing the space heating energy with
System 1. On the other hand, the district heating solutions
(System 3) dominated most of the oil fire boiler solutions
(System 2). The reason is that, there is no considerable differ-
ence in the price between the two systems, while the emission
factor of System 3 is less than that for System 2.
It is to note that the minimum required air temperature
inside the zones, 218C, was maintained in all the optimal
solutions with the different types of envelope parameters and
systems. This is achieved by the controller acting on the
zone-heating units. The higher set-point (248C) of the zone
air temperature is applicable only during the summer season,
when the ground-source heat pump system with free cooling
heat-exchanger (System 5) is selected as a solution, because
it is the only system which offers the cooling option.
However, a procedure is made available in all cases to open
the small window at each zone via a proportional-integral
controller when the indoor air temperature (T
i
,8C) is
.248C and the outdoor air temperature is lower than the
indoor air temperature. This procedure emulates the human
behaviour to open the window and provides natural venti-
lation when overheating occurs.
It can be seen from the results shown in Figure 1that
System 5 was not selected in the optimal solutions. The optim-
isation algorithm preferred to invest in better properties of the
building envelope and in other types of systems, which have
lower price and no cooling, instead of investing in the higher
price system (System 5). Same reason applies for not getting
solutions with window shading. As a result, the indoor air
temperature increased above 248C during summer for the
obtained solutions. In order to evaluate this overheating, a
degree-hour parameter (DH
24
,in8C h) is defined as the sum-
mation of the degrees higher than 248C at the warmest zone at
each hourly time step in a 1-year simulation (8760 h)
DH24 ¼X
i¼8760
i¼1
ðTi24Þdtð3Þ
Ti24 0
where T
i
is the indoor air temperature [8C] and dtis a 1-h
time period.
It is noted that overheating occurred in all the optimal sol-
utions with a range of DH
24
from 4181 to 62548C h, while the
reference value for the initial design case without a cooling system
is 24008C h. Minimum overheating occurred at solution number
41, which has a maximum value of U
bldg
(0.496 W/m
2
K) and
CO
2
-eq emission.
6CONCLUSIONS
By using simulation-based optimisation, the multi-objective
optimisation solver succeeded in finding designs of the house
and its HVAC systems that have lower CO
2
-eq emissions and
investment costs compared with the initial designs. The
heating system seems to have the largest influence on the
objective values and the optimal solutions. Other variables
characterising the insulation level and tightness of the building
envelop as well as the system features have a smaller effect.
Using CO
2
emissions and investment cost as objectives led to
optimal solutions where the building insulation and tightness
caused increased room air temperatures and overheating
during the summer season. Therefore, to limit this phenom-
enon, there is a need to include a constraint for the maximum
value of the allowable overheating of air temperature inside the
zones in the problem set-up. This will be considered in the
Figure 3. Average U-value of the house (U
bldg
) for the optimal solutions.
M. Hamdy et al.
4of 5International Journal of Low-Carbon Technologies 2010, 00, 1 – 5
at Helsinki University of Technology on April 26, 2010 http://ijlct.oxfordjournals.orgDownloaded from
continuation work of this study, which will be done by includ-
ing a thermal comfort criterion as a problem constraint in the
multi-objective optimisation process.
REFERENCES
[1] http://www.equa.se/eng.ice.html.
[2] Heljo J, Laine H. Electrical heating and Heat pumps as consumers
of electricity and causes of emissions in Finland. Report 2005:2
(in Finnish). Tampere University of Technology, Institute of Construction
Economics.
[3] Hasan A, Vuolle M, Siren K. Minimisation of life cycle cost of a detached
house using combined simulation and optimisation. Build Environ 43:
2022–34.
[4] Hamdy M, Hasan A, Siren K. Combination of optimisation algorithms for
a multi-objective building design problem. In: Proceedings of Building
Simulation ’09 Conference, IBPSA, 2009, pp. 173–9.
[5] Deb K. Multi-objective Optimization Using Evolutionary Algorithms. John
Wiley and Sons, 2001.
[6] http://www.tamar.fi.
Optimum design of a house and its HVAC systems
International Journal of Low-Carbon Technologies 2010, 00, 1 – 5 5of 5
at Helsinki University of Technology on April 26, 2010 http://ijlct.oxfordjournals.orgDownloaded from