applying a multi-objective optimization approach for design of low-emission cost-effective dwellings

lable at ScienceDirect

Building and Environment 46 (2011) 109e123

Contents lists avai

Building and Environment

journal homepage: www.elsevier .com/locate/bui ldenv

Applying a multi-objective optimization approach for Design of low-emissioncost-effective dwellings

Mohamed Hamdy*, Ala Hasan, Kai SirenAalto University School of Science and Technology, Department of Energy Technology, P.O. Box 14400, FI-00076 Aalto, Finland

a r t i c l e i n f o

Article history:Received 7 May 2010Received in revised form5 July 2010Accepted 8 July 2010

Keywords:CO2-eq emissionsInvestment costSummer overheatingBuildingOptimizationPareto front

* Corresponding author. Tel.: þ358 9470 23161; faxE-mail address: [email protected] (M. Hamdy).

0360-1323/$ e see front matter � 2010 Elsevier Ltd.doi:10.1016/j.buildenv.2010.07.006

a b s t r a c t

Modern buildings and their HVAC systems are required to be not only energy-efficient but also producefewer economical and environmental impacts while adhering to an ever-increasing demand for betterenvironment. Research shows that building regulations which depend mainly on building enveloperequirements do not guarantee the best environmental and economical solutions. In the current study,a modified multi-objective optimization approach based on Genetic Algorithm is proposed andcombined with IDA ICE (building performance simulation program). The combination is used to mini-mize the carbon dioxide equivalent (CO2-eq) emissions and the investment cost for a two-storey houseand its HVAC system. Heating/cooling energy source, heat recovery type, and six building envelopeparameters are considered as design variables. The modified optimization approach performed effi-ciently with the three studied cases, which address different summer overheating levels, and a set ofoptimal combinations (Pareto front) was achieved for each case. It is concluded that: (1) compared withinitial design, 32% less CO2-eq emissions and 26% lower investment cost solution could be achieved, (2)the type of heating energy source has a marked influence on the optimal solutions, (3) the influence ofthe external wall, roof, and floor insulation thickness as well as the window U-value on the energyconsumption and thermal comfort level can be reduced into an overall building U-value, (4) to avoidmuch of summer overheating, dwellings which have insufficient natural ventilation measures couldrequire less insulation than the standard (inconsistent with energy saving requirements) and/or addi-tional cost for shading option.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

The European Commission announced plans for a EuropeanUnion energy policy that included a unilateral 20% reduction(compared with emission levels of 1990) in greenhouse gas emis-sions by 2020. Buildings contribute to about 35% of carbon dioxideemissions, which is closely related to climate change [1]. The maincontributor to the total heating-related CO2 emissions of 725 Mt/a from the EU building stock in 2002 was the residential sector(77%) while the remaining 23% originated from non-residentialbuildings. In the residential sector, single-family houses representthe largest group and are responsible for 60% of the total CO2emissions, equivalent to 435 million ton per year [2].

Aiming to environmental building solutions, the updatedEuropean regulations stipulate well-insulation (U-values) and heatrecovery requirements. This has resulted in significant energysavings for heating, especially in northern Europe: for example,

: þ358 9470 23418.

All rights reserved.

in Germany it has lead up to 30% energy savings compared to theprevious standards, in France to 10% savings and in Ireland to22e33% savings. Thermal insulation of buildings (external walls,roof and floor) and double-pane windows (even triple glazing withlow-e and argon in northern countries like the Baltic States, Finlandand Sweden) reduce annual energy consumption for space heating,by lowering heat losses through the building’s envelope [3]. Energyconsumption in insulated buildings may be 20e40% less than innon-insulated buildings [4].

Although most of the current European building regulationshave well measures for energy saving, they cannot guarantee thebest environmental solutions. Based on models representative forthe range of the Norwegian district heating plants, calculationsshowed that heating-related CO2 emissions in residential buildingsconnected to the district heating grid and with an energy standardin accordance with the new building regulations are lower than forsimilar buildings with a low-energy standard and with heatingbased on electricity [5]. The primary energy use and the CO2emissions depend strongly on the source of energy supply. A studyby Gustavsson and Joelsson [6] shows that a single-family house

mailto:[email protected]

www.sciencedirect.com/science/journal/03601323

http://www.elsevier.com/locate/buildenv

http://dx.doi.org/10.1016/j.buildenv.2010.07.006



House plan

(143 m2

floor area)

Lower floor

Upper floor

Stair

Bed R.

Living Room

Bed R.

Bed R. Bed R.

Bath R.

WC

Kitchen

Stair

Stair

Three-zones

simplified model

Glassing area 5 m 2

Small window for summer cooling (0.9 m2)

This side is adjacent to another building

6.8 m

Stair

N

a- Original floor plan b- Three zones simplified model

Fig. 1. Two-storey house (143 m2).

M. Hamdy et al. / Building and Environment 46 (2011) 109e123110

from the 1970s heated with biomass-based district heating andwith cogeneration has 70% lower operational primary energy usethan if heated with fuel-based electricity.

Without considering the economics, the research would lead tocostly environmental solutions. For low-emission, cost-effectivesolutions, Rolfsman’s study [7] addressed two energy savingmeasures (e.g., extra insulation and new types of windows) andthree heating energy sources (electrical, district heating andground source heat pump). With the assumed costs, the measuresdid not appear economical. The low-emission cost-effective solu-tion is difficult and time-consuming to achieve using iterative trialand error technique. As an alternative, a suitable simulation-basedoptimization approach can be utilized. The increase in the invest-ment cost corresponding to improving the building performance isconsidered in Hasan et al’s study [8]. The study minimizes the lifecycle cost (LCC) as a single objective, providing one optimal solutionfor each studied case. A study by Wang et al. [9] implementeda multi-objective Genetic Algorithm optimization model for single-storey office building, providing a set of design alternatives for botheconomical and environmental criteria. The study focused only onthe building envelope. However, it was mentioned that moreparameters can be optimized if the scope was expanded to covermechanical systems. Verbeeck and Hens’ study [10] performed lifecycle optimization for externally low-energy dwelling concepts.The study implemented a two-step, multi-objective evolutionaryGA algorithm. For a good balance between computational time andapproximation of the ideal Pareto front or trade-off curve, 6000evaluations were performed. Furthermore, an in-depth analysisand fine-tuning was necessary. Legally non-binding buildingenergy codes, voluntary building environmental performanceassessment schemes, and incentive-based schemes that providesubsidies to reduce the costs of improvements are reviewed inRef. [11].

Expanding the design space solution to include the availableenergy sources and ventilation heat recovery types besides thebuilding envelope parameters has become necessary for environ-mental building designs. However, the costs should be consideredfor economical solutions. The large number of design variables aswell as the non-linear relations between variables poses a complexchallenge for low-emission, cost-efficient building solutions.During the last two decades, evolutionary computation techniques,such as genetic algorithms (GA), have been receiving increasingattention regarding their potential as optimization techniques forsuch complex problems. However since GA starts searching byrandomly sampling within an optimization solution space and thenuses stochastic operators to direct a process based on objectivefunction values, a large number of generations are usually requiredto achieve an acceptable Pareto front. Furthermore, a high quality ofoptimal solutions (a continuous Pareto front) cannot be guaranteedby using a certain number of generations as a stopping criterion[12].

The current study implements a three-phase multi-objectiveoptimization approach (PR_GA_RF). The approach aims to reducethe random behavior of the genetic algorithm (GA) by using a goodinitial population from the preparation phase (PR). After a lownumber of generations, the refine phase (RF) starts using fast andrealistic stopping criteria, considering good diversity for optimalsolutions. The approach is combined with IDA ICE 3.0 (BuildingPerformance Simulation Program) to minimize the CO2-eq emis-sions related to building energy consumption and the investmentcost of the design variables. A two-storey house in the cold climateof Finland is selected as a case study.

In order to achieve environmental cost-effective building solu-tions, a wide solution space is suggested, including four heatingenergy sources (electrical heating, oil fire boiler, district heating,

and ground source heat pump), free cooling option, and three typesof ventilation heat recovery systems as well as various options forsix building envelope parameters: building tightness, insulationthickness of the external wall, roof, and floor as well as the windowand shading types.

Since the building energy consumption depends significantly onthe desired thermal comfort level, three studied cases areproposed. The first case disregards summer overheating, aiming toextreme optimal trade-off relation in terms of CO2-eq emission andinvestment cost. The second case assumes the summer overheatinglevel of the non-cooling initial design (building envelope withU-values based on C3-2007 [13]) as a constraint function. The thirdcase addresses a higher level of thermal comfort by assuming thesummer overheating level of the initial designwith a cooling optionas a constraint.

2. House and its HVAC system

2.1. General description

A typical Finnish two-storey house, located in Helsinki, is takenas a case study. The gross floor area of the house is 143 m2. Theglazing area represents about 15% of the floor area. Each storey hasan internal height of 2.5 m. The two stories are connected bya staircase (Fig. 1a). The internal gains due to people, lighting andelectric appliances are assumed according to annual values speci-fied by the Finnish building code D5 [14] and taken into thecalculation as a profile with hourly values.

The ventilation system consists of an air handling unit (AHU)which supplies fresh air to the bedrooms and living room anddraws the exhaust air from the bathrooms and the kitchen witha cross air-to-air heat recovery system. The AHU heater keeps thesupply air temperature at 18 �C when the incoming outdoor airtemperature is lower than this temperature. The average exhaust

Table 1Design variable.

Variables Type Five initialdesigns

Lowerbound

Upperbound

Insulation thicknessof external wall [m]

Continuous 0.124 0.024 0.424

Insulation thicknessof roof [m]

Continuous 0.21 0.11 0.51

Insulation thicknessof floor [m]

Continuous 0.14 0.04 0.44

Windows type Discrete 1 1 5Heat recovery Discrete 2 1 3Shading type Discrete 1 1 2Building tightness type Discrete 1 1 5Heating/cooling

system typeDiscrete From 1 to 5 1 5

Table 2Window type.

WindowType

U-value[W/m2 K]

Total solartransmission [%]

Shading coefficientfactor

Price[V/m2]

1 1.4 44 0.656 1802 1.1 44 0.656 1853 1 34 0.53 2054 0.85 29 0.482 2405 1.1 28 0.437 210

M. Hamdy et al. / Building and Environment 46 (2011) 109e123 111

air flow from the whole house is equal to 0.65 air changes per hour,which is in accordance with the Finnish Building code D2 [15].

The building tightness, insulation thickness of the external wall,roof, and floor as well as window and shading type are assumed asdesign variables. An oversized heating system is used in simulationto achieve the minimum acceptable indoor air temperature(21�1 �C) for different building envelope solutions. No heating/cooling system is used for the staircase.

2.2. Initial designs

Five initial designs are assumed with different heating/coolingsystems: electrical heating (Sys. 1), oil fire boiler (Sys. 2), districtheating (Sys. 3), ground source heat pump GSHP (Sys. 4), and GSHPwith a free cooling option (Sys. 5). The cooling option is assumedonly with the 5th initial design. The heat recovery has an annualefficiency of 70%. Based on the National Building Code of Finland,C3-2007 [13], U-values 0.24, 0.15, 0.24, and 1.4 W/m2 K are assumedfor external walls, roof, ground floor, and window respectively. Thewindow type has a shading coefficient SC of 0.656. The initialsolutions are assumedwith a shading option and building tightnessn50¼ 4 1/h where n50 is the number of air changes per hourequivalent to the air leakage rate with a 50 Pa pressure differencebetween indoor and outdoor.

2.3. Simplified model including operable window

To reduce the execution time of simulation, a simplified model(Fig. 1b) is used to represent the house by three zones: lower floor,upper floor, and staircase. The zones are considered as open planareas. On the north and south directions, two windows are used ateach storey to represent the total fixed glazing area of the building(20 m2). In addition, a small window (0.9 m2) is proposed andinstalled at the middle of each storey with a special PI-controller.The two small windows open proportionally when the outdoor airtemperature is less than the indoor air temperature and the latter ishigher than 24 �C. The controller tries to emulate human behaviorby opening the window to improve the thermal comfort byincreasing the ventilation in summer.

3. Formulation of the optimization problem

3.1. Design variables

Aiming to low-energy cost-effective concepts for single-familyhouses in Finland, a sensitivity analysis was performed [16]addressing insulation thickness, window type, building tightness,and efficiency of heat recovery as design parameters. In a recentFinnish optimization study [8], nearly the same parameters withdifferent assumptions were addressed, and the life cycle cost wasconsidered as a single-objective function. Different types of energysources are used in Finland. The primary greenhouse gas emissionfactors (related to three major greenhouse gases: CO2, sulphur, andnitrogen.) for different types of energy supplied to the buildingswere evaluated in Ref. [17].

The current study considers eight design variables: heating/cooling energy source, ventilation heat recovery type, buildingtightness, window type, shading option as well as insulationthickness of the external wall, roof, and floor. The environmentalimpact is estimated considering emission factors from Ref. [17]. Theeight design variables and their investment costs are presented inTables 1e6. Table 1 presents the initial values, lower bounds, upperbounds, and types (discrete or continuous) of the eight designvariables as well as the ranges of the external wall, roof, and floorinsulation thickness. Five window types are described in Table 2.

Three different ventilation units are presented in Table 3. Table 4presents the shading options: external blind, horizontal lathsshading and no shading option. Improving the building tightness isdone by careful work and more strict control on the site.This creates an additional cost. It is assumed that the tightnessn50¼ 4 1/h is the reference valuewith no additional costs. A smallertightness value creating an additional cost (V/floor m2) is shown inTable 5. Five types of heating/cooling systems with differentemission factors are shown in Table 6: four heating systems (directelectrical heating, fuel oil boiler, district heating, or GSHP) and onlyone heating/cooling option (GSHP with the free cooling option).The ground source heat pump (GSHP) is a brine-to-water systemwith a compressor driven by electricity. The brine flows in a deepborehole using the ground as the heat source. Cooling is achievedwithout compressor, while the compressor is kept in the freecooling mode, by cold water which flows between the chilled floorsand an intermediate heat exchanger. The free cooling upgrade costsan additional 1000 V. However, it provides powerful cooling, usingnegligible amount of energy to run a circulating pump.

3.2. Objective functions

The aim of this study is to achieve low-emission, cost-effectivedesign solutions. Therefore CO2-eq emissions related to heatingenergy and the investment cost related to the suggested designvariables are selected as two objective functions to be minimized.The first objective, CO2-eq emissions [kg/m2 a], is calculated by thefollowing equation.

CO2-eq ¼ Q � EF=h (1)

where Q is the total heating energy, EF is the primary greenhousegas emission factor, and h is the heating system efficiency (Sys.1e3)or the COP of the heat pump (Sys. 4 and 5), see Table 6.

The second objective, the investment cost, is the total cost of theinvestments related to the eight design variables. Different typesof insulation (mineral wool, blow-in wool, and polyurethane) areused in the external wall, roof, and ground floor. The insulationshave prices of 56.3, 32.5, and 100 V/m3, and thermal conductivity of

Table 3Price of the AHU including different types of heat recovery system.

Type Efficiency [%] Specification Price [V]

1 60 Plate heat exchanger 31722 70 Rotating wheel 34433 80 Rotating wheel 3715

Table 5Building tightness type.

Type Specificationn50 [l/h]

Price for additionalwork [V/m2]

1 4 02 3 53 2 124 1 225 0.5 30


0.035, 0.05, and 0.026 W/m2 K, respectively [8]. Only one shadingoption (External blind, horizontal laths) is assumed with the pricefrom [18]. The investment costs of other design variables areassumed to be in line with market prices and presented in Tables 2,3, 5, and 6.

Table 6Heating/cooling system types.

3.3. Constraint function

The HVAC energy consumption (consequently the CO2-eq emis-sions) depends on the criterion of thermal comfort, i.e., disregardingthe summer overheating allows additional insulation for maximumspace heat energy saving. To assess the level of summeroverheating,degree-hour DH24 is used and defined as the summation of theoperative temperature degrees higher than 24 �C at the warmestzone during a one-year simulation period (8760 h), as follows:

DH24 ¼Xi¼8760

i¼1

dH24 (2)

dH24 ¼ ðTi � 24ÞDt when Ti � 24 > 0

dH24 ¼ 0 when Ti � 24 � 0

where Ti is the operative temperature [�C] at the centre of thewarmest zone (upper floor) and dt is a one hour time period [h].

In order to investigate the influence of thermal overheating onthe optimal solutions, three case studies (Cases 1, 2 and 3) areproposed. Case 1 disregards the summer overheating, aiming toextreme optimal trade-off relation between the objectives. In Case2, the degree-hours DH24 value of the initial design withouta cooling option (DH24¼ 2400 �C h) is considered as a constraintfunction. For a higher thermal comfort level, DH24¼1000 �C h (theindoor degree-hours of the initial design with a cooling option) isassumed as a constraint function for Case 3. The three studied casescan be summarized as follows:

Objective 1: Minimum CO2-eq emissions related to heatingenergy.

Objective 2: Minimum investment costs for the eight designvariables.

Constraint:

Case 1 (no constraint);Case 2 (DH24� 2400 �C h);Case 3 (DH24�1000 �C h).

Find: optimal combinations between the design variables (Xi)where i¼ 1e8.

Table 4Shading type.

Type MSC MSSC Description Price [V/m2]

1 0.14 0.09 External blind,horizontal laths

200

2 1 1 No shading 0

MSC: multiplier for shading coefficient; MSSC: multiplier for solar shadingcoefficient.

Method: a modified multi-objective optimization approach(PR_GA_RF). (More details about this approach can be found in thefollowing section).

4. The modified optimization approach

In a previous study [12], two combinations between the opti-mization algorithms were developed. The First is called PR_GA. It isa two-phase multi-objective optimization combination that worksunder the MATLAB environment. Briefly, in PR_GA the GeneticAlgorithm (GA) from the MATLAB 2008a Genetic and Direct SearchToolbox [19] was modified to be able to deal with discrete andcontinuous variables. Then it was combined with a deterministicoptimization solver (FMINCON, single-objective deterministicsolver from MATLAB optimization ToolBox) in order to provide GAwith a good collection of individuals as an initial population. Thisprocess is called the preparation phase (PR). Themajor advantage ofPR_GA is that it tries to reduce the random behavior of GA in anattempt to obtain good solutions with a lower number of evalua-tions (simulation runs).

The second combination is called GA_RF. It can be used for highquality, enhanced optimal solutions. As a hybrid optimizationsolver, GA_RF utilizes Fgoalattain (Multi-objective optimizationsolver from MATLAB 2008a optimization toolbox) after using themodified GA (MATLAB GA able to deal with discrete and continuousdesign variables) in order to improve the quality and diversity ofobtained results. The advantage here is that GA_RF refines onlya certain number of the obtained solutions exhibiting significantdiversity. This reduces the time of the optimization process byavoiding much of mostly similar evaluations. The other advantageis that the Pareto solutions are enhanced by sorting the history-results of the refining steps.

The current study proposes a combination between the twopredefined approaches. The new combination is called PR_GA_RF.This combination provides the advantages of the two modifiedapproaches (PR_GA and GA_RF). However, since the PR_GA_RF isa multi-step approach, many of the evaluations are repeated. Thiscould be avoided easily if the new combination (PR_GA_RF) wasmodified to determine the values of the objective and constraintfunctions of the repeated design, i.e., variable combinations directlyfrom the history data.

Type Type Price [V/m2] EF/h [kg CO2-eq/kWh]

1 Direct electric radiatora 30 0.459/12 Oil fire boilera 94 0.267/0.93 District heatinga 101 0.226/14 GSHPa 126 0.459/3b

5 GSHP with free cooling 133 0.459/3b

EF: emission factor.a Without cooling system.b h¼ COP in case of GSHP system.

IDA ICE

(Building performance Simulation program)

Input file

(IDALISP)

Outputs files

(energy consumption, indoor temperature)

CO2-eq,

Investment cost,

and DH24

calculator

(MATLAB m-files)

Optimizer

(Modified MATLAB optimization approach)

Data source

(emission factors and prices)

ReadWrite

WriteRead

MATLAB Environment

Fig. 2. Components and their relationships for simulation-based optimization.


In the current study, the preparation phase (PR) used FMINCONto minimize the CO2-equivent emissions then the investment costas single objectives in two separate optimization steps imple-menting a constraint function on the degree-hour DH24 for Cases 2and 3. The history of the PR’s iterations is sorted based on the twoobjective functions providing the GAwith a good initial population.The refine phase (RF) addressed only 10 optimal solutions from theGA Pareto front. The refined optimal solutions were selectedconsidering different investment costs dividing the Pareto frontwith fixed step (maximum investment cost�minimum invest-ment cost)/10. In order to obtain the final Pareto front, the history ofthe refine phase (RF) is sorted applying the non-dominatedconcept.

5. The simulation-optimization approach

PR_GA_RF is combined with IDA ICE (IDA Indoor Climate andEnergy 3.0 program [20]) under the MATLAB environment asshown in Fig. 2. IDA ICE is a whole-building dynamic simulationprogram that makes simultaneous performance assessments of allissues fundamental to building design: shape, envelope, glazing,HVAC systems, controls, light, indoor air quality, comfort, energy

10

15

20

25

30

35

40

10 20 30 40 50CO

2-eq em

GSHP

(Sys.4)

District

(Sys.3)

Oil fi

th

gi

ee

ht

fo

ts

oc

tn

em

ts

ev

nI

01

[s

el

ba

ir

av

ng

is

ed

3]

€

Utopia point

Fig. 3. Non-cooling initial designs and the Pareto

consumption, etc. The accuracy of IDA ICE was assessed through theIEA solar heating and cooling program, Task 22, subtask C [21].IDA ICE 3.0 was chosen as one of the major 20 building energysimulationprograms, which were subjected to analysis andcomparison [22].

The combination between PR_GA_RF and IDA ICE was used toperform the optimization process for the three cases studied. Thework was done by a computer (Intel� core �2 Quad CUP 2.40 GHzprocessor. 3061 MB RAM) with the Windows Vista system.A simulation run took on average 2.5 min.

6. Results and discussion

6.1. Case 1

In this case the optimization approach is employed to obtain anextremely optimal relation between the two objective functions.No constraint function is assumed on summer overheating, whichprovides possibilities for more energy saving, e.g., using more cost-effective insulation to reduce the space heating energy. For thispurpose, 1310 simulation runs are performed: 290 for the prepa-ration phase (PR), 720 for the GA phase using 40 population

60 70 80 90 100

Cadidate solutionsPareto frontInitial designsUtopia point

ission [kg/m2

a]

Electrical radiator

(Sys.1)

re boiler (Sys.2)

front solution of Case 1 (DH24� 6250 �C h).


individuals (Pop) and 18 generations (Gen), and 300 for the refinephase (RG). Fig. 3 shows all the candidate solutions from the opti-mization history, four non-cooling initial designs (Table 1: heatingsystem from 1 to 4), 41 optimal solutions (Pareto front), and theutopia point (minimum investment cost and minimum CO2-eqemissions). The values of the objective functions corresponding tothe utopia point are calculated by assuming the lowest price foreach design variable (lower bounds of insulation thickness,window type 1, non-shading option, building tightness n50¼ 4 l/h,the lowest efficient heat recovery type 1, and electrical radiation asa heating system) to estimate the minimum investment cost, andby assuming the best heating energy saving measures (upperbounds of insulation thickness, window type 4, non-shadingoption, building tightness n50¼ 0.5 1/h, and the highest efficientheat recovery) as well as GSHP as a heating system to determine theminimum CO2-eq emissions. The utopia point represents animaginary target which cannot be achieved because of the trade-offbetween the objectives. The utopia point is used here to check theextremes of the obtained Pareto front.

The cooling option is only available in Sys. 5 (GSHP witha cooling option). This requires additional investment cost (1000 V)compared with Sys. 4 (GSHP without a cooling option). Since thiscase is performed without a constraint on summer overheatinglevel (DH24), there is no reason for additional investment ina cooling option. Therefore Sys. 5 is not selected in any of the 41optimal combinations. The same reason applies for not gettingsolutions with window shading.

Disregarding the summer overheating provides a set of solutions(Pareto front)which are better than the initial ones in terms of lowerCO2-eq emissions and investment cost. However, the obtainedsolutions have higher DH24 levels (DH24 from 4181 to 6254 �C h)than the non-cooling initial ones (DH24¼ 2400 �C h). Also, it isworthwhile to mention that the obtained solutions on the Paretofront are found to be classified according to the heating energysource. This reveals that the heating systemhas a stronger influenceon the low-emission cost-effective solutions than the other designvariables. This phenomenon can be explained by the following:

1. The amount of CO2-eq emissions is the product of the emissionfactor (EF) and the total heating energy (Q), see Equation (1).The emission factor (EF) depends only on the type of theheating source, see Table 6.

2. The total heating energy (Q) consists of space heating (qs),domestic hot water heating (qw), and system heat loss (ql). Thedomestic water heating and system heat loss are assumedconstants. Their related CO2-eq emissions (EF� qw and EF� ql)are functions of the emission factor (EF) of the heating system.Based on the 41 obtained solutions (Pareto front), the domestichot water heating together with the system heat loss is(36.5 kWh/m2 a), which is 25e52% of the total heating energy.

3. The investment cost of the heating system is often higher thanthe cost of the other design variables. According to the41 obtained solutions, the price of an electrical radiator heatingsystem (the cheapest heating system) and the price of a GroundSource Heat Pump (the most expensive heating system)represent 11% and 50% of the maximum investment cost(38 500 V), respectively.

It is worthwhile also to mention that without changing theheating system, there is only one way to reduce the CO2-eq emis-sion, namely by reducing the energy consumption (Q). Since thedomestic hot water heating (qw) and system heat loss (ql) areassumed constants, optimal combinations between the insulationthicknesses, window type, building tightness, and heat recoverysystem are found by the optimizer for each heating system (Sys.1, 2,

3, and 4) giving low-energy cost-effective solutions. In the optimalsolutions of Fig. 3, the space heating energy (AHU heating coil plusspace heating device energy) comes in the descending order fromright to left, and as shown in Fig. 4. It should be noted that this isachieved by finding cost-effective combinations between buildingenvelope parameters to get a lower building U-value (Ubldg). TheUbldg represents the mean U-value for the whole-building envelopedefined by the following equation:

Ubldg ¼�Uwall � Awall þ Uroof � Aroof þ Ufloor

�Afloor þ Uwindow � Awindow

�.Atotal ð3Þ

where

Atotal ¼ Awall þ Aroof þ Afloor þ Awindow

From Fig. 4, it can be seen that, for each system, lower Ubldgprovides lower space heating energy. The variations in the selectedtypes of heat recovery and building tightness (Fig. 5) explain thesmall deviation in the relation between the results of Ubldg andspace heating in Fig. 4. The average Ubldg of the optimal solutions is0.238 W/m2 K, and about 80% of the solutions have Ubldg less thanthat for the initial design (Ubldg¼ 0.3 W/m2 K). The minimumbound for Ubldg in the solution space is 0.131 W/m2 K (solution 1).However for Sys.1, reducingUbldg to less than 0.17 W/m2 K (solution23) was not an optimal solution because a lower emission heatingsystem (oil fire boiler) was available. Fig. 4 indicates that theoptimization solver changed the heating system from Sys. 1 (solu-tion 23, Ubldg¼ 0.17 W/m2 K) to Sys. 2 (solution 22, Ubldg¼ 0.34 W/m2 K) instead of continuing in reducing Ubldg with Sys. 1 fromsolutions 41 to 23. In addition, at solution (23), building tightness(n50¼11/h) and heat recovery efficiency (h¼ 70%) were selected,while the solution space includes higher building tightness(n50¼ 0.5 1/h) as well as better heat recovery (h¼ 80%), whichprovides higher savings in the space heating energy. This illustratesthat changing the heating system type from Sys. 1 to Sys. 2 isa better solution for the two objective functions than continuing inreducing the space heating energy with Sys. 1. It is also to be notedthat the district heating solutions (Sys. 3) dominated the oil fireboiler solutions (Sys. 2) except for one solution. The reason is thatwhile the emission factor of Sys. 3 is less than that for Sys. 2 there isno considerable difference in the price between the two systems.

Compared with the initial design (Sys. 1), 32% lower CO2-eqemissions and 26% lower investment are achieved on the Paretofront. The minimum required air temperature (21�1 �C) inside theupper and lower storey for all the optimal building envelopesolutions is achieved. A procedure is made available to open a smallwindow at each storey via a PI-controller when the indoor airtemperature is >24 �C and the outdoor air temperature is lowerthan the indoor air temperature. However, the indoor operativetemperature is raised above 25 �C for 13% of the year hours witha maximum temperature of 33.2 �C. This minimum overheatingoccurred at solution number 41, which has the maximum value ofUbldg (0.496 W/m2 K) in the optimal solution and the maximumCO2-eq emission (70.5 kg CO2-eq emission/m2). The next twostudied cases minimize the CO2-eq emissions and investment costwhile trying to avoid much of summer overheating. Section 8compares between the three cases presenting the duration curvesof the operative temperature and mean air temperature for them.

6.2. Case 2 (DH24� 2400 �C h)

DH24¼ 2400 �C h is the summer overheating level of the fournon-cooling initial designs (initial design values in Table 1, heating

0

25

50

75

100

125

0

0.1

0.2

0.3

0.4

0.5

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41Optimal Solutions

m/

hW

k[

yg

re

ne

gni

ta

eh

ec

ap

S2

]a.

GSHP

(Sys.4)

District

(Sys.3)

Electrical radiator

(Sys.1)

Oil fire boiler (Sys.2)

No shading

Initial designs 91.5 kWh/m2.a

Initial designs

0.3 W/m2

K

eu

la

V-

Ug

dlb

m/

W[

2

.]

K

Fig. 4. Space heating energy of the optimal design as a percentage of the space heating energy of the initial design.


systems from 1 to 4). In order to achieve low-emission cost-effec-tive solutions at that level of summer overheating, DH24� 2400 �Cis implemented as a constraint function. In this case, the modifiedmulti-objective optimization approach PR-GA-RF (Section 4) per-formed 2787 evaluations: 297 for the preparation phase (PR), 1200for GA phase, and 1290 for refine phase (RF) achieving 64 optimalsolutions. Fig. 6 shows those solutions in addition to the four non-cooling initial designs.

To check the extremes of the obtained Pareto front, the utopiapoint is determined as shown in Fig. 6. Because of the summeroverheating limit of Case 2 the values of the objective functionscorresponding to the utopia point are found by applying a modifiedsingle-objective optimization solver (Fmincon solver, from theMATLAB 2008a optimization toolbox, modified to deal withdiscrete and continuous variables) to determine the minimumamount of CO2-eq emissions then the minimum investment cost attwo separate steps using DH24� 2400 �C h as a non-linearconstraint function. According to the utopia point in Fig. 6, the PR-GA-RF optimization approach achieved the extreme minimums ofthe two objectives separately at the two terminals of the Paretofront. It is also worthwhile to mention that the optimal solutions ofthis case are found to be classified according to the type of a heatingsystem similar to Case 1. However system 5 (GSHP with coolingoption) is selected in 32 optimal solutions.

System 5 is the only system in the solution space that offersa cooling option. Furthermore, it has the highest investment cost.The cooling system is sized (i.e., the potential of free cooling system

is limited) for the initial design (Table 1) which has a poor buildingtightness (n50¼ 4 1/h) and shading option (Table 4). This meansthat additional insulation, higher building tightness, and/or non-shading solutions may cause summer overheating even if thecooling option is selected. The challenge was to find optimalcombinations of building envelope parameters which achieve thepredefined thermal comfort criterion (DH24� 2400 �C h) witha smaller increase in energy (consequently a smaller increase in theCO2-eq emissions) and a smaller increase in the investment cost(e.g., no mechanical cooling solutions).

Fig. 7 shows that the optimization solver kept theDH24� 2400 �C h for all the 64 optimal solutions, and half of thosesolutions are without a cooling system. This is achieved by usingthe shading option for all the non-cooling solutions (33e64). Theshading option was also used for the free-cooling ones (1e9) whenimplementing additional insulation (Ubldg� 0.17 W/m2 K) for morespace heating energy saving. For each heating/cooling system,optimal combinations between the window type and insulationthicknesses of the external wall, floor, and ceiling were selected,making a clear relation between the overall heat transmissioncoefficient Ubldg and energy consumption, as shown in Fig. 8. Theinfluence of the other design parameters is causing some irregu-larity in the relation.

The building envelope’s discrete variables played a big role inattaining the defined level of the thermal comfort(DH24� 2400 �C h), with and without a cooling option. Forexample, to avoid much of summer overheating, the building

0

1

2

3

4

5

0

1

2

3

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41

Optimal Solutions

ep

yT

ss

en

th

gi

Tg

ni

dl

iu

B

GSHP

(Sys.4)

District

(Sys.3)

Electrical radiator

(Sys.1)


No shading

ep

yT

yr

ev

oc

eR

ta

eH

Fig. 5. Average U-value of the building (Ubldg) for the optimal solutions of Case 1 (DH24� 6250 �C h).

15

20

25

30

35

40

45

10 20 30 40 50 60 70 80 90 100

Initial dsigns (DH24 = 2400 C.h)Pareto front (DH24 ≤ 2400 C.h) Utopia point (DH24 = 2400 C.h at minimum CO2-eq emissions)

GSHP without cooling (Sys. 4)

District heating(Sys. 3)


Electrical Radiator(Sys. 1)

Solutions 46:64

Solutions 35:45

Solutions 33:34

Solutions 10:32

Solutions 1:9

Utopia point


th

gi

ee

ht

fo

ts

oc

tn

em

ts

ev

nI

01

[s

el

bai

ra

vn

gi

se

d3

]€

CO2-eq emission [kg/m

2.a]

District heating (Sys.3)

GSHP (Sys.5)

5

Fig. 6. Non-cooling initial designs and the Pareto front of Case 2 (DH24� 2400 �C h).


0

800

1600

2400

1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64Optimal Solutions

HD

42

]h.

C[

Electrical Radiator

(Sys.1)

Oil fire boiler

(Sys.2)

GSHP with free cooling

(Sys.5)

District heating

(Sys.3)With shadingWithout shadingWith shading

Fig. 7. DH24 for the 64 optimal solutions of Case 2 (DH24� 2400 �C h).


tightness type 1 (n50¼ 4 1/h) and the shading option are selectedfor 80% and 66% of the optimal solutions, respectively and as shownby Fig. 9. In addition, window types 1 and 2, which have the highestshortwave shading coefficient (SSC¼ 0.656: less solar heatingenergy), are selected for 88% of the optimal solutions. Low expensesof these combinations are the other reasons behind the selection.

Next we analyze the performance of the optimization solver inselecting various design parameters. To obtain minimum solutionsof investment cost and CO2-eq emissions, the Electrical Radiator(Sys. 1), the lower price heating system, is found to be optimal

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

1 4 7 10 13 16 19 22 25 28 3

0

25

50

75

100

125

150

175GSHP with free cooling

(Sys.5)

Optim

Without shadingWith shading

m/

hW

k[

yg

re

ne

gni

ta

eh

ec

ap

S2

]a.

Initial designs 91.5 kWh/m2.a

Initial designs

0.3 W/m2

K

eul

aV

-U

gdl

bm

/W

[2

.]

K

Fig. 8. Ubldg and the space heating energy of the o

(optimal solution from 64 to 46) with incrementally additionalinsulation to minimize the space heating energy (Fig. 8), andconsequently CO2-eq emissions (Fig. 6). However, since any addi-tional insulation would increase the summer overheating, theoptimization solver switched to the Oil Fire Boiling heating system(Sys. 2) at solution 45, avoiding any increase in the building enve-lope insulation (Fig. 8). This provides a higher reduction in CO2-eqemissions (Fig. 6), because Sys. 2 has a lower emission factor thanSys. 1 (Table 6) and keeps DH24 below 2400 �C h as shown in Fig. 7.This requires a significant increase in the investment cost (the

1 34 37 40 43 46 49 52 55 58 61 64

Electrical Radiator

(Sys.1)

Oil fire boiler

(Sys.2)

al Solutions

District heating (Sys.3)

With shading

ptimal solutions of case 2 (DH24� 2400 �C h).

0

1

2

Optimal Solutions

District heating

(Sys.3)

0

1

2

3

4

5

0

1

2

3

1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64

Electrical Radiator

(Sys.1)

Oil fire boiler

(Sys.2)

GSHP with free cooling

(Sys.5)

With shadingWithout shadingWith shading

ep

yT

yr

ev

oc

eR

ta

eH

ep

yT

ss

en

th

gi

Tg

ni

dl

iu

Be

py

Tw

od

ni

W

Fig. 9. Building tightness, window, and heat recovery type of the optimal solutions for Case 2 (DH24� 2400 �C h).


difference between the prices of Sys. 2 and Sys. 1), explainingthe discontinuity (investment cost step) in the Pareto front at8620 kg/a of CO2-eq emissions as exhibited in Fig. 6. A higherefficiency heat recovery types 2 and 3 (Table 3) are selected for

30

31

32

33

34

35

36

16 18 20

Better Solutions

th

gi

ee

ht

fo

ts

oc

tn

em

ts

ev

nI

01

[s

el

bai

ra

vn

gi

se

d3

]€

CO2-eq e

Fig. 10. Initial designs and the Pareto f

lower heating energy solutions: from 46 to 50 for the ElectricalRadiator (Sys. 1) and from 1 to 23 for GSHP (Sys. 5) optimal solu-tions. This provides lower CO2-eq emission solutions, while anyadditional insulation is avoided to keep DH24 within an acceptable

22 24 26 28

Pareto front (DH ≤ 1000 C. h)

Initial design (DH = 1000 C. h)

Alternative Solutions

mission [kg/m2.a]

ront of Case 3 (DH24�1000 �C h).

Optimal Solutions

HD

42

]h

.C

[

Sys.5 (GSHP with free cooling )

Building tightness type 1(n50 = 4 1/h)

With shading

Ubldg

< 0.3 W/m2

K Ubldg

≥ 0.3 W/m2

K

0

200

400

600

800

1000

1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58

Better Solutions Alternative

Solutions

Alternative

Solutions

Fig. 11. DH24 for the optimal solutions of Case 3 (DH24�1000 �C h).

0

25

50

75

100

125

150

0

0.1

0.2

0.3

0.4

0.5

0.6

1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58

Sys. 5 (GSHP with free cooling )


With shading

Ubldg

< 0.3 W/m2

K Ubldg

≥ 0.3 W/m2

K


Solutions

Alternative

Solutions

eu

la

V-

Ug

dlb

m/

W[

2

.]

K

Optimal Solutions

Initial designs 91.5

kWh/m2.a

Initial designs

0.3 W/m2

K

m/

hW

k[

yg

re

ne

gn

it

ae

he

ca

pS

2]

a.

Fig. 12. Ubldg and the space heating energy of the optimal solutions for Case 3 (DH24�1000 �C h).


0

1

2

3

4

5

0

1

2

3

1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58

Sys.5 (GSHP with free cooling )


With shading

Ubldg

< 0.3 W/m2

K Ubldg

≥ 0.3 W/m2

K


Solutions

Alternative

Solutions

ep

yT

wo

dn

iW

ep

yT

yr

ev

oc

eR

ta

eH

Optimal Solutions

Fig. 13. Window and heat recovery type of the optimal solutions for Case 3 (DH24�1000 �C h).


limit (Fig. 7). Since the prices of the Oil Fire Boiling (Sys. 2) andDistrict heating system (Sys. 3) are close, the cheapest heatrecovery system (type 1) is selected for all the optimal solutions ofthose two systems, giving them the opportunity to compete witheach other. Since the Oil Fire Boiler (Sys. 2) has a lower price, itdominates most of the District Heating (Sys. 3) solutions. On theother hand, the GSHP Free-Cooling optimal solutions (32 to 10)dominate another part of the District Heating solutions savingmuch of the investment cost by selecting the non-shading option.However, the optimal solutions from 1 to 9, in addition to usingextreme insulation (Ubldg� 0.17W/m2 K) for the lowest CO2-eqemissions, use the shading option to maintain the thermal comfortcriterion.

Sys. 4 has the same emission factor (EF) as Sys. 5 (GSHPwith freecooling option). However to maintain the DH24 constraint, a highexpense shading option is necessarily for Sys. 4 (GSHP withouta cooling option) solutions. Since the shading option is moreexpensive than the free cooling option of Sys. 5, all Sys. 4 solutionsare dominated by Sys. 5 ones. This explains the absence of Sys. 4among the optimal solutions. For the same reason, the Sys. 5optimal solutions (1e32) dominate the initial designs of Sys. 4(Fig. 6).

6.3. Case 3 (DH24�1000 �C h)

In theprevioussection,DH24�2400 �Ch isassumedasaconstraintfunction. This leaded to thin the operative temperature exceeds 24and 25 �C for 18.5 and 8% of the year hours, respectively, with 30.5 �C

maximum temperature. For a higher thermal comfort level,DH24�1000 is assumed as a constraint function in Case 3. Since themajor part of the solution space does not offer any cooling option,a larger number of evaluations, i.e., 3500 (325 PRþ 40 Pop X 40Genþ 1575 RF), was needed to achieve high level thermal-comfortsolutions using strictly the constraint function (DH24�1000 �C).Fig. 10 exhibits 58 optimal solutions for this case representing thetrade-off relation between the two objectives. Those solutions arecompared with the initial design (Table 1, Sys. 5 GSHP with freecoolingoption)whichhasDH24of1000 �C h. It is found thatonly threeoptimal solutions provide better designs than the initial one; theother 55 solutions can be considered as alternatives.

Since some amount of overheating is allowed, the shadingoption 1 (less direct solar heat gain) and the leakiest building type(n50¼ 4 1/h) in addition to Sys. 5 (the only system which hascooling option) are selected for all the optimal solutions. This keepsthe summer overheating under the desired threshold(DH24¼1000), as shown in Fig. 11.

The optimization solver tried to minimize the CO2-eq emissionsusing optimal combinations between the building envelope param-eters and heat recovery type. The Ubldg of those combinations isarranged in a descending order from right to left (Fig. 12). The initialdesign has Ubldg of 0.3 W/m2 K. This value is used to classify theoptimal solution as shown in the following.

Fig. 13 shows that for optimal solutions 27e58, window types 1and 2 and heat recovery type 1 (low price combination) are selectedwith Ubldg� 0.3 W/m2 K (less insulation cost) giving a lowerinvestment cost than the initial design does. A higher efficiency

Fig. 14. Brute force, feasible solutions, and Pareto front for case 3 (DH24�1000 �C h).


heat recovery (type 2) is utilized to obtain a set of solutions (21e26)which produce smaller amounts of CO2-eq emissions (Fig. 10)without an increase in summer overheating (Fig. 11). For morereduction in CO2-eq emissions, more insulation is used(Ubldg< 0.3 W/m2 K). Moreover, window type 1 (SSC¼ 0.656) thenwindow type 5 (SSC¼ 0.437) are selected to reduce the spaceheating demand by increasing the direct and indirect solar heatgain respectively. The influence of the U-values of those windowtypes 1 and 5 (1.4 and 1.1 W/m2 K) is shown in Fig. 12 where theresistance of the insulation thicknesses is nearly constant (optimalsolutions from 1 to 20). The lower U-values of window types 2 and4 (1 and 0.85 W/m2 K respectively) are not used, thus avoidingadditional overheating in order to keep the building atDH24�1000 �C h as defined by the constraint function. The highestefficiency heat recovery type 3 is used, providing more reduction inthe heating energy (consequently CO2-eq emissions) withoutincreasing the summer overheating. This is shown by the optimalsolutions (Ubldg< 0.3 W/m2 K) in Fig. 13.

7. Comparison with results from a brute-force search method

A brute-force searchmethod is implemented to check the resultsobtained by the modified optimization approach (PR_GA_RF) for

Fig. 15. Duration curves of the operative temperature (a) and the mean air tem

Case3. Brute-force search is anexhaustive search that systematicallyenumerates all possible candidate solutions.

The predefined case study includes eight design variables. Threeof them are continuous and five are discrete. Table 1 presents theupper and lower bounds of the design variables. If 5 cmis considered as an exhaustive-search step for the thickness ofinsulation in the wall, roof, and floor, and all combinationsbetween the discrete variables are taken into account, then8� 8� 8� 5� 3� 2� 5� 5¼ 384000 simulation runs are neededto get all possible candidate solutions. The execution time of onesimulation run is about 2.5 min. This means that 666 days would berequired to get the brute-force search results for the predefinedproblem. To make the brute-force search feasible, the problem sizeneeds to be limited by using problem specifying heuristics, whichhelp to reduce the number of candidate solutions to a manageablesize. We can get some indications from the PR_GA_RF results thatwill make the brute-force search fall within a preferable set ofsolutions. For example, in Case 3 the heating/cooling system 5,shading option, and building tightness type 4 are selected in all theoptimized solutions. However, window types 3 and 4 are notselected (Fig. 13). This can reduce the brute-force candidate solu-tions significantly. More reduction is obtained by using a larger stepfor the continuous variable (8 cm instead of 5 cm for theinsulations’ thicknesses). The reduced brute-force search method

perature (b) at the warmest zone (upper floor) for the three studied cases.

Fig. 16. Pareto fronts of the three studied cases (DH24� 6250, 2400, and 1000 �C h) and contribution of the heating/cooling system types in each case.


needs only 5� 5� 5� 3� 3�1�1�1¼1125 simulation runs. Asimilar procedure for reduction was not applicable for the first twostudied Cases 1 and 2.

For Case 3, the obtained brute-force (1125 candidate solutions)is presented in Fig. 14 in square symbols while all the candidatesolutions which satisfy the thermal comfort limit (DH24�1000 �C h) are symbolized by circles. By indicating the obtainedsolutions of Case 3 in the figure, it can be seen that the obtainedsolutions lie on the front side of the circle points (feasible solu-tions). This verifies the dominating nature of the obtained optimalsolutions.

8. Comparison of the three cases

The preceding cases addressed three different levels of over-heating (DH24� 6250, 2400, and 1000 �C h). Fig. 15 shows theduration curve of the operative and mean air temperatures at thewarmest zone. The optimal solutions 41 (with the minimumoverheating in Case 1, DH24¼ 4181 �C h), 46 (with the upper boundoverheating in Case 2, DH24¼ 2400 �C h), and 20 (with the upperbound overheating in Case 3, DH24¼1000 �C h) are selected,respectively, to evaluate the temperature duration curves for thethree studied cases.

The energy consumption of building depends significantlyon thecriteria set for the indoor environment (e.g., thermal comfort crite-rion). However, the emission production is not only dependent onenergy consumption but also on the heating/cooling energy sourceaswell as on the efficiencyof theHVAC systems. Fig.16 compares theoptimal solutions (trade-off relations) of the studied cases in terms

of CO2-eq emissions and investment cost, considering the threedifferent levels of overheating inpresenting the significant influenceof heating system type on the optimal solutions.

From some observations in Fig. 16 at the value of CO2-eq emis-sions (50 kg CO2-eq/m2 a), a higher investment cost is needed inCase 2 compared with Case 1 in order to lower the overheatingindex DH24. This can be shown by points (A) and (B). Point (A) hasDH24 of 5400 �C h which is higher than that for point (B), which is2200 �C h. For the latter, a lower building tightness, shading optionand relatively thin insulation were implemented to decrease theamount of summer overheating. However, this required a lot ofspace heating energy (points A and B have space heating energiesof 70 and 130 kWh/m2 a, respectively). To keep the same level ofCO2-eq emissions (equal environmental impact), the optimizationsolver selected the Oil Fire Boiling system (Sys. 2) which hasa lower emission factor (EF¼ 0.267 kg/kWh) for point (B) insteadof the Electrical Radiator (Sys. 1) (EF¼ 0.459 kg/kWh) which wasthe selection of point (A). This required an additional investmentcost of 9150 V (price difference between the heating systems 1and 2). Furthermore, the shading option costs point (B) additional4000 V to decrease the direct solar radiation. However, a 3150 V

investment cost was saved by using less insulation and less effi-cient building tightness. As a result, an additional cost (10 000 V)was needed for higher thermal comfort conditions (point B) tomaintain the same impact on the environment (50 kg CO2-eq/m2 a).

Finally, it is worthwhile to mention that, on average Case 1 usedlower Ubldg (average Ubldg¼ 0.24 W/m2 K) to attain minimumamounts of heating energy (consequently CO2-eq emissions).


However, Case 2 implemented higher Ubldg (averageUbldg¼ 0.38 W/m2 K) for fewer summer overheating solutions(DH24� 2400 �C h) with and without a cooling option. The inter-mediate values of Ubldg (average Ubldg¼ 0.3 W/m2 K) with thecooling option are selected in Case 3.

9. Conclusions

Many of the building energy regulations have only buildingenvelope (U-values) and heat recovery requirements. The impor-tance of this study is that it extends the scope to include the supplyheating energy source and the shading option as well, aiming toachieve low-emission cost-effective thermal-comfort solutions formodern buildings.

For this purpose, a modified multi-objective optimizationapproach is proposed and combined with IDA ICE (buildingperformance simulation program) addressing three studied caseswith different thermal overheating levels. A set of optimal solutionsis obtained and analyzed. The analysis shows the physical meaningbehind the optimizer’s selections (decisions) and helps to under-stand the simultaneous influence of the design parameters on thebuilding emissions, investment cost, and thermal comfort. Themost important conclusions are

� The influence of the supply heating system type on the resultsis more significant than the influence of the other designvariables. Using a low-emission supply heating system reducesnot only the emissions related to space heating energy, but alsothe emissions related to domestic hot water and system heat-ing loss energies, which are considerable in cold climatecountries.

� For environmental and indoor thermal-comfort considerations,investing in high-price low-emission supply heating systemcould be a better solution than investing in additional insu-lation and other low-energy measures.

� Additional insulation reduces the space heating energyconsumption. However, this increases the overheating duringthe summer. High efficient heat recovery units are a goodsolution for ‘low emission’, ‘low summer overheating’ dwell-ings. For economical and practical considerations, heatrecovery unit with annual efficiency in between 60 and 80% isa reasonable selection.

� Without considering the summer overheating level, the opti-mization approach achieved a set of solutions which aresignificantly better than the initial designs using additionalinsulation. This means that additional insulation could berequired if operable windows (free cost solution) are availablefor sufficient natural ventilation and cooling.

� In cold climates, acceptable summeroverheating levels could beachieved without a cooling option. However, cooling systemand shadingoption are required forhigh thermal comfort levels.

The simulation-based optimization approaches show a greatpotential for the solution of multi-objective building design prob-lems, and can be used in the design phase to give a better under-standing for the performance of the building and its HVAC systems.

Acknowledgement

The authors would like to acknowledge the financial support ofthe Finnish National Technology Agency (TEKES), as part of theMASI programme, as well as the following supporting companies:Optiplan Oy, Pöyry Building Servicees Oy, Saint-Gobain Isover Oy,Skanska Oy, and YIT Oyj.

References

[1] Vine E. Opportunities for promoting energy efficiency in buildings as an airquality compliance approach. Energy 2003;28(4):319e41.

[2] Carsten P, Boermans T, Harnisch J. Mitigation of CO2 emissions from the EU-15building stock beyond the EU directive on the energy performance of build-ings. Ecofys GmbH. Eupener Straße, 59 50933 Cologne, Germany; 2006.

[3] Balaras Constantinos A, Droutsa Kalliopi, Dascalaki Elena,Kontoyiannidis Simon. Heating energy consumption and resulting environ-mental impact of European apartment buildings. Energy and Buildings2005;37:429e42.

[4] Balaras CA, Droutsa K, Argiriou AA, Asimakopoulos DN. Potential for energyconservation in apartment buildings. Energy and Buildings 2000;31(2):143e54.

[5] Thyholt M, Hestnes AG. Heat supply to low-energy buildings in districtheating areas: analyses of CO2 emissions and electricity supply security.Energy and Buildings 2008;40:131e9.

[6] Gustavsson L, Joelsson A. Life cycle primary energy analysis of residentialbuildings. Energy and Buildings 2010;42:210e20.

[7] Rolfsman B. CO2 emission consequences of energy measures in buildings.Building and Environment 2002;37:1421e30.

[8] Hasan A, Vuolle M, Siren K. Minimisation of life cycle cost of a detached houseusing combined simulation and optimization. Building and Environment2008;43(12):2022e34.

[9] WangW, Zmeureanu R, Rivard H. Applying multi-objective genetic algorithmsin green building design optimization. Building and Environment2005;40:1512e25.

[10] Verbeeck G, Hens H. Life cycle optimization of extremely low energy dwell-ings. Journal of Building Physics 2007;31:143.

[11] Lee WL, Yik FWH. Regulatory and voluntary approaches for enhancingbuilding energy efficiency. Progress in Energy and Combustion Science2004;30(5):477e99.

[12] Hamdy M, Hasan A, Siren K. Combination of optimization algorithms fora multi-objective building design problem. IBPSA: 11th International BuildingPerformance Simulation Association Conference, Glasgow-UK; 2009.

[13] The National Building Code of Finland C3. Thermal insulation in a building,regulations. Helsinki: Ministry of the Environment; 2007: p. 9.

[14] The National Building Code of Finland D5. Calculation of power and energyneeds for heating of buildings, guidelines. Helsinki: Ministry of the Environ-ment; 2007: p. 72.

[15] D2 Finland Code of building Regulation. Indoor climate and ventilation ofbuildings, regulation and guidelines. Helsinki: Ministry of Environment;2003.

[16] Pylsy P, Kalema T. Concepts for Low-Energy Single-Family Houses. TampereUniversity of Technology. Department of Mechanics and Design. ResearchReport 1; 2008.

[17] Heljo J, Laine H. Report 2. Institute of Construction Economics. TampereUniversity of Technology: Finland; 2005.

[18] www.tamar.fi.[19] Deb K. Multi-objective optimization using evolutionary algorithms. John

Wiley and Sons; 2001.[20] Sahlin P, Eriksson L, Grozman P, Johnsson H, Shapovalov A, Vuolle M. Whole-

building simulation with symbolic DAE equations and general purposesolvers. Building and Environment 2004;39(8):949e58.

[21] Achermann M, Zweifel G. RADTEST radiant cooling and heating test cases.A Report of Task 22, Subtask C. Building Energy Analysis Tools. ComparativeEvaluation Tests, IEAdInternational Energy Agency, Solar Heating and Cool-ing Programme; April 2003.

[22] Crawley DB, Hand JW, Kummert M, Griffith BT. Contrasting the capabilities ofbuilding energy performance simulation programs. In: Proceedings of the 9thIBPSA Conference, Montreal, Canada; 2005. p. 231e238.

http://www.tamar.fi

applying a multi-objective optimization approach for design of low-emission cost-effective dwellings

Documents