Simulation-based evolutionary optimization for energy-efficient layout plan design of high-rise residential buildings

Simulation-based evolutionary optimization for energy-efficient layout plan design of high-rise residential buildings

Journal of Cleaner Production 231 (2019) 1375e1388 Contents lists available at ScienceDirect Journal of Cleaner Production journal homepage: www.els...

5MB Sizes 0 Downloads 10 Views

Journal of Cleaner Production 231 (2019) 1375e1388

Contents lists available at ScienceDirect

Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro

Simulation-based evolutionary optimization for energy-efficient layout plan design of high-rise residential buildings Vincent J.L. Gan a, H.K. Wong a, K.T. Tse a, Jack C.P. Cheng a, Irene M.C. Lo a, b, C.M. Chan a, * a b

Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Hong Kong Institute for Advanced Study (IAS), The Hong Kong University of Science and Technology, Hong Kong

a r t i c l e i n f o

a b s t r a c t

Article history: Received 20 September 2018 Received in revised form 10 April 2019 Accepted 27 May 2019 Available online 28 May 2019

Buildings consume 40% of global energy, in which residential buildings account for a significant proportion of the total energy used. Previous studies have attempted to optimize the layout plan of residential buildings for minimizing the total energy usage, mainly focusing on low-rise houses of a regular shape and having a limited number of design variables. However, layout design for high-rise residential buildings involves the complicated interaction among a large number of design variables (e.g., different types of flats with varying configurations) under practical design constraints. The number of possible solutions may increase exponentially which calls for new optimization strategies. Therefore, this study aims to develop an energy performance-based optimization approach to identify the most energyefficient layout plan design for high-rise residential buildings. A simulation-based optimization method applying the evolutionary genetic algorithm (GA) is developed to systematically explore the best layout design for maximizing the building energy efficiency. In an illustrative example, the proposed optimization approach is applied to generate the layout plan for a 40-storey public housing in Hong Kong. The results indicate that GA attempts to maximize the use of natural-occurring energy sources (e.g., wind-driven natural ventilation and sunlight) for minimizing 30e40% of the total energy consumption associated with air-conditioning and lighting. The optimization approach provides a decision support basis for achieving substantial energy conservation in high-rise residential buildings, thereby contributing to a sustainable built environment. © 2019 Published by Elsevier Ltd.

Keywords: Building design optimization Built environment Energy efficiency Genetic algorithm High-rise residential building Layout plan

1. Introduction The building sector has a considerable impact on the environment, since it accounts for one third of greenhouse gas emissions and 40% of the energy consumption worldwide (UNEP, 2009). In a high-rise high-density city such as Hong Kong, buildings can even account for 60% of the carbon emissions and 90% of total electricity consumption (HKEPD, 2012). As the residential building sector consumes approximately 30% of the total energy used worldwide (Allouhi et al., 2015), it becomes important to consider the energy performance in residential building design in order to reduce the global energy usage. Wan and Yik (2004a) investigated energy enduse characteristics of residential buildings, which shows that air-

* Corresponding author. E-mail addresses: [email protected] (V.J.L. Gan), [email protected] (H.K. Wong), [email protected] (K.T. Tse), [email protected] (J.C.P. Cheng), [email protected] ust.hk (I.M.C. Lo), [email protected] (C.M. Chan). https://doi.org/10.1016/j.jclepro.2019.05.324 0959-6526/© 2019 Published by Elsevier Ltd.

conditioning and lighting contributed a significant proportion of the electricity used in building operation. As such, many studies have focused on the improvement of air-conditioning and lighting in residential buildings (Yik and Bojic, 2006; Yik and Lun, 2010). Peuportier et al. (2013) and Gan et al. (2018) proposed methods using thermal simulation and life cycle assessment to account for the energy efficiency in building design, considering various architectural and technical characteristics (e.g., envelope insulation materials). Yik and Lun (2010) showed that around 25% of the energy consumption in a residential building can be saved with the use of natural ventilation. The preceding case studies have demonstrated considerable energy reductions by adopting certain energy-efficient measures. Some studies have also described the use of advanced analytical methods (such as optimization) to further improve the building energy efficiency (Dovì et al., 2009; Hamdy et al., 2013; Tuhus-Dubrow and Krarti, 2010). It was found that designing buildings with better shape, at the right orientation, can reduce the energy consumption by as much as 30e40% (Cofaigh EO et al., 1999). Wang (2005) and

1376

V.J.L. Gan et al. / Journal of Cleaner Production 231 (2019) 1375e1388

Wang et al. (2006) proposed an optimization method using evolutionary genetic algorithm (GA) to explore the optimal building shape and the envelope characteristics (such as windows) in the exterior surface of a multi-story building. The GA is an efficient search heuristic where the fittest individuals are selected for reproduction in order to search for a diverse range of qualified design solutions (Wetter and Wright, 2004). Similarly, Fesanghary et al. (2012) focused on the optimization of the envelope characteristics in a low-rise residential house using a GA multi-disciplinary optimization method, aimed at minimizing the life cycle cost and carbon emissions. Gagne and Andersen (2012) applied GA to maximize the daylight performance by varying the building façade design (such as windows, overhangs and shading devices), considering the trade-off between the lighting energy consumption and the indoor illuminance for a good visual environment. Some studies (Ascione et al., 2016; Wright et al., 2002) have emphasized the optimization of air-conditioning and ventilation systems in buildings (e.g., water flow and fan size) for improved thermal discomfort and lower operating cost. Hamdy et al. (2011) optimized the configuration of building systems (e.g., energy sources, heat recovery types, cooling systems, etc.) in single-family houses using GA multi-objective optimization. The studies aforementioned emphasized the optimization of building shape, envelope characteristics, and building system and control (Evins, 2013). Some studies have tried to optimize the building layout plan, with the aim of exploring the physical arrangement of spaces in a building in order to obtain the best overall performance (Bao et al., 2013). Bojic et al. (2001) evaluated and compared the energy performance of two representative flat layout designs for high-rise residential buildings with various thermal construction methods and insulation materials. Merrell et al. (2010) established a stochastic optimization method based on Bayesain network trained via real-world data for optimizing the layout plan of living rooms, bedrooms, kitchen etc. in a low-rise house. Flager et al. (2009) integrated energy simulation with GAbased multi-objective optimization for optimizing the configuration of an individual room (such as the dimension of structural components) for less energy consumption and structural cost. The preceding studies have demonstrated innovative design methods for energy reductions in an individual flat/room, low-rise houses, and multi-story residential buildings. For high-rise residential buildings, most engineers still rely on empirical methods and trial-and-error processes to explore the energy-efficient design. The layout design for high-rise residential buildings involves the complicated interaction of large numbers of design variables (such as different flats with various configurations) under real-life design constraints, so the number of possible solutions would increase exponentially with size of the problem (Gan et al., 2019a). Specifically, a typical high-rise residential building often comprises various parts (often referred to as wings) with each part consisting of a large number of flats of different sizes to meet the needs of tenants and site constraints. The combination of different types of flats can result in a large number of possible designs and substantially change the building layout plan, which in turn affects the interaction between the building and the ambient environment, as well as the overall energy performance. This critical problem has not yet been solved in the literature, and therefore a new optimization strategy is required for exploring the optimal layout plan for high-rise residential buildings. Another challenge for the layout optimization of high-rise buildings is the timevarying ambient environment (e.g., temperature, wind, and atmospheric pressure) that changes significantly with the building height (Pan et al., 2017). The strong wind and complex ambient environmental conditions impose considerable impacts on the thermal load and the energy performance in buildings, which have to be incorporated during the optimization process.

In this paper, attempts have been made to develop an efficient energy performance-based optimization approach to identify the optimal layout of high-rise residential buildings for minimizing the energy demand during building operation (such as air-conditioning and lighting). The optimization problem is mathematically formulated and the evolutionary GA is used to generate a diverse range of layout plans with consideration of various design constraints. The effects of ambient wind and environmental conditions on the thermal load of a building are quantified during the optimization process. A customized program is developed to support the wideranging exploration of the design space and to identify the most energy-efficient layout plan design for high-rise residential buildings. The proposed optimization approach provides insights regarding the critical parameters relevant to the ultimate energy performance of high-rise buildings, and serves as a decision support tool for minimizing the building energy consumption, contributing to a sustainable built environment. 2. Methodology The proposed approach follows the process of simulation-based evolutionary optimization (Gan et al., 2019a). As Fig. 1 shows, it starts by creating an initial population of building layout which is then applied to generate a representative set of alternative layout plans via crossover and mutation in the evolutionary GA. Energy simulation is then performed to determine the electricity consumptions associated with air-conditioning and lighting for each layout plan. The energy performance results are subsequently used to guide the GA towards finding the optimum among a range of design solutions. The developed optimization approach has several advantages. First, the proposed approach applies GA, a stochastic metaheuristic, for the wide-ranging exploration of the large design space to identify the optimal design solution, which is more robust than the empirical trial-and-error methods. The proposed simulation-based optimization approach evaluates each individual layout plan in energy simulation engines, taking into consideration the local weather profile and climatic data. The complex effect of exterior environment on building energy consumption can be evaluated during the optimization process. The detail processes are explained in following sections. 2.1. Problem formulation 2.1.1. Representation of building layout plan A typical high-rise residential building often comprises various parts (often referred to as wings) with each part consisting of different types of flats. As shown in Fig. 2, we first parameterize each building layout plan as G4. Each layout plan G4 is represented as the union of a set of wings, Rw, which can be described by its geometric attributes (i.e., width mw, length nw, story height hw) and position attributes (i.e., origin (aw, bw) at the end of one side of the corridor, and the rotation angle qw). To facilitate interpretation, the attributes of wing Rw are encoded as a vector of six variables [mw, nw, hw, aw, bw,qw]. Each individual wing Rw can be further represented by the union of a set of flats. Fwl stands for a specific flat l in wing w, which is parameterized using its size attributes (i.e., flat area zwl, width gwl, length pwl), position attributes (i.e., center (xwl, ywl) and rotation mwl related to the origin of wing w), as well as the functional attributes (i.e., number of occupants wwl). Likewise, the attributes of flat Fwl are encoded as a vector of seven variables [zwl,gwl, pwl, xwl, ywl, mwl, wwl]. Given the attributes above-mentioned, a parameterized layout plan G4 can be interpreted as a union of multiple wings (i.e., ∪w¼1WRw), where each wing Rw is further represented by a union of different flats (i.e., {F 1w ; F 2w ; …; F Lww }), as follows:

V.J.L. Gan et al. / Journal of Cleaner Production 231 (2019) 1375e1388

1377

Fig. 1. Framework for the proposed evolutionary optimization approach.

RW ¼ fF 1W ; F 2W ; …; F LWW g

(

G4 ¼n∪W w¼1 fR1 ;R2 ;…;R o W gn

R1 ¼ F 11 ;F 21 ;…;F L11 ;R2 ¼ F 12 ;F 22 ;…;F L22

o

o n ;RW ¼ F 1W ;F 2W ;…;F LWW (1)

where W refers to the total number of wings in the layout plan and Lw stands for the total number of flats in the particular wing w. As such, Eq. (1) represents the configuration of a building layout plan containing different wings Rw for every w∈W, and each wing Rw w consists of a specific flat arrangement, i.e., ∪Ll¼1 F lw ðw ¼ 1; 2; …; WÞ. By changing the wing and the flat attributes, the GA generates new layout plans with different shapes, orientations and flat arrangements for subsequent evaluation and comparison. 2.1.2. Objective function Eq. (2) refers to the objective function, aimed at guiding the GA engine to minimize the total energy consumption from the major operational stage activities (i.e., air-conditioning and lighting). Eqs. (3) e (9) determine how the thermal load and lighting load are quantified. The layout plans generated by the GA need to satisfy practical design constraints (such as floor area, number of occupants), which are described in detail in Section 2.1.3. Specifically, a candidate layout plan is considered for evaluation as long as it satisfies the design constraints, and the optimum is the one with the least total energy consumption. 4

minimize TECðG Þ ¼

" Lw W X T X X w¼1 l¼1

t¼1

Llw ðtÞ

P lw

þ

T X t¼1

AC lw ðtÞ

Q lw ðtÞ

#

COP lw (2)

subject to:

Llw ðtÞ ¼

( AC lw ðtÞ

¼

Q lw ðtÞ ¼

l

1; Elw ðtÞ < Ew ðtÞ L Rlw ðtÞ ¼ 1 0; Otherwise

I X i¼1

(3)

l

1; T lw ðtÞ  T w ðtÞ L Olw ðtÞ ¼ 1

(4)

l

0; T lw ðtÞ < T w ðtÞ r Olw ðtÞ ¼ 0

CQ i ðtÞ þ

J X j¼1

SQ j ðtÞ þ

D X d¼1

IQ d ðtÞ þ

U X

NQ u ðtÞ

(5)

u¼1

wherein TEC(G4) represents the total energy consumption of airconditioning and lighting for a building layout plan G4 (kWh), w stands for a specific wing in the building layout plan, l represents a specific flat in wing w, t refers to a particular time step (hour). Lwl(t) refers to the operating status of lighting at time t for flat l in wing w (the value is equal to 1 h if the lighting is turned on and zero if the lighting is switched-off), Pwl represents the lighting power (kW). ACwl(t) refers to operating status of the air-conditioning at time t for flat l in wing w (the value is equal to 1 h if the airconditioning is turned on and zero if switched-off), Qwl(t) stands for the thermal load at time t for flat l in wing w (kW), and COPwl is the coefficient of performance for the air-conditioning system. Eq. (2) evaluates the total amount of energy consumed by the thermal and lighting control systems for different wings W over a period of time T. As Eq. (3) shows, the operating status Lwl(t) of the lighting depends on the indoor illuminance (Ewl(t) in lm/m2) that is incident on a reference surface specified by users, the minimum illuminance (from user-input) required to maintain a comfortable visual environment Ewl(t) (lm/m2), and the status of the occupantsRwl(t) (i.e., one means that flat l is occupied and the

1378

V.J.L. Gan et al. / Journal of Cleaner Production 231 (2019) 1375e1388

Fig. 2. Representation of building layout plan. (a) Geometric and position attributes of a wing. (b) Size, position and functional attributes of a flat.

occupant(s) is/are awake, whereas zero refers to other situations). The lighting is turned on when the indoor illuminance Ewl(t) is lower than the minimum requirement Ewl(t) and the occupant(s) is/are active, i.e., Rwl(t) ¼ 1. Otherwise, lighting is turned off for energy-saving purposes. The indoor illuminance can be obtained via daylighting modeling, taking account of the location, orientation, window size, etc. of the building. Eq. (4) shows that the operating status ACwl(t) of the airconditioning depends on the average indoor temperature Twl(t) in  C, the thermostat set-point temperature Twl(t) in  C, and the flat occupancy Owl(t) (i.e., one means occupied and zero means unoccupied). The air-conditioning is turned on when the indoor temperature Twl(t) is higher than the thermostat set-point temperature Twl(t) and the flat is occupied, i.e., Owl(t) ¼ 1. Otherwise, the air-conditioning is turned off for the purpose of energy-saving. As shown in Eq. (5), the thermal load Qwl(t) is calculated as the total amount of heat gain or loss from various sources that needs to be controlled to maintain an acceptable indoor air temperature. As shown in Fig. 3, the heat exchange for a flat mainly consists of: (i) heat gain by conduction CQi(t) through opaque (like walls, floors and ceilings) and transparent elements (like windows); (ii) heat

gain of solar radiation SQj(t)through transparent elements; (iii) heat gain from human bodies, household equipment, and lighting IQd(t); and (iv) heat gain due to infiltration and/or ventilation air NQu(t). It should be noted that the heat gain may be absorbed by thermal mass (e.g., walls and slabs) during the day and released into flat throughout the night, creating a time delay and a change in the predicted heat gain. Heat-gain weighting factor method (Kerrisk, 1981) is used to quantitatively account for the time shift and to adjust the predicted heat gain/thermal load (i.e., how much heat is stored and released for every time step). The models for weighting factor calculation consider the room characteristics such as total heat capacity, material property (e.g., density, solar absorbability), thickness and surface area of the element, and energy incident on the surface. Heat gain by conduction CQi(t) through an element i (kW) can be calculated as follows: 0

CQi ðtÞ ¼

T ðtÞ  T lw ðtÞ

1 h 1 þ Li k1 i þh 0

Ai

(6)

in which Twl(t) refers to the indoor air temperature at time t for flat

V.J.L. Gan et al. / Journal of Cleaner Production 231 (2019) 1375e1388

1379

Fig. 3. Heat exchange between an individual flat and the outdoor ambient environment.

l in wing w ( C). T′(t) is the air temperature for the outdoor environment or other flats attached to flat l ( C). h represents the air heat transfer coefficient for flat l whereas h' refers to the heat transfer coefficient for the outdoor environment or the other flats attached to flat l in kW/(m2  C). Li is thickness of the element i (m), ki is the thermal conductivity of element i in kW/(m  C), and Ai stands for the effective area of i (m2). In the case of a building element facing the outdoor environment, the heat conduction also needs to consider solar radiation flux incident on the element. The solar radiation absorbed by the element and flowing into the flat F D F can be quantified as f½SD i ðtÞ gi þSi ðtÞ gi  ti  DQIR g Ai and added to CQi(t). SiD(t) and SiF(t) are the direct and diffuse solar radiation received per unit area (kW/m2), considering solar direction, building orientation (incident angle), overhang, etc. giD and giFare the absorbability for direct and diffuse solar radiation. ti is the coefficient of shading for element i, and DQIR stands for the infrared radiation thermal loss (kW/m2). Solar radiation through a transparent element j at a particular time t (kW) can be quantified as follows:

SQj ðtÞ ¼

h

i F D F SD j ðtÞ aj þ Sj ðtÞ aj Aj tj

ACH V 3600

(9)

in which rair is the air density (kg/m3), Vwl(t) represents the volumetric airflow rate at time t for flat l in wing w (m3/s), CT stands for the specific heat of air in J/(kg C), Twl(t) refers to the indoor air temperature for flat l in wing w ( C),T′(t) is the air temperature for the outdoor ambient environment or for the flats attached to flat l ( C). The volumetric airflow rate mainly depends on the number of air changes per hour (ACH) and the volume of the flat (V). The ACH value can be predicted from the building surface pressures, driven by wind and buoyancy effects. Since high-rise buildings are exposed to time-varying wind speeds and directions, the building surface pressures and the ACH for a particular flat changes from time to time which in turn influences the heat transfer by infiltration and the energy performance. The proposed optimization approach uses an physical modeling algorithm, i.e., the ShermanGrimsrud algorithm (Deru and Burns, 2003; Sherman and Grimsrud, 1980), to take account of the effect due to the timevarying wind environment.

(7)

in which SjD(t) and SjF(t) are the direct and diffuse solar radiation striking the transparent element (kW/m2). ajD and ajFare the transmittance for direct and diffuse solar radiation. Aj stands for the effective area of element j (m2), and tj refers to the coefficient of shading for j. IQk(t) is the heat gain from an indoor heat source k (such as occupants, household equipment and building systems like lighting) (kW). IQk(t) is calculated by summing all the heat sources K and then multiplying by their corresponding per-unit heat generation rate (such as kW/person for occupants and kW/m2 for the lighting intensity). NQu(t) refers to the heat gain due to infiltration and/or ventilation through an opening u (kW), which can be calculated via the formulae below:

 0  NQu ðtÞ ¼ CT rair V lw ðtÞ T ðtÞ  T lw ðtÞ

V lw ðtÞ ¼

(8)

2.1.3. Building design constraints Any layout needs to conform to building design constraints before evaluating its energy performance. This study formulates the design constraints as a set of equality/inequality equations {c1,c2, …} over the parameterized layout plan G4, e.g., 1(G4) ¼ 1 means that the layout meets the first constraint whereas 1(G4) ¼ 0 indicates an invalid layout. A set of design constraints arises out of building (planning) regulations, functional requirements, or livability considerations. For building (planning) regulations, a layout should lie inside the designated site boundary and the building corridor should have a maximum length to facilitate fire evacuation. Specifically, Eq. (10) shows that any flat Fwl belonging to wing Rw and layout plan G4 should be located inside the site boundary denoted by ∪B. Eq. (11) constrains the distance between any flat Fwl and the closest exit E (such as a staircase or elevator lobby), in which Cx stands for the maximum allowable length considering the accessibility and fire evacuation.

1380

V.J.L. Gan et al. / Journal of Cleaner Production 231 (2019) 1375e1388

F lw 2 ∪B⇔ dRw 2G4 ;

F lw 2Rw

(10)

  Distance F lw ; E  Cx ; l ¼ 1; 2; …Lw ; w ¼ 1; 2; …W:

(11)

Other design constraints include functional requirements (e.g., a layout must satisfy the target floor area, the geometric attributes should not exceed a certain specified aspect ratio, etc.) and livability considerations (e.g., a layout should accommodate a targeted number of occupants, the angle between different wings should not exceed a minimum value to provide privacy, etc.). The functional requirements and livability constraints can be formulated as: Lw W X X

zlw  Cε

(12)

w¼1 l¼1

€  zl  C ; l ¼ 1; 2; …L ; w ¼ 1; 2; …W: C w s s w

(13)

€  mw  C ; w ¼ 1; 2; …W: C b b nw

(14)

Lw W X X

wlw  C d

(15)

w¼1 l¼1





l Rw ; R*w  Cl

(16)

Eqs. (12)e(14) are functional requirements: Eq. (12) specifies Lw W P P that the total area of all flats in a layout plan ( zlw Þ must be w¼1 l¼1

more than or equal to the target floor area Cε. Eq. (13) is used to control the area (zwl) for each individual flat, wherein C€sand Csrefer to the minimum and maximum flat areas of Fwl. Eq. (14) represents the geometric constraints limiting the aspect ratio (mw/nw) for different wings, in which C€band Cbstand for the

minimum and maximum allowable values. Eqs. (15) and (16) are livability constraints: Eq. (15) denotes that the expected number of Lw W P P occupants ( wlw ) should be more than or equal to the target w¼1 l¼1

number of residents Cd. Eq. (16) limits the angle l between two adjacent wings Rw and Rw* to be more than or equal to a minimum allowable value Cl. In addition to the design constraints abovementioned, more design constraints can be added by users and strategically set as active or inactive in order to guide the GA to produce the desirable layout plans. 2.2. Exploration of optimal design using genetic algorithm (GA) This study applies GA to explore a diverse set of initial and alternative layout plans {G4}. It requires specific user inputs as well as the appropriate definition of numerical optimization like population size and crossover/mutation probability. Regarding the user input, the configuration of flat units that form the layout plan of a residential building needs to be provided first. This includes the dimension and detailed arrangement of the flat units as well as the construction details such as wall thickness and insulation layer conductivity. Moreover, users need to provide a weather file containing the local climate data, the indoor occupancy pattern, and the minimum requirement for daylighting and ventilation. Users can have a flexible definition on the design constraints (such as the building site boundary, the target floor area, the number of occupants, etc.) to control the GA optimizer in generating different types of layout plans. The information is then utilized to produce a set of representative layout plans, complying with the pre-defined site conditions and design constraints. Each individual layout plan is evaluated in verified energy simulation engines to determine its energy consumption associated with air-conditioning and lighting. Once the energy performance of each layout plan is evaluated, the results are subsequently used to guide the GA towards finding the optimum among a range of design solutions. As shown in Fig. 4, generation of a building layout plan consists of two stages: (i) a combination of different wings to form the

Fig. 4. (a) Interpretation and (b) exploration of building layout plans using GA.

V.J.L. Gan et al. / Journal of Cleaner Production 231 (2019) 1375e1388

1381

Fig. 5. Overview of building layout plan and site boundary for the original design, modified from our preliminary studies (Gan et al., 2019a, 2019b).

building shape at a specific orientation, and (ii) an arrangement of different flats into each individual wing. This two-stage layout optimization is tested, which shows reduced complexity due to the possible large number of design variables while providing the flexibility to produce a diverse range of layout designs (Gan et al., 2019b). In the GA optimization, each building layout plan is represented by a bit string, in which every digit interprets a specific characteristic of the layout plan. The bit string consists of two major parts, the first of which describes the building-level geometry including the number of wing and elevator lobby/stair, the building orientation, and the shape of the building (e.g., angle between two adjacent wings). The second part interprets the detailed flat arrangement for each individual wing, including the size, number, and assignment for different kinds of flats. As shown in Fig. 4, the GA first produces an initial population of randomly generated building layout plans called parents. Each parent layout is represented by a string of digits, which are altered by crossover and mutation operators in the GA to form a new generation of building layout designs called offspring. To produce the new generation of offspring solutions, a pair of parent layout plans are randomly selected using the stochastic uniform method and the crossover function exchanges a certain part of the bit strings between the selected parent layout plans. Alternatively, mutation stochastically alters one or more digit values in the bit string to produce a new layout plan. This is an iterative process with the population of building layouts in each iteration called a generation. The offspring layout plans from the GA are subsequently evaluated in energy simulation engines to determine their electricity consumptions associated with air-conditioning and lighting. The GA optimizer also checks whether the offspring solutions

meet all the design constraints. The energy simulation results are then used to guide the GA towards finding the most energyefficient design option. Specifically, once the energy performance of each individual layout plan is evaluated, the fittest solution with the least energy consumption can be determined. If the fittest solution converges to the same value for a certain number of generations (e.g., ten generations) and has no further improvement, the optimizer terminates and provides the optimum solution to users. Otherwise, a selection procedure is performed to choose a number of fittest layout plans from the current generation for the next iteration of the algorithm. The GA can be repeated until the optimum layout plan with the least energy consumption is found.

3. Illustrative example In the illustrative example, the GA optimization is implemented in a customized program using MATLAB (MathWorks, 2018). DOE 2.2, a verified energy simulation engine, is taken to compute the hourly thermal load and predict the building energy consumption (DOE-2.2, 2017). As shown in Fig. 5, a 32-story T-shaped public housing in Hong Kong is selected as the original design for evaluating the optimized results. Each floor consists of 18 standardized flats including eight 1/2 Person flats, eight 1-Bedroom flats, and two 2-Bedroom flats. As Table 1 shows, the configuration of the standardized flats (like floor area, maximum occupant, window size, etc.) have been designed to make good use of natural ventilation and lighting (Cheng, 2011; Legislative Council Panel on Housing, 2013). The standardized flats form three separate wings, which are then connected by an elevator lobby in the middle of the building. Two original design scenarios (SI and SII) are used to

Table 1 Standardized flats used in the illustrative example. Configuration

1-2 Person flat

1 Bedroom flat

2 Bedroom flat

3D overview Floor plan a Max No. of occupants Floor area (m2 GFA)

2 14.5

3 30.9

4 35.9

a The dimensions of the standardized flats are created with references to the design of public housing in Hong Kong (Cheng, 2011; Legislative Council Panel on Housing, 2013).

1382

V.J.L. Gan et al. / Journal of Cleaner Production 231 (2019) 1375e1388

Table 2 Scenarios considered in the analysis. Design constraints

a

Original Design

Site boundary No. of occupants b Number for each type of flat c Accessibility d Flat symmetry along corridor Natural ventilation

Optimized Design

SI

SII

S1

S2

S3

S4

✓ 48 ✓ 24 m ✓ 7

✓ 48 ✓ 24 m ✓ ✓

7 48 7 7 ✓ 7

7 48 7 7 ✓ ✓

✓ 48 ✓ 24 m ✓ ✓

✓ 48 ✓ 24 m 7 ✓

a The site boundary for the optimized designs S3 and S4 are the same as the original design SI and SII. b The number of occupants for the four optimized designs (S1eS4) should be equal to or more than that for the original design (i.e., 48). c The constraint on the number for each type of flat ensures that the optimized designs have a similar combination of flat types as the typical public housing design in Hong Kong. d The accessibility constraint ensures the maximum distance between any flat and the nearest elevator lobby or staircase is within the limiting value of 24 m in accordance with the Building Regulation of Hong Kong (Building Ordinance, 2012).

interpret the situations without natural ventilation and with natural ventilation, respectively. In SI, the windows are closed all the time, and building occupants rely on air-conditioning for fresh air supply and space cooling. In SII, the windows are kept opened for natural ventilation. If the indoor air exceeds the set point temperature, air-conditioning is turned on for fresh air supply and space cooling. Energy simulation is performed for the T-shaped public housing, taking into consideration the local weather, typical occupancy pattern, material types, etc. The energy simulation results are then verified with the surveyed electricity consumption from literature and are further compared with the optimized results to evaluate the amount of energy savings. Table 2 summarizes the constraints for the two original design scenarios (SI and SII) and four optimized designs (S1, S2, S3 and S4). S1 considers only the constraint of the total number of occupants on each floor and assumes that the building entirely relies on mechanical ventilation. Based on S1, S2 further enables the use of natural ventilation to evaluate the effects of air ventilation. Both S3 and S4 consider natural ventilation as well as the constraints for the number of occupants, site boundary and accessibility. The site

boundary and the number of occupants used in S3 and S4 are the same as that in the original design (SI and SII). The accessibility constraint ensures the maximum distance between any flat and the closest elevator lobby or staircase is within the limiting value of 24 m in accordance with the Building Regulations of Hong Kong (Building Ordinance, 2012). Compared to S3, S4 does not consider the symmetry of flats on both sides of the corridor. The three types of standardized flats (see Table 1) are used in the generation of building layout plans for S1eS4. The following parameters are specified in GA during the optimization process: population size ¼ 100, crossover probability ¼ 0.8, mutation probability ¼ 0.02, and the maximum number of generation ¼ 500. The energy performance of each layout plan generated by GA is evaluated with certain pre-defined inputs, including material property (such as thermal conductivity, specific heat capacity and density), equipment type (lighting power, air-conditioned & lighting zones, etc.), and occupancy pattern. The lighting system in the residential building does not consider dimming function. The cooling seasons are from April to October (Yik and Lun, 2010), with a suggested thermostat set point temperature equivalent to 25.5  C (Kam, 2009). The weather file for Hong Kong is obtained from an open building performance simulation tool (Green Building Studio, 2018).

4. Results and discussion Fig. 6 shows the annual electricity consumption associated with air-conditioning and lighting for each design scenario. The energy performance results are validated with the surveyed data and the electricity energy trends in Hong Kong's residential housing environment, as shown in Table 3. The annual electricity consumption for Hong Kong's public housing is around 117.4 ± 42.3 kWh/m2 (Cheung et al., 2014), in which around 45% of the electricity consumption is used for air-conditioning and lighting (EMSD, 2016). Therefore, the annual electricity demand for air-conditioning and lighting calculated from the surveyed data is around 52.8 ± 19.0 kWh/m2. Since the surveyed data considers the daily impact of natural ventilation, it will be compared with simulation results for original design SII and optimized designs S2eS4 which have modeled the natural ventilation effect. The floor area of the flat units is 429 m2 for the original design SII (29,200 kWh), and

Fig. 6. Annual electricity consumption associated with air-conditioning and lighting for the original and optimized designs.

V.J.L. Gan et al. / Journal of Cleaner Production 231 (2019) 1375e1388

1383

Table 3 Comparison of simulated electricity consumption for the original & optimized designs with surveyed electricity usage data in Hong Kong's residential housing environment. Calculated from surveyed data Electricity demand for air-conditioning & lighting (kWh/m2) a

a

52.8 ± 19.0

Design scenarios in this study 54.6 to 68.1

The figures are calculated from the surveyed electricity usage data in Hong Kong's residential housing environment (Cheung et al., 2014; EMSD, 2016).

Fig. 7. Variation of monthly electricity consumption for the original and optimized designs.

ranges from 346 to 426 m2 for the optimized designs S2eS4 (19,690e23,750 kWh). Therefore, the electricity consumption per floor area can be calculated as 54.6e68.1 kWh/m2, which agree well with the surveyed electricity demand in literature. It can be concluded that our proposed optimization method can provide accurate estimations for the energy consumption in high-rise residential buildings. The optimized results are further compared with the energy consumption of the original design to measure the amount of energy reduction. Details are discussed in the following sections.

4.1. Original design SI (without natural ventilation) and SII (with natural ventilation) In SI, the occupants fully rely on air-conditioning for fresh air supply and space cooling, whereas SII considers switching between air-conditioning and natural ventilation using an indoor set point temperature of 25.5  C as the threshold. As Fig. 6 shows, enabling natural ventilation in buildings can help reduce the thermal load and the air-conditioning usage by 17% from 35,000 kWh (SI) to 29,200 kWh (SII). The results also indicate that air-conditioning accounts for over 80% of the total energy demand (24,000e30,000 kWh) and significantly outweighs the impact of lighting (5,000 kWh), therefore should be emphasized in design. The large proportion of air-conditioning is mainly due to the subtropical weather of Hong Kong. Specifically, occupants heavily rely on the use of air-conditioning because the average outdoor temperature is relatively high, especially during the hot and humid summer. Fig. 7 shows the variation of monthly electricity consumption for different design scenarios. The electricity consumption starts to increase from April, approaches the maximum in July, and

decreases to a relatively lower level in October. Compared to SI, SII enables natural ventilation and saves the electricity consumption by around 31e47% (1,320e1,450 kWh) from April to May and 20e36% (1,000e1,650 kWh) from September to October. However, the energy reduction between June and August is not significant at all, with less than 1% (40 kWh) of the electricity saving. Fig. 8 compares the cooling electricity consumption of selected flats in April and July for SI and SII. It can also be seen that the electricity consumption is reduced by approximately 54e66% in April and only 1% in July. To explain the reason, Fig. 9 shows the variation of monthly average temperature and the thermostat set point temperature. The outdoor temperature is relatively lower than the thermostat set point temperature (25.5  C) at around 40e50% of the time (e.g., evening) in April, May, September, and October, when natural ventilation could provide cooler air to reduce the heat inside flats. Therefore, the electricity savings in April, May, September, and October are considerable. In contrast, the average outdoor temperature from June to August is much higher than 25.5  C regardless of the time, so natural ventilation mainly carries hot air into the flats if the windows are opened. In such a case, utilizing natural ventilation is not beneficial for maintaining a comfortable thermal environment and reducing the electricity consumption.

4.2. Optimized design S1 and S2 The only constraint used in S1 and S2 is the total number of occupants, the limiting value of which is the same as in the original design. No site boundary constraint has been imposed because these two designs aim to study the optimal layout if there is no site constraint. As shown in Fig. 10, the optimized layout for S1 (without natural ventilation) has an elongated I-shape with two linearly

1384

V.J.L. Gan et al. / Journal of Cleaner Production 231 (2019) 1375e1388

Fig. 8. Air-conditioning electricity consumption of selected flats in April, July and Annual Average for the original design SI (without natural ventilation) and SII (with natural ventilation).

Fig. 9. Thermostat set point temperature in this study and the natural ventilation efficiency for different months (the local temperature data are obtained from Green Building Studio, 2018). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

connected wings (located from west to east), consisting of 24 smallsized flats for 1 or 2 person(s). Compared to the electricity consumption of original design, S1 substantially reduces the annual electricity consumption by 33% to 23,390 kWh (see Fig. 6). As Fig. 7 shows, the monthly electricity consumption is reduced by 30e38% (1,000e2,500 kWh). This is because the majority of the flats in S1 face either south or north, which minimizes the heat gain due to solar radiation coming from the east and west directions. As such, the thermal load and the energy consumption associated with airconditioning for optimized S1 are greatly reduced over the entire cooling season. Furthermore, the layout plan for S1 implies that small-sized flats are preferred in residential building design, because they are less energy-demanding and easier to be cooled than large-sized flats. Specifically, a small 1/2 Person flat requires only one air-conditioning unit, whereas the large-sized flats (e.g., 2Bedroom flats) have more rooms and require multiple airconditioning devices to control the thermal load. Therefore, GA has a preference for small-sized flats to form the layout plan, as long as it satisfies the design constraints considered in optimization. As Fig. 11 shows, the optimal layout for S2 (with natural

ventilation) is the same as S1, but the electricity consumption of S2 (i.e., 19,690 kWh) is 16% less than S1 due to the use of natural ventilation (see Fig. 6). Fig. 12 shows the prevailing wind speeds and directions for Hong Kong over the cooling season. Theoretically, an elongated I-shape layout (located from southwest to northeast) with the majority of flats facing east and southeast can maximize the benefit due to natural ventilation, because there is more time in a year (32%) that cooler air (<25.5  C) comes from east or southeast. However, such arrangements may also increase the heat gain due to solar radiation from the east direction, thereby offsetting the benefit of natural ventilation. The GA needs to strike a balance between the heat exchange via both natural ventilation and solar radiation. The final results indicate that an I-shape layout with two linearly connected wings (located from west to east) is the best, since it also minimizes the heat gain due to solar radiation without significantly compromising the natural ventilation, leading to the most energy saving for air-conditioning. As Fig. 7 shows, the optimized S2 use less electricity consumption than S1, reducing around 1,300e1,900 kWh (30e41%) from April to May and 2,300e2,600 kWh (18e32%) from September to October.

V.J.L. Gan et al. / Journal of Cleaner Production 231 (2019) 1375e1388

1385

Fig. 10. Layout plan for optimized design S1.

4.3. Optimized design S3 and S4 The purpose of S3 and S4 is to study the layout plan, considering more practical constraints as in the original design. Figs. 13 and 14 show the optimized layout plans for S3 and S4, taking account of natural ventilation, number of occupant, site constraints, etc. in optimization. The total electricity consumption for optimized S3 and S4 are 23,750 and 22,490 kWh respectively, which are 19% and 23% less than that of the original design SII (with natural ventilation) (see Fig. 6). Fig. 7 indicates that there is a significant reduction in the electricity consumption for S3 and S4. The monthly electricity saving between June and August is 1,300e1,800 kWh (reducing by 24e27%), which is more much more than the saving from April to May (200e600 kWh) and from September to October (200e700 kWh). Since utilizing natural ventilation from June to August is not beneficial for lowering the indoor temperature (as the outdoor temperature is too high), the reduced electricity consumption in S3 and S4 can mainly be attributed to the reduction in solar radiation. The optimized layouts for S3 and S4 change to a Tshape with three separate wings in order to meet the site boundary constraints. There are twelve 1/2 person flats, six 1-Bedroom flats, and two 2-Bedroom flats for S3. Regarding S4, the number of flat is similar to that of S3 except that S4 misses one 1/2 person flat. Largesized flats (e.g., 1-Bedroom and 2-Bedroom flats) are used because utilizing only small-sized flats (such in S2) will make the wings too long and violate the accessibility constraint. As a result of using large-sized flats, the entire layout plan consumes relatively more electricity than S2. The optimized layout plan becomes less efficient

in terms of the total electricity consumption and the solution optimality is sacrificed. Since large-sized flats are more difficult to be cooled, the GA engine allocates the large-sized flats at the east and the southeast corners of the layout plan. There are more periods in the year when wind carries cool air from east, southeast and southwest directions (see Fig. 12) which can be used to control the air temperature in large-sized flats. Another design implication can be drawn in S4. The design constraints for S4 are the same as for S3 except allowing the design of asymmetric wing. In S3, no large flats are located at the southwest corner, which is also a prevailing wind direction. In S4, however, large flats are no longer located mainly at the east or southeast corners, but are also located at the southwest corner to gain the benefits of natural ventilation. The reason is that in S3 under the restriction of the symmetrical wing, the optimal position of a flat does not solely depend on the flat itself, but also its twin on the opposite side of the corridor. For example, a flat facing east can benefit from natural ventilation, but its twin flat under the symmetrical wing restriction must face west and does not have significant natural ventilation. Without the restriction of the symmetrical wing, different types of flats can be located in their optimal locations to further reduce the energy consumption in the building. 4.4. Recommendations for layout design and optimization The four proposed designs in the illustrative example are optimized, but subject to different kinds of design constraints.

Fig. 11. Layout plan for optimized design S2.

1386

V.J.L. Gan et al. / Journal of Cleaner Production 231 (2019) 1375e1388

Fig. 12. Theoretical orientation of the elongated I-shape layout plan and the prevailing wind directions over the cooling season, AprileOctober (the local wind data are obtained from Green Building Studio, 2018). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 13. Layout plan for optimized design S3.

Specifically, S3 and S4 take account of more real-life design constraints (such as building site boundary, number of occupants, accessibility, etc.), which are the same as the original public housing design. S1 and S2 are optimized designs without the consideration of accessibility and site boundary in optimization. If design constraints can be changed in optimization (such as S1 and S2), the GA optimizer may have more freedom to explore the design space and can generate more energy-efficient designs in response to the exterior environment. In general, layout plan design and optimization for high-rise residential buildings should focus on maximizing the use of natural-occurring energy sources (such as

natural ventilation and sunlight) to replace the use of energyconsuming equipment such as air-conditioning and lighting. Trade-off effects may exist between air-conditioning and lighting e reducing sunlight and solar radiation mitigates the air-conditioning energy consumption, but it may require more lighting for maintaining a good visual environment. The proposed evolutionary optimization approach strives a balance between air-conditioning and lighting loads, and tries to maximize both the thermal and daylight performance of the layout plan. In most cases, the layout design in sub-tropical cities like Hong Kong should emphasize airconditioning because the hot and humid weather will increase the

V.J.L. Gan et al. / Journal of Cleaner Production 231 (2019) 1375e1388

1387

Fig. 14. Layout plan for optimized design S4.

need for air-conditioning, making it significantly outweigh the impact of the other systems, like lighting. Solar radiation and natural ventilation are the two most critical thermal load components for determining air-conditioning energy consumption. Therefore, it is important to optimize the layout plan for minimizing the heat gain of solar radiation while maximizing wind capture for natural ventilation in buildings. First of all, using the elongated I-shape layout (located from west to east) can minimize the solar radiation and the thermal load, leading to substantial reduction in the energy consumption. If there is no restriction on the type of flat, utilizing small-sized flats to form the layout plan is better because they are less energy-demanding and relatively easier to be cooled. Allocating large flats to face the prevailing wind directions (such as east or southeast in Hong Kong) is an energy-efficient practice, because larger flats are more difficult to be cooled and more wind-driven natural ventilation can help control the indoor air temperature. In many cases, flats facing prevailing wind directions can maximize the effect of natural ventilation, but this benefit may be offset by the increased solar radiation. In such circumstances, the proposed optimization approach can be used to analyze the combined effects between solar radiation and natural ventilation, as well to generate the optimum layout plan. 5. Conclusions This paper presents an evolutionary GA optimization approach to identify the optimal layout plan for minimizing the energy consumption of high-rise residential buildings, considering various design constraints. A customized program is developed to systematically explore the most energy-efficient design of high-rise residential buildings. In an illustrative example, the computer program is used to examine a diverse set of energy-efficient building designs and provide insights regarding the critical parameters relevant to building energy efficiency. First of all, smallsized flats are preferred in high-rise residential buildings because they are less energy-demanding and relatively easier to be cooled. If site boundary is considered, an I-shape layout (located from west to east) can minimize the heat gain of solar radiation and can minimize the thermal load. If design constraints (such as site boundary and accessibility) are considered, the layout plan may change to less energy-efficient shapes for meeting the design constraints and solution optimality may be sacrificed. In Hong Kong, allocating large flats on the east or east southeast corners represent an

energy-efficient design, because large flats are more difficult to be cooled and there are more periods in the year when wind carries cooler air from the east to the southeast direction. The actual layout plan may differ for regions with different weathers, site conditions, materials, and occupancy patterns, but the same optimization method can be used to find the corresponding optimal design options. Based on the detailed user-inputs, such as local climate data and site configurations, the proposed evolutionary optimization approach can achieve the optimal layout plan and identify the corresponding energy-efficient designs. Acknowledgement The work described in this paper was partially supported by the Research Grants Council of the Hong Kong Special Administrative Region, China (project no. 16200714). Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.jclepro.2019.05.324. References Allouhi, A., El Fouih, Y., Kousksou, T., Jamil, A., Zeraouli, Y., Mourad, Y., 2015. Energy consumption and efficiency in buildings: current status and future trends. J. Clean. Prod. 109, 118e130. Ascione, F., Bianco, N., De Stasio, C., Mauro, G.M., Vanoli, G.P., 2016. Multi-stage and multi-objective optimization for energy retrofitting a developed hospital reference building: a new approach to assess cost-optimality. Appl. Energy 174, 37e68. Bao, F., Yan, D.-M., Mitra, N.J., Wonka, P., 2013. Generating and exploring good building layouts. ACM Trans. Graph. 32 (4), 122. Bojic, M., Yik, F., Sat, P., 2001. Influence of thermal insulation position in building envelope on the space cooling of high-rise residential buildings in Hong Kong. Energy Build. 33 (6), 569e581. Building Ordinance, 2012. Building (Planning) Regulations Cap.123F. Hong Kong. Cheng, I., 2011. Affordable social housing: modular flat design for mass customization in public rental housing in Hong Kong. In: Proceedings of SB11 Helsinki World Sustainable Building Conference. Helsinki, Finland, 18-21 Oct. Cheung, C., Mui, K., Wong, L., Yang, K., 2014. Electricity energy trends in Hong Kong residential housing environment. Indoor Built Environ. 23 (7), 1021e1028. Cofaigh, E.O., Fitzgerald, E., Alcock, R., McNicholl, A., Peltonen, V., Marucco, A., 1999. A Green Vitruvius - Principles and Practice of Sustainable Architecture Design. James & James. Science Publishers) Ltd, London. Deru, M., Burns, P., 2003. Infiltration and Natural Ventilation Model for WholeBuilding Energy Simulation of Residential Buildings. National Renewable Energy Laboratory, Department of Energy, US. DOE-2.2, 2017. Building Energy Use and Cost Analysis Tool. James J. Hirsch, US. Dovì, V.G., Friedler, F., Huisingh, D., Klemes, J.J., 2009. Cleaner energy for sustainable

1388

V.J.L. Gan et al. / Journal of Cleaner Production 231 (2019) 1375e1388

future. J. Clean. Prod. 17 (10), 889e895. EMSD, 2016. Hong Kong energy end-use data. Electrical & Mechanical Services Department (EMSD), Hong Kong. Evins, R., 2013. A review of computational optimisation methods applied to sustainable building design. Renew. Sustain. Energy Rev. 22, 230e245. Fesanghary, M., Asadi, S., Geem, Z.W., 2012. Design of low-emission and energyefficient residential buildings using a multi-objective optimization algorithm. Build. Environ. 49, 245e250. Flager, F., Welle, N., Bansal, P., Soremekun, G., Haymaker, J., 2009. Multidisciplinary process integration and design optimization of a classroom building. J. Inf. Technol. Constr. 14, 595e612. Gagne, J.M.L., Andersen, M., 2012. A generative façade design method based on daylighting performance goals. J. Build. Perform. Simul. 5 (3), 141e154. Gan, V.J.L., Deng, M., Tse, K.T., Chan, C.M., Lo, I.M.C., Cheng, J.C.P., 2018. Holistic BIM framework for sustainable low carbon design of high-rise buildings. J. Clean. Prod. 195, 1091e1104. Gan, V.J.L., Lo, I.M.C., Tse, K.T., Wong, C.L., Cheng, J.C.P., Chan, C.M., 2019a. BIM-based integrated design approach for low carbon green building optimization and sustainable construction. In: Proceedings of the 2019 International Conference on Computing in Civil Engineering (i3CE). Atlanta, Georgia, USA, 17-19 June. Gan, V.J.L., Tse, K.T., Cheng, J.C.P., Lo, I.M.C., Chan, C.M., 2019b. Site-specific modular design optimization for high-rise residential buildings. In: Proceedings of the Modular and Offsite Construction Summit. Banff, Alberta, Canada, 21-24 May. Green Building Studio, 2018. Weather Data of Hong Kong. Autodesk green building Studio. Available at. https://gbs.autodesk.com/GBS/ (accessed on Januanry 2018). Hamdy, M., Hasan, A., Siren, K., 2011. Applying a multi-objective optimization approach for design of low-emission cost-effective dwellings. Build. Environ. 46 (1), 109e123. Hamdy, M., Hasan, A., Siren, K., 2013. A multi-stage optimization method for costoptimal and nearly-zero-energy building solutions in line with the EPBD-recast 2010. Energy Build. 56, 189e203. HKEPD, 2012. Greenhouse Gas Emission in Hong Kong by Sector. Hong Kong Environmental Protection Department, Hong Kong. Kam, M.S., 2009. 25.5 Deg C and Human Comfort. Electrical & Mechanical Services Department, Hong Kong. Kerrisk, J.F., 1981. Weighting factors. In: York, D.A., Cappiello, C.C. (Eds.), DOE-2 Engineers Manual. Lawrence Berkeley Laboratory, University of California,

Berkeley, US. Legislative Council Panel on Housing, 2013. Modular Flat Design for Public Housing Development. The Hong Kong Housing Authority Legislative Council, Hong Kong. MathWorks, 2018. MATLAB R2018b. The MathWorks Inc., US. Merrell, P., Schkufza, E., Koltun, V., 2010. Computer-generated residential building layouts. ACM Trans. Graph. 29, 181. Pan, W., Qin, H., Zhao, Y., 2017. Challenges for energy and carbon modeling of highrise buildings: the case of public housing in Hong Kong. Resour. Conserv. Recycl. 123 (Suppl. C), 208e218. Peuportier, B., Thiers, S., Guiavarch, A., 2013. Eco-design of buildings using thermal simulation and life cycle assessment. J. Clean. Prod. 39, 73e78. Sherman, M.H., Grimsrud, D.T., 1980. Infiltration-Pressurization Correlation: Simplified Physical Modeling. Lawrence Berkeley National Laboratory, University of California, US. Tuhus-Dubrow, D., Krarti, M., 2010. Genetic-algorithm based approach to optimize building envelope design for residential buildings. Build. Environ. 45 (7), 1574e1581. UNEP, 2009. Buildings and Climate Change e Summary for Decision Makers. United Nations Environment Programme (UNEP). Wan, K.S.Y., Yik, F.W.H., 2004. Building design and energy end-use characteristics of high-rise residential buildings in Hong Kong. Appl. Energy 78 (1), 19e36. Wang, W., Rivard, H., Zmeureanu, R., 2006. Floor shape optimization for green building design. Adv. Eng. Inf. 20 (4), 363e378. Wang, W., Zmeureanu, R., Rivard, H., 2005. Applying multi-objective genetic algorithms in green building design optimization. Build. Environ. 40 (11), 1512e1525. Wetter, M., Wright, J., 2004. A comparison of deterministic and probabilistic optimization algorithms for nonsmooth simulation-based optimization. Build. Environ. 39 (8), 989e999. Wright, J.A., Loosemore, H.A., Farmani, R., 2002. Optimization of building thermal design and control by multi-criterion genetic algorithm. Energy Build. 34 (9), 959e972. Yik, F., Boji c, M., 2006. Application of switchable glazing to high-rise residential buildings in Hong Kong. Energy Build. 38 (5), 463e471. Yik, F.W., Lun, Y.F., 2010. Energy saving by utilizing natural ventilation in public housing in Hong Kong. Indoor Built Environ. 19 (1), 73e87.