Applied Energy 171 (2016) 336346

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Design optimization of office building envelope configurations

for energy conservation
Yu-Hao Lin a, Kang-Ting Tsai b, Min-Der Lin c, Ming-Der Yang d,
a
Centre for Environmental Restoration and Disaster Reduction, National Chung Hsing University, 250 Kuo-Kuang Rd., Taichung 402, Taiwan
b
Program of Landscape and Recreation, National Chung Hsing University, 250 Kuo-Kuang Rd., Taichung 402, Taiwan
c
Department of Environmental Engineering, National Chung Hsing University, 250 Kuo-Kuang Rd., Taichung 402, Taiwan
d
Department of Civil Engineering, National Chung Hsing University, 250 Kuo-Kuang Rd., Taichung 402, Taiwan

h i g h l i g h t s

Building Envelope Energy Load (ENVLOAD) is essential for green building design.

Office Building Envelope design Model (OBEM) was developed based on ENVLOAD.

OBEM provides envelope configurations for architects reference.

Tabu search links OBEM to optimize an office building envelope configuration.

Optimized design reduces construction cost under energy conservation regulations.

a r t i c l e i n f o a b s t r a c t

Article history: Designing envelope configurations of office building with the low construction cost and energy consump-
Received 28 August 2015 tion is a discrete optimization problem. The configuration is currently determined merely on architects
Received in revised form 4 March 2016 experiences resulting in an inefficient expense or by building energy performance simulation which is
Accepted 5 March 2016
time-intensive and involves complex processes. Based on an efficient regression equation to substitute
Available online 23 March 2016
complex energy simulators, this study developed an Office Building envelope Energy performance and
configuration Model (OBEM) to provide envelope configurations, including construction material, sun-
Keywords:
shade type, sunshade length, window number, and window length and width for architects reference.
Green building
Tabu search (TS)
Also, Tabu search, which is effective in solving discrete optimization problems, was integrated with
Optimization OBEM into an Optimal OBEM decision support system (OPOBEM). The OPOBEM was applied to a real
Building envelope office building construction for optimizing its envelope configuration at minimum construction budget
Energy conservation under the energy conservation regulations of green buildings. The result shows that the optimized instal-
lation of sunshade type efficiently reduces solar heat gain according to the high variation of the sunshade
coefficient, thus achieves the goal of energy conservation and reduces the envelope costs. Compared with
architects manual estimation, the optimized envelope design realizes nearly 41% budget savings, thus
demonstrating the feasibility of the proposed OPOBEM.
2016 Elsevier Ltd. All rights reserved.

1. Introduction consumption [5]. Specifically, precise design of building envelope

Constructing green buildings with nearly zero-energy is a major
objective in energy policies worldwide [1]. Building envelope fea-
tures considerably affect the energy efficiency of buildings, indoor
environmental quality and the thermal comfort for human [24];
thus, the effective evaluation of the thermal performance of build-
ing envelopes is crucial in the reduction of air-conditioning energy
Corresponding author. Tel.: +886 4 2284 0440x214; fax: +886 4 22862857.
E-mail address: mdyang@nchu.edu.tw (M.-D. Yang).
can significantly help improve energy efficiencies of building [6].
Several building envelope energy efficiency standards, such as
Overall Thermal Transfer Value (OTTV) equation and Perimeter
Annual Load (PLA), were proposed and revised to apply to different
buildings in different countries according to climate type, analysis
period (such as summer or whole year), or building type [5,7,8].
The OTTV is doubtful to be selected as an effective index in Taiwan,
where cooling load predominates [9]. The PLA proposed by Japan
government is defined as annual thermal load of perimeter spaces
within 5 m of exterior wall [10]. Comprehensive building thermal

http://dx.doi.org/10.1016/j.apenergy.2016.03.018
0306-2619/ 2016 Elsevier Ltd. All rights reserved.
 Y.-H. Lin et al. / Applied Energy 171 (2016) 336346
Nomenclature
Ac annual cooling air-conditioning hours, h/yr
Ai areas of the window glass in the air-conditioned zone in
ith sector, m2
A0i areas of the window glass in the non-air-conditioned
zone in ith sector, m2
AFp total air-conditioning floor areas in building perimeter
zones, m2
Bi wall areas in the air-conditioned zone in the ith sector,
m2
B0i wall areas in the non-air-conditioned zone in ith sector,
m2
Ci glass curtain areas in the air-conditioned zone in ith
sector, m2
C 0i glass curtain areas in the non-air-conditioned zone in
ith sector, m2
Cg unit costs of window glass, NTD/m
Cw unit costs of wall, NTD/m
Cu unit costs of glass curtain, NTD/m
Cr unit costs of roof, NTD/m
Cs unit costs of sunshade board, NTD/m
Di roof areas in the air-conditioned zone in ith sector
D0i roof areas in the non-air-conditioned zone in ith sector,
m2
DH annual degree-hours based on monthly average temper-
ature (298 K), K h/yr
dr sunshade depth rate, %
Ei sunshade board areas in the ith sector, m2
FWi total width of building in ith sector, m
G annual indoor heat gain, W h/m2 yr
IHk isolation-hours on k orientation of building, W h/m2 yr
Ki sunshade coefficient of window in the ith sector
performance simulation models, such as whole building energy
simulation program and Transient Systems Simulation Program
(TRNSYS), were used to facilitate estimating building energy per-
formance [1113]. However, the operation of such simulation pro-
grams is time-intensive and involves complex processes [14,15]. To
simplify the complex processes of estimating the building energy
performance and apply quickly to practical problems, Magnier
and Haghighat used TRNSYS to create a building energy database
for training artificial neural networks (ANNs) and combined a
simulation-based ANNs with a multiobjective algorithm for opti-
mizing building design [16]. Similarly, some studies presented
simple regression expression for predicting annual building energy
performance based on the simulation results of commercial build-
ing energy programs [1719]. Chou and Chang extended the
regression concept to predict a peak cooling load of building [20].
In Taiwan, a building Envelope Energy Load (ENVLOAD) regression
equation was developed by modifying the PLA with local climatic
data to estimate building envelope energy performance [9,18,21
23]. The ENVLOAD representing the total annual cooling and heat-
ing load in perimeter of buildings per unit floor area [9,18] was
established based on case studies of hundreds of buildings with
different climatic contexts in tropical and subtropical monsoon
regions [23]. A low ENVLOAD value indicates low building envelope
energy demand and high energy conservation [24] so to be a design
index for green buildings [21,25,26]. However, the ENVLOAD is cur-
rently calculated by architects manual determination based on the
architectural blueprint that limits its applicability. In addition,
architects often design building envelopes for energy conservation
on the basis of their experience, but such subjective approaches
337
Li heat loss coefficient of building envelope in the ith sec-
tor, W/m2 K
Mki insolation gain coefficient on k orientation of the build-
ing envelope in the ith sector, dimensionless
Ni window number of building in the ith sector
OR calculated window opening rates of the building, %
ORo original design window opening rates of the building, %
Rl requested lower limit of ENVLOAD, kW h/m2 yr
Ru requested upper limit of ENVLOAD, kW h/m2 yr
SLi sunshade board length in the ith sector, m
Ti sectors number of the building envelope, dimensionless
Tu increment in the average room temperature, K
Ul thermal conductivity of the wall, W/m2 K
Um thermal conductivity of the glass curtain, W/m2 K
Un thermal conductivity of the roof, W/m2 K
Vairai air-conditioned zone areas in the ith sector, m2
Vunairai non-air-conditioned zone areas in the in the ith
sector, m2
Wi window width in the ith sector, m
WLi window length in the ith sector, m
Greek letters
gi solar transmittance of glass, dimensionless
Acronyms
OBEM Office Building envelope Energy performance and con-
figuration Model
OPOBEM Optimal Office Building envelope Energy performance
and configuration Model
TRNSYS TRaNient SYstems Simulation program
ENVLOAD annual building Envelope energy Load, kW h/m2 yr
may not yield the optimal design features [27] and may result in
a high construction budget. An optimized building envelope is
required to achieve a high energy performance of the building in
green building design [28]. Studies used optimization approaches
to aid architects in selecting the optimal building envelope features
[29,30]. For example, Fesanghary et al. minimized energy con-
sumption by altering building envelope materials, including insu-
lation type, roofing, and window type, size, and glazing [31]. In
other words, efficient building envelope performance assessment
techniques and energy consumption indices are essential in green
building design.
In building design optimization, architects must sometimes
assign integer or discrete values to building design variables
[32,33]. Heuristic algorithms, such as genetic algorithm and Tabu
search (TS), were demonstrated to be a useful optimizer to solve
continuous and discrete optimization problems [3437]. Ha et al.
used TS to manage the power consumption in a home automation
system by determining the starting time of some services [38].
Meanwhile, TS was also found more effective than other optimizers
for resolving mixed-integer nonlinear programming [39]. Few
researches utilized TS to optimize a building design, and none of
studies reported on optimizing office building envelope features
in Taiwan. In addition, office buildings are considered as the major
growing source of energy consumption in urban area [5]. Hence,
this study is to develop an Optimal Office Building envelope Energy
performance and configuration Model (OPOBEM). First, Office
Building envelope Energy estimation and configuration Model
(OBEM) was built based on the ENVLOAD equation. The OBEM pro-
vides office building ENVLOAD and useful office building envelope
338 Y.-H. Lin et al. / Applied Energy 171 (2016) 336346
configurations for satisfying architects design, including window
number, window length, window width, type of sunshade board,
sunshade board length, window glass material, wall material, roof
material, glass curtain material, and envelope cost. The OPOBEM
integrates TS optimizer with the OBEM for minimizing the envel-
opes construction cost and satisfying simultaneously regulations
requirements of green building. The validity of OPOBEM was
demonstrated in designing a real office building; two design sce-
narios were investigated and cost efficiency analysis of the build-
ing performed at various requested ENVLOAD values was
implemented for illustrating the feasibility of OPOBEM as an archi-
tectural reference.
2. Methodology
2.1. ENVLOAD, building envelope configurations OBEM, and case study
Since temperature difference between indoor and ambient con-
ditions predominates the energy consumption of building in coun-
tries in frigid and temperate zones, heat supply and heat transfer
efficiency in envelope material are mainly concerned in building
energy conservation [9]. In contrary, energy consumption for cool-
ing is greater than for heating in office buildings in subtropic and
tropic zones, such as Taiwan. Therefore, the efficient designs of
office building envelope, including solar shading, thermal insula-
tion, and building tightness, are firstly concerned and followed
by the selection of material. The ENVLOAD takes into account the
boundary conditions, including indoor temperature, outdoor tem-
perature, sunlight, solar radiation, office building orientation, the
internal gains (heating, occupation, lighting), natural ventilation,
climate, envelope material, thermal inertia, and air-conditioner
annual sensible heat load [14]. The climatic differences in Taiwan
must be considered in three groups corresponding to northern,
central, and southern parts as Fig. 4. Therefore, the ENVLOAD as
Eq. (1) involves two meteorological parameters (DH, IHk) and three
architectural design parameters (G, L, Mki) to interpret the energy
Fig. 1. Components of OBEM and connection between the exogenous variables and exogenous variables.
loads behaviors in office buildings. G is the annually cumulative
amount of internal loads within perimeter zones.
To provide architect design reference, the OBEM was developed
based on Eq. (1) to provide useful architectural design information
(endogenous variables), including window number, window
length, window width, type of sunshade board, sunshade board
length, window glass material, wall material, roof material, glass
curtain material, building ENVLOAD value, and envelope cost as
Fig. 1, in which different envelope materials have different thermal
conductance, solar transmittance, and cost. Such endogenous vari-
ables determining the exogenous variables and the connection
between the exogenous variables and exogenous variables are
shown in Fig. 1. For example, the conductivity of wall determined
by the selected wall material affects the heat loss coefficient of
building envelope. Therefore, the ENVLOAD value and building
envelope cost can be calculated through the exogenous variables
and exogenous variables.
Before building an OBEM, an initial architectural blueprint
should be required and implemented by architects. The OBEM
was combined with TS optimizer to form the OPOBEM, and the
detailed flowchart of OBEM and OPOBEM were shown in Fig. 2.
First, the collected data, such as floor area and building location,
were input to the model. The exogenous variables (decision vari-
ables for TS optimizer) for the ENVLOAD, such as envelope material
and window number for each building sector and orientation, were
input, and the endogenous variables, such as window area, solar
transmittance, and thermal conductance of the envelope, were
determined. The sunshade coefficient (Ki) for window was derived
from depth rate, type of sunshade, and building orientation; depth
rates in the three sunshade configurations are depicted in Fig. 3,
and Table S1 in Supplementary Material lists the coefficient of
the sunshade effect, Ki, for this case study. For example, if the depth
rate, type of sunshade, and building orientation is 0.05, horizontal,
north, respectively, Ki is 0.93. Next, Mki and Li are calculated for
air-conditioned and non-air-conditioned zones in building, respec-
tively. DH and IHk are determined and related to the building ori-
entation and location respectively. Finally, the ENVLOAD value for
 Y.-H. Lin et al. / Applied Energy 171 (2016) 336346 339

office building is calculated using Eq. (1). In addition, OR and sha- wall material, glass curtain material, and roof material are shown
dow area are obtained using Eq. (10). The detail mathematical for- in Table S3S5; Total Ci, C 0i , Di, and D0i are, 1513.92, 4841.13,
mulation of OBEM is estimated using Eqs. (1)(14) [40]. 2893.31, and 531.48 m2, respectively; gi and detail candidate kinds
for window glass are shown in Table S2.
X
Ti
ENVLOAD 20370 2:010  G 0:033  Li  DH
i
! 2.2. TS algorithm
X
4 X
Ti
1:079  Mki  IHk 1
k i TS, a heuristic approach proposed by Glover and Languna [42],
G 13:5  Ac 2 needs short and long memory to search solutions, and has been
successfully applied to a number of combinatorial optimization
Ac 1661 118  Tu  3:1  Tu2 3
problems. TS contains five main algorithm steps, including move,
13:5
Tu 4 tabu list, aspiration criteria, intensification, and diversification
L X [43,44]. The flowchart of TS used in this study is presented in
Li 1:011 fU l  Bi  B0i  Ai U m  C i  C 0i Fig. 5. First, initial solutions (decision variables), including window
U n  D  D0 0:5  U l  B0i U m  C 0i U n  D0 g=AFp 5 number, window length, window width, sunshade board length,
Mki Vairai Vunairai =AFp 6 sunshade board width, window glass material, wall material, roof
X X material, and glass curtain material as shown in Fig. 1, were ran-
Vaira K i  gi  Ai 0:035  U l  Bi  B0i  Ai  domly generated from the interval between the lower and upper
 0

0:035  U n  Di  Di 7 boundary. For example, there are three candidates of sunshade
  type in the OBEM (as Fig. 3) that are coded as: 1 (horizontal sun-
Vunaira 0:5  0:035  U l  B0i 0:035  U m  C 0i 0:035
shade), 2 (vertical sunshade), and 3 (grid sunshade). In other
 U n  D0  8
words, the lower and upper boundary for sunshade type is 1 and
Ai Ni  Li  W i 9 3, respectively, and initial solution for sunshade type can be
   
OR Ai A0i = Ai A0i Bi B0i  100 10 selected randomly from 1, 2, or 3. Similarly, the variables of win-
K i Fsunshade type; depthrate; orientation 11 dow glass, wall, roof, and glass curtain materials are coded as 1
to the candidate type of available materials; the candidate number
For horizontal sunshade, of glass, wall, and roof are 58, 23, and 19, respectively; other dis-
dr SLi =W i 12 crete variables are directly encoded as integer numbers represent-
ing the number of unit lengths, and the actual size is the product of
For vertical sunshade, the number and unit length. The initial solution moves to create
several neighborhood candidate solutions and to calculate the
dr SLi =Li 13
objective value (ENVLOAD); the type of move depends on the prop-
For grid sunshade, erty of problems [44], and an add/drop moving process was used in
this study. For example, a random number 3 is generated for sun-
dr SLi =W i SLi =Li =2 14
shade board length in the first iteration and indicates that sun-
Here, the nomenclature and definition of the symbols for each shade board length is 3 unit lengths; the neighborhood candidate
equation are shown in Nomenclature. The coefficients in Eqs. solution of sunshade board length generated by the move is 2
(1)(8) are regression coefficients and have been popularly used and 4 unit lengths, and the corresponding objective value (cost)
for office buildings in Taiwan. In the ENVLOAD equation, Eq. (1) is 20 and 40 NTD, respectively. In this study, the local optimum
has a constant of 20370 in W h/m2 yr and dimensionless coeffi- among the candidate solutions (2 unit lengths) was chosen next,
cients, including 2.010, 0.33, and 1.079. However, different build- and the solution (2 unit lengths) is marked as tabu on the tabu
ing categories, such as hotel and office buildings, have different list (long memory). The tabu list records the past moves to prevent
regression coefficients and have been set for the buildings in Tai- cycling searches and entrapment in local optima [45], and each
wan [41]. Fig. 3 depicts the three sunshade configurations in the local optimum in each iteration is recorded on the elite solution list
OBEM; each building orientation (k) was divided into several sec- (long memory list), whose size is set as 300 iterations in this case
tors (i) according to the initial architectural blueprint, and the study. The short tabu lists and long elite lists (short and long mem-
building envelope elements were estimated in each sector. In addi- ory) demonstrate intensified and diversified searches, respectively
tion, building structures in Taiwan is column structure system to [46]. The tabu list in this study records solution attributes, values
bear the building weight, and the change of faade window area of decision variables, and the tabu list size is set to the number
wouldnt affect the major structural design of the building. The sur- of decision variables. If the local optimum in this move is tabu, it
rounding obstruction possibly shelters the building from sunlight is checked against the aspiration criteria. The aspiration criterion
and affects the ENVLOAD value, but it is not considered in the in the case study is set to accept the local optimum as a new solu-
OBEM. tion when the local optimum is better than the recorded optimum
An office building in Chiayi city, west Taiwan, located at an alti- on the elite solution list. Otherwise, the optimal solution is
tude of 30 m, was chosen for the case study. The photograph of the replaced by the second optimal solution among the candidate solu-
building and elevation drawings in each orientation is shown in tions. If this second optimal solution is not tabu, it is accepted as a
Fig. 4. The building has 13 aboveground floors and two under- new solution, and the tabu list is updated. If a new solution is bet-
ground floors, and horizontal sunshade was designed in this build- ter than the recorded optimum, the recorded optimum is replaced
ing; other data and material configurations of the envelope are by the new solution. Above iteration continues until the maximum
listed in Tables 13. In the study case, Fig. 1 shows that the build- termination criterion is satisfied, which is set to 100,000 iterations
ing has 4 orientations and the building envelope in each orienta- in this study.
tion is divided into 3 sectors (Ti). AFP, Ac, DH, IHk, and total To preserve search aggressiveness and diversity, intensification
building envelope area are shown in Table 1; the candidate kinds and diversification strategies were implemented before termina-
of wall material, glass curtain material, and roof material, and unit tion if the iteration satisfied specific conditions. The intensification
cost (Cw, Cu, and Cr) and thermal conductivity (Ul, Um, and Un) for strategy records a subset of elite solutions with high-quality local
340 Y.-H. Lin et al. / Applied Energy 171 (2016) 336346

Fig. 2. Flowchart of the OPOBEM.

optima obtained from the solutions of visited solution. If an algo-
rithm is unable to obtain a more optimal solution after a specified
number of iterations, the TS selects the most optimal solution
among these elite solutions as the new solution for next iteration.
For example, when the number of iterations approaches 1000 in
this study, the TS selects the most optimal solution among these
elite solutions as the new solution for next iteration. Contrary to
intensification, diversification strategies diversely search unvisited
regions to explore solutions that considerably differ from those
already obtained. If the current recorded optimum remains unim-
proved after a specified number of iterations, diversification is
applied by randomly perturbing the current recorded optimum
to randomly generate a new solution for the next iteration to
explore unvisited solution spaces. For example, if the recorded
optimum remains unimproved in 600 consecutive iterations in this
study, the recorded optimum will be increased or decreased ran-
domly to generate a new solution, through which the TS restarts
a new iteration.
2.3. OPOBEM
As illustrated in Fig. 2, the optimization process in the OPO-
BEM involves collecting the required data related to the build-
ing envelope and determining objectives based on the
 Y.-H. Lin et al. / Applied Energy 171 (2016) 336346 341

Horizontal sunshade Vertical sunshade Grid sunshade

SL
i

SL i

SL i
Wi
Wi

WLi Li
Depth rate =
Depth rate = SLi/Wi Depth rate = SLi/WLi (SLi/Wi+SLi/WLi)/2
Fig. 3. Available sunshade type in the OBEM.

Fig. 4. Layout of study case and building facade in different orientation.

architects requests, followed by adjusting the TS procedures For horizontal sunshade,

and determining the minimum building envelope construction
Ei Ni  W i  SLi 17
cost (ENVCOST) at the required ENVLOAD value. The construction
cost is the sum of the costs of the window area, window glass For vertical sunshade,
material, sunshade board, and wall, roof, and glass curtain Ei Ni  Li  SLi  2 18
materials of the building envelope. Furthermore, the mathemat-
ical formulation of the OPOBEM in a building orientation is For grid sunshade,
expressed in Eqs. (15)(22). Ei Ni  W i  SLi Ni  Li  SLi  2 19
Minimize ENVCOST
     Subject to
ENVCOST R Ai A0i  C g Bi B0i  C w C i C 0i
 Ni  W i 5 FW i 20
 C u Di D0i  C r Ei  C s 15 Rl 6 ENVLOAD 6 Ru 21
Ei Fwindow size; sunshade style; sunshade board length 16 ORo 6 OR 22
342 Y.-H. Lin et al. / Applied Energy 171 (2016) 336346

Table 1

536.4

987.8

532.2
1032.0
Basic data and original design plan for case study building.

Area
(m2)
Orientation (k) Mka IHkb Mk  IHk (W h/m2 yr)
North 0.0316 267,000 8435.01

Opening rate,
East 0.0176 564,000 6789.47
South 0.0219 464,500 10,151.21

Window

RO0 (%)
West 0.0173 392,700 9751.11

21.4

11.6

11.6
20.0
Roof 0.0135 1,039,000 14,076.23
Heat loss of the envelope, Li (W/m2 K) 6.72
Average increment of indoor temperature, Tu 2.01

2.6  1.2

2.8  1.2
2.8  1.2
1.2  2.0
1.8  2.0

2.6  2.0
1.2  2.0
2.0  1.2

2.0  1.2
18.2.0
(K)

Sizeb
Annual cooling air conditioner operation time 1885.45
(h), Ac
Annual indoor heat gain, G (W h/m2 yr) 25,453.51
Exterior area, AFp (m2)

Number
13,139.52
Degree-hours based on monthly temperature 16,100

12

16
averages, DH (K h/yr)

Sunshade board

6
2
5
9
7

1
5
2
Building envelope area (m2) 19,007.52
Total floor area (m2) 55,124.21

36,448,360
Horizontal
Horizontal

Horizontal
Horizontal
Horizontal
Scenario II

Sunshade
ENVLOAD, (kW h/m2 yr) 87.45

3093.44
Vertical
Vertical
Vertical

1629.4
Envelope cost ($NTDc)

87.67
61,856,720

style

Grid
Grid

41
a
Coefficient of solar heat gain at specific orientation of the building,
dimensionless.

234.2

124.3

233.7

124.3
b
Isolation-hours based on monthly weather averages at specific orientation

Area
(m2)
(W h/m2 yr).
c
New Taiwanese Dollar; 1 NTD = 0.033 US dollar.

rate, RO (%)
Window
Opening
Here, the nomenclature and definition of the symbols for each

4.86

2.68

4.75

2.68
equation are shown in Nomenclature. The ENVLOAD value is
derived from Eq. (1). Therefore, the nine variable categories of deci-

1.2  2.4

2.8  1.2
2.8  1.2
2.8  1.2

2.8  1.2
2.8  1.2
2.8  1.2
2.8  1.2
1.2  2.0

1.2  2.0
sion variable are the window number, window length, window Sizeb
width, window glass material, sunshade board length, sunshade
board width, wall material, roof material, and glass curtain mate-
rial, as shown in Fig. 1. In this study, the ranges of window length,
Number

window width, type of sunshade board, sunshade board length,

window glass material, wall material, roof material, and glass cur-
Sunshade board

2
3
1
1
2
2
3
1
1
2
tain material are 17, 19, 13, 116, 158, 123, 119, and 15,

34,745,500
Horizontal
Horizontal
Horizontal

Horizontal
Horizontal
Horizontal
Horizontal
respectively. The detail candidate kinds and unit cost for window
Scenario I

Sunshade

Vertical
Vertical

Vertical

glass material and sunshade board (Cg and Cs) are shown in

716.63
208.08
86.94
style

43.8
Table S2S3. These nine categories of decision variables are
encoded as integer numbers in the TS for estimating the ENVLOAD
and ENVCOST values. Such OPOBEM is referred to as a mixed-
533.2

533.2
1020.6

984.0
Area
(m2)

integer programming problem, but discrete variables generally

render the optimization problem nonconvex, discontinuous and
therefore more difficult to solve [32] and [47]. The OPOBEM
rate, RO0 (%)

focuses on optimally selecting the building envelope material,

Window
Opening

facade window size, and number. This change affects building

21.2

11.5

11.5
20.0

envelope energy performance but not the structural design of the

building. The TS, OBEM, and OPOBEM were coded in FORTRAN
and implemented using Visual Studio 6.0 on a personal computer.
1  1.6
1  1.6
1  1.6
1  1.6
1  1.6
1  1.6
1  1.6
1  1.6
1  1.6
1  1.6
Sizeb

3. Results and discussion

Number

New Taiwanese Dollar; 1 NTD = 0.033 US dollar.

16
22

16
16
22

16
10

10

The original envelope design, derived from OBEM, of the study

Sunshade board

7
Original design

case is detailed in Table 1; different building orientations have dif-

61,856,720
Horizontal
Horizontal
Horizontal
Horizontal
Horizontal
Horizontal
Horizontal
Horizontal
Horizontal
Horizontal

ferent heat gains (Mk  IHk) from the sunshade factors because of
Sunshade
Scenario

3071.07
3075.2
Results for different design scenarios.

differences in the window area and number of incident sunlight.

87.45
style

The estimated ENVLOAD value and construction cost were

-

87.45 kW h/m2 yr and 61,856,720 NTD, respectively. Table 2 shows

a (front)

a (front)

Cost reduction ratio (%)

a (back)

a (back)

OR0 in the northern, eastern, southern, and western orientations

Total widow area (m2)
a and c

a and c
Sectorc

ENVLOAD (W h/m2 yr)

Envelop cost ($NTDa)

being 21.2%, 11.5%, 20.0%, and 11.5%, respectively; high OR usually

Sunshade area (m2)

Height  width.
b

results in high heat gain of envelope that possibly needs more

c

Refer to Fig. 4.

operation time of air-conditioner to reach thermal comfort.

Orientation

Because the estimated ENVLOAD satisfies the energy conservation

regulations for green buildings in Taiwan (<92 kW h/m2 yr), the
North

South

West
Table 2

Item

East

ENVLOAD value is constrained to the original value of

c
b
a

87.45 0.5 kW h/m2 yr to compare the optimal results in the two

Table 3
Building envelope material of original and optimal design for case study.

Envelope category Scenario

Original Design Scenario II
Window glass Material On-Line reflective sandwich glass Single-layer on-Line reflective glass
Thick (m) OLB6 + pvb + 6 Blue reflective glass + 6
Solar transmittance, gi 0.39 0.4
Unit Cost ($NTD/m2) 1634 1307
Wall Material Composition Thick (m) Thermal resistance Composition Thick (m) Thermal resistance
Exterior gas film 1/23 Exterior gas film 1/23
Aluminum plate 0.006 1/210 Ceramic tile 0.01 1/1.3
Air layer 0.086 Cement mortar 0.015 1/1.5

Y.-H. Lin et al. / Applied Energy 171 (2016) 336346

Aluminum plate 0.0016 1/210 Red tile 0.23 1/0.8
Glass wool 0.03 1/0.042 Cement mortar 0.01 1/1.5
Air layer 0.086 Interior gas film 1/9
Gypsum board 0.012 1/0.17
Interior gas film 1/9
Thermal conductivity (W/m2 K) 0.9 2.14
Unit cost ($NTD/m2) 2700 1350
Glass Curtain Material Composition Thick (m) Thermal resistance Composition Thick (m) Thermal resistance
Exterior gas film 1/23 Exterior gas film 1/23
Glass 0.008 1/0.78 Glass 0.008 1/0.78
Air layer 0.086 Air layer 0.086
Fiber calcium silicate 0.025 1/1.2 Fiber corrugated sheet 0.004 1/1.2
Interior gas film 1/9 Rock wool heat insulator 0.032 1/0.042
Fiber corrugated sheet 0.004 1/1.2
Interior gas film 1/9
Thermal conductivity (W/m2 K) 2.4 0.98
Unit cost ($NTD/m2) 2200 1720
Roof Material Composition Thick (m) Thermal resistance Composition Thick (m) Thermal resistance
Exterior gas film 1/23 Exterior gas film 1/23
Granite quartz tile 0.01 1/1.3 Above 2500 psi concrete 0.05 1/1.4
Cement mortar 0.015 1/1.5 Foaming PS or PU board 0.025 1/0.028
Light weight concrete waterproofing membrane 0.08 1/0.8 PU 0.005 1/0.05
Perforated asphalt felt 0.002 1/0.05 Cement mortar 0.015 1/1.5
Reinforced concrete 0.01 1/0.11 Reinforced concrete 0.15 1/1.4
Interior gas film 0.15 1/1.5 Cement mortar 0.015 1/1.5
Exterior gas film 1/7 Interior gas film 1/7
Thermal conductivity (W/m2 K) 1.61 0.74
Unit cost ($NTD/m2) 2500 1860

343
344 Y.-H. Lin et al. / Applied Energy 171 (2016) 336346
Fig. 5. Flowchart of the Tabu search.
scenarios. In addition, even though building envelope with a low
window number (low OR) leads to low heat gain, energy consump-
tion, and construction cost of building envelope, such external out-
look of the envelope could not be popularly accepted by most
architects or decision-makers. Therefore, scenario I ignored the
OR constraint and optimized envelope configurations to approach
directly the objective of envelope cost minimization. Compared
with scenario I, scenario II optimized the envelope configurations
with window opening rate and ENVLOAD value similar to those
in the original design. The optimized envelope configurations
involve construction material, sunshade type, sunshade length,
window number, and window length. To illustrate the feasibility
of OPOBEM used as architectural reference, a cost-effectiveness
analysis was implemented on the optimized design plans in sce-
nario II as shown in Fig. 6.
3.1. Scenario I: optimized design compared with the architects manual
design
Table 2 reports the optimal design result of scenario I; the ENV-
LOAD and total cost of the building envelope is 86.94 kW h/m2 yr
and 34,745,000 NTD, respectively. Neglecting the window opening
rate constraint, this optimal design employs cheaper materials
than those in the original design (OR0), particularly for the wall
material; nevertheless, the original ENVLOAD value is maintained.
The construction cost is reduced by approximately 43.8%, from
61,856,720 to 34,745,500 NTD. Compared with the original design,
the window number, opening rate, and area, are lower, particularly
for windows in the northern and southern orientations. Although
two sunshade types are adopted, the sunshade area is decreased
so to reduce the envelope cost. In addition, windows with different
types of sunshades in different orientations attributes to the differ-
ent insolation (IHk) in each orientation.
3.2. Scenario II: optimal design with a similar window opening rate of
original design (RO0)
To investigate fairly the feasibility of the TS optimizer in solving
the envelope design optimization problem, the TS was conducted
to optimize the design using window opening rate and ENVLOAD
value similar to those in the original design. Fifty-four decision
variables in the OPOBEM were evaluated, and the solution space
comprised approximately 2.38  1050 possible designs of the
building envelope. Table 2 reports the optimal design result for
 Y.-H. Lin et al. / Applied Energy 171 (2016) 336346
Fig. 6. Building envelope costs in different ENVLOAD requests.
Scenario II. A cost reduction of approximately 41% in the estimated
solution compared with the architects manual design demon-
strates that the TS satisfactorily solves the envelope construction
cost optimization problem with the specific demand of envelope
conservation energy. Compared with the original design, an
increase of only 22 m2 in the window area corresponded to a
decrease of approximately 1446 m2 in the sunshade area. A lower
sunshade coefficient (Ki) means higher efficiency of sunshade for
windows; it depends on the building orientation, window number,
and sunshade type and board size. In the northern orientation, the
highest variation in the sunshade coefficient is observed for verti-
cal sunshade type. Therefore, windows in the northern orientation
adopt vertical sunshade to efficiently reduce incident sunlight. By
contrast, grid sunshade is adopted in the southern orientation to
increase sunshade because it yields the highest variation of the
sunshade coefficient in the southern orientation. These results
demonstrate that optimized sunshade assists in reducing the solar
heat gain through windows and the sunshade area efficiently, thus
conserving energy and reducing envelope costs.
In addition, the detailed material configurations in the original
design and in Scenario II are listed in Table 3. Compared with the
original design, the optimal design in Scenario II in envelope
entails a lower unit cost and thermal transmittance except for
the wall material. Nevertheless, this design yields the similar ENV-
LOAD as in the original design and satisfies green building regula-
tions in Taiwan, demonstrating that expensive materials are not
necessary to construct green buildings in Taiwan; such low costs
can encourage people and businesses to construct green buildings
in Taiwan.
3.3. Cost-effectiveness analysis for different ENVLOAD value
To explore the relationship between the construction cost and
the ENVLOAD, the costs at different ENVLOAD values are estimated
through the OPOBEM. Under the same conditions as Scenario II, the
ENVLOAD values are set to 110%, 90%, 80%, and 70% in the optimal
design, that is, the ENVLOAD values are restricted to lower than
96.4, 78.8, 70.1, 61.4, and 52.6 kW h/m2 yr, respectively, as shown
in Fig. 6. A low ENVLOAD value entails a high envelope cost: a
request for high envelope energy conservation requires a high con-
struction budget. The wall cost appears relatively high, but a low
variance rate was maintained for the envelope cost in the opti-
mized design. For ENVLOAD value lower than 61.4, the sunshade
board cost increases drastically to reduce the solar heat gain
through windows. In addition, glass curtain cost maintains a low
variance rate in the envelope cost, and the glass curtain uses
almost the same material in the optimized and high ENVLOAD
345
designs, indicating that the glass curtain in this building envelope
is relatively easy to design. Similarly, roof uses the same material
due to the constant roof area and the objective of least cost,
although 15 candidate materials are available.
In addition, a high window area indicates a high opening rate of
the envelope and usually leads to high solar heat gain through win-
dows, thus requiring a high budget to satisfy the requested ENV-
LOAD. However, comparing the designs in Scenarios I and II,
restricting the window opening rate did not result in an obvious
variance of envelope cost, possibly because the unit cost of glass
and wall materials are similar in this study and the objective is
to obtain the least envelope cost. In other words, the window
opening rate, envelope cost, and ENVLOAD index in building envel-
ope design are conflicting objectives; therefore, the optimal design
of the building envelope is considered a multiobjective optimiza-
tion problem and solved using a multiobjective optimizer.
4. Conclusions
The design of low-energy green buildings has been attracting
increasing attention in both academic and professional fields.
However, energy simulation of building envelopes is time-
intensive and complex. A simplified energy conservation index
for designing a green building, ENVLOAD, has been proposed to
estimate the energy performance of office building envelopes in
Taiwan; a low ENVLOAD value indicates low building envelope
energy demand and high energy conservation. However, this
envelope configuration design is a highly nonlinear and nonconvex
optimization problem. This study develops an optimal office build-
ing envelope design decision support system, namely OPOBEM.
First, the OBEM is built based on the ENVLOAD equation and con-
siders some conditions design including the window open rate,
envelope material, and window sunshades in the building faade.
The design information with high flexibility in the OBEM can
increase the applicability of the ENVLOAD. The TS combined with
the OBEM to search the least cost of building envelopes at the
required ENVLOAD value. A real envelope design of office building
in Chiayi city, Taiwan, was used to illustrate the feasibility of the
OPOBEM. Furthermore, two scenarios were tested through OPO-
BEM and compared with the original design, and cost-
effectiveness analysis clarified the relationship between envelope
cost and the ENVLOAD index.
The OPOBEM successfully provides the optimal envelope mate-
rial, sunshade type, sunshade length, window number, window
length and width for the building envelope in each orientation
for architect reference. The OPOBEM provided an acceptable build-
ing envelope design at 41% cost lower than in the original design
346 Y.-H. Lin et al. / Applied Energy 171 (2016) 336346
proposed by architects. To efficiently reduce the solar heat gain
through windows, sunshade should be optimally used according
to the high variation of the sunshade coefficient, thus achieving
the goal of energy conservation and reducing the envelope costs.
The OBEM and OPOBEM based on the ENVLOAD can be applied to
other countries in tropical and subtropical monsoon zones, if the
local climatic and building data are sufficient to modify the ENV-
LOAD equation. In addition, the optimized design information with
high flexibility in the OPOBEM increase the applicability of the
ENVLOAD equation. The window opening rate, cost, and ENVLOAD
for designing a green building envelope are conflicting objectives,
and the optimal design of the building envelope is considered as
a multiobjective optimization problem. In addition, a modified
ENVLOAD equation can be established by other countries in tropical
and subtropical monsoon zones if local climatic and building data
are sufficient. Building envelope energy conservation, more factors
in green building design, such as air-conditioner, can also be inte-
grated into the OPOBEM in the future. Simultaneously, a multiob-
jective algorithm can be applied to resolve the tradeoff among
opening rate, cost, and energy conservation of the building
envelope.
Acknowledgements
The authors would like to thank the National Science Council of
Taiwan for the partial financial support of this research under Pro-
ject NSC100-2221-E-507-005, NSC-102-2221-E-005-003-MY3.
This work is supported in part by the Ministry of Education, Tai-
wan, under the ATU plan.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.apenergy.2016.
03.018.
References
[1] Cheng C-L. Evaluating water conservation measures for Green Building in
Taiwan. Build Environ 2003;38:36979.
[2] Hwang R-L, Lin T-P, Chen C-P, Kuo N-J. Investigating the adaptive model of
thermal comfort for naturally ventilated school buildings in Taiwan. Int J
Biometeorol 2009;53:189200.
[3] Mndez Echenagucia T, Capozzoli A, Cascone Y, Sassone M. The early design
stage of a building envelope: Multi-objective search through heating, cooling
and lighting energy performance analysis. Appl Energy 2015;154:57791.
[4] Huang Y, Niu J-l, Chung T-m. Study on performance of energy-efficient
retrofitting measures on commercial building external walls in cooling-
dominant cities. Appl Energy 2013;103:97108.
[5] Yu J, Tian L, Xu X, Wang J. Evaluation on energy and thermal performance for
office building envelope in different climate zones of China. Energy Build
2015;86:62639.
[6] Sozer H. Improving energy efficiency through the design of the building
envelope. Build Environ 2010;45:258193.
[7] Huang J, Deringer J. Status of energy efficient building codes in Asia. Hong Kong
SAR: The Asia Business Council; 2007. p. 69.
[8] Devgan S, Jain AK, Bhattacharjee B. Predetermined overall thermal transfer
value coefficients for Composite, HotDry and WarmHumid climates. Energy
Build 2010;42:184161.
[9] Yang KH, Hwang RL. Energy conservation of buildings in Taiwan. Pattern
Recogn 1995;28:148391.
[10] Yu J, Yang C, Tian L, Liao D. Evaluation on energy and thermal performance for
residential envelopes in hot summer and cold winter zone of China. Appl
Energy 2009;86:197085.
[11] Mustafaraj G, Marini D, Costa A, Keane M. Model calibration for building
energy efficiency simulation. Appl Energy 2014;130:7285.
[12] Hachem C, Athienitis A, Fazio P. Energy performance enhancement in
multistory residential buildings. Appl Energy 2014;116:919.
[13] Chantrelle FP, Lahmidi H, Keilholz W, Mankibi ME, Michel P. Development of a
multicriteria tool for optimizing the renovation of buildings. Appl Energy
2011;88:138694.
[14] Lin H-T, Wang J-C. The design index of building energy conservation and CAD
program BEEP in Taiwan. In: Proc Eighth Int IBPSA Conf. p. 114.
[15] dos Santos GH, Mendes N. Analysis of numerical methods and simulation time
step effects on the prediction of building thermal performance. Appl Therm
Eng 2004;24:112942.
[16] Magnier L, Haghighat F. Multiobjective optimization of building design using
TRNSYS simulations, genetic algorithm, and Artificial Neural Network. Build
Environ 2010;45:73946.
[17] Chou S, Wong Y. Predicting energy performance of commercial buildings in
Singapore. ASHRAE Trans (United States) 1986;92.
[18] Yang KH, Hwang RL. The analysis of design strategies on building energy
conservation in Taiwan. Build Environ 1993;28:42938.
[19] Chou S, Chang W. Development of an energy-estimating equation for large
commercial buildings. Int J Energy Res 1993;17:75973.
[20] Chou S, Chang W. Large building cooling load and energy use estimation. Int J
Energy Res 1997;21:16983.
[21] Yang K-H, Hwang R-L. An improved assessment model of variable frequency-
driven direct expansion air-conditioning system in commercial buildings for
Taiwan green building rating system. Build Environ 2007;42:35828.
[22] Huang K, Lin H. Design standard of energy conservation for building HVAC
systema simplified method of chiller cooling capacity estimation base on
building envelope energy conservation index in Taiwan. Building Simulation.
City: IBPSA: Montreal, Canada 2005:43542.
[23] Huang K-T, Lin H-T. Development of simplified estimation method of chiller
energy use for office buildings in Taiwan. ASHRAE Trans 2007;113.
[24] Hwang R-L, Shu S-Y. Building envelope regulations on thermal comfort in glass
facade buildings and energy-saving potential for PMV-based comfort control.
Build Environ 2011;46:82434.
[25] Huang Y, Niu J-l. Energy and visual performance of the silica aerogel glazing
system in commercial buildings of Hong Kong. Constr Build Mater
2015;94:5772.
[26] Yang M-D, Lin M-D, Lin Y-H, Tsai K-T. Multiobjective optimization design of
green building envelope material using a non-dominated sorting genetic
algorithm. Appl Therm Eng 2016. http://dx.doi.org/10.1016/j.
applthermaleng.2016.01.015 [In press].
[27] Bichiou Y, Krarti M. Optimization of envelope and HVAC systems selection for
residential buildings. Energy Build 2011;43:337382.
[28] Mazo J, El Badry AT, Carreras J, Delgado M, Boer D, Zalba B. Uncertainty
propagation and sensitivity analysis of thermo-physical properties of phase
change materials (PCM) in the energy demand calculations of a test cell with
passive latent thermal storage. Appl Therm Eng 2015.
[29] Rapone G, Saro O. Optimisation of curtain wall faades for office buildings by
means of PSO algorithm. Energy Build 2012;45:18996.
[30] Tuhus-Dubrow D, Krarti M. Genetic-algorithm based approach to optimize
building envelope design for residential buildings. Build Environ
2010;45:157481.
[31] Fesanghary M, Asadi S, Geem ZW. Design of low-emission and energy-efficient
residential buildings using a multi-objective optimization algorithm. Build
Environ 2012;49:24550.
[32] Nguyen A-T, Reiter S, Rigo P. A review on simulation-based optimization
methods applied to building performance analysis. Appl Energy
2014;113:104358.
[33] Wetter M, Polak E. A convergent optimization method using pattern search
algorithms with adaptive precision simulation. Build Serv Eng Res Technol
2004;25:32738.
[34] Namuli R, Jaumard B, Awasthi A, Pillay P. Optimisation of biomass waste to
energy conversion systems for rural grid-connected applications. Appl Energy
2013;102:101321.
[35] Yang M-D. A genetic algorithm (GA) based automated classifier for remote
sensing imagery. Can J Remote Sens 2007;33:20313.
[36] Yang M-D, Su T-C. Automation model of sewerage rehabilitation planning.
Water Sci Technol 2006;54:22532.
[37] Lin Y-H, Chen Y-P, Yang M-D, Su T-C. Multiobjective optimal design of
sewerage rehabilitation by using the nondominated sorting Genetic
algorithm-II. Water Resour Manage 2016;30:487503.
[38] Ha LD, Ploix S, Zamai E, Jacomino M. Tabu search for the optimization of
household energy consumption. Information Reuse and Integration, 2006 IEEE
International Conference on: IEEE; 2006. p. 8692.
[39] Tsai J-F, Lin M-H. Global optimization of signomial mixed-integer nonlinear
programming problems with free variables. J Glob Optim 2008;42:3949.
[40] Technical codes for energy conservation design of office buildings.
Construction and Planning Agency. Taiwan: Ministry of the Interior; 2005.
[41] Huang K-T, Lin H-T. Design standard of energy conservation for building
HAVC-a simplified method of chiller capacity estimation based on building
envelope energy conservation index in Taiwan. In: The 9th International
Building Performance Simulation Association Conference, Canada. 45342.
[42] Glover F, Laguna M. Tabu search. Springer; 1999.
[43] Lin M-D, Sung Y-H, Chu C-W, Lin Y-H. Water distribution network
optimization using heuristic algorithms. In: Proc 2007. p. 578094.
[44] Glover F, Laguna M. Tabu Search. Springer; 2013.
[45] Laguna M. A guide to implementing tabu search. Investigacin Operativa
1994;4:525.
[46] Al-Sultan KS, Fedjki CA. A tabu search-based algorithm for the fuzzy clustering
problem. Pattern Recogn 1997;30:202330.
[47] Hemker T, Fowler K, Farthing M, von Stryk O. A mixed-integer simulation-
based optimization approach with surrogate functions in water resources
management. Optim Eng 2008;9:34160.