Content uploaded by Suraparb Keawsawasvong
Author content
All content in this area was uploaded by Suraparb Keawsawasvong on Nov 03, 2023
Content may be subject to copyright.
Eng. Sci., 2023, 26, 953
© Engineered Science Publisher LLC 2023 Eng. Sci., 2023, 26, 953  1
Engineered Science
DOI: https://dx.doi.org/10.30919/es953
Office Building Envelope Design Optimization by Modified
Competitive Search Algorithm for Energy Saving
Thira Jearsiripongkul1,* Mohammad Ali Karbasforoushha,2 Mohammad Khajehzadeh,3 Suraparb Keawsawasvong3 and Chanachai
Thongchom4
Abstract
Building envelopes for green buildings should be designed with low energy consumption and low construction costs. A
modified competitive search algorithm is used for the optimization of the office building design due to the long runtime of
simulation tools such as EnergyPlus and TRNSYS. For the minimization of the cost of construction at the needed energy
conservation, the envelope configuration, such as window numbers, walls, glass curtain walls, etc., is optimized. Comparing
the proposed algorithm with some others, the cost decreased for the optimum design of the building structure at the needed
energy load value. The number of iterations is also reduced by the proposed approach. Moreover, the overall area of the
window is increased, which has resulted in the natural ventilation being more proper. Since the ratio of the glass curtain wall
is increased, it can result that the indoor lighting being better. Per unit area of the envelope, the value of the energy load is
smaller and the total cost is lower for the proposed method in comparison with other algorithms, considering that the
opening rate of the window is much the same. The total cost decreased by 37.1% in comparison with the initial design. It can
be observed that the MCSA is more efficient than the other compared methods in energy saving of the building in this paper.
Keywords: Optimization; Modified competitive search algorithm; Construction cost; Building envelope; Energy conservation.
Received: 25 July 2023; Revised: 27 August 2023; Accepted: 30 August 2023.
Article type: Research article.
1. Introduction
In recent years, buildings have consumed a major portion of
energy around the world.[1,2] In particular, buildings globally
account for higher than 40% of the overall energy use.[3]
Although the achieved energy efficiency advantages, it is
expected to increase energy consumption in buildings due to
population growth.[4,5] Nevertheless, it is revealed that the
building sector has the potential for the highest energy and
economic savings because there are solutions for its
improvement that are economically proper and profitable.[6]
Energy conservation actions are increasingly required to be
performed in existing buildings because of the low rate of
replacement of these buildings (almost 0.07% yearly).[7] A
major amount of energy is consumed in office buildings.[8]
Therefore, it is important for the optimization of the energy
use and cost of these building types. The target of designing
green buildings is to improve indoor environments with low
energy consumption, which has been considerably expanded
recently.[9] Still, there are many thousands of office buildings
with green certifications worldwide.[10] Different parameters
affect the energy performance of building envelopes, such as
orientation, window shading, area of window and glazing, roof
and wall insulations, and weather conditions.[11,12] That is to
say, it is needed to evaluate several combinations of
parameters to design green buildings.[13]
Different research has been carried out in recent decades to
optimize buildings.[14] In Ref. [13], to optimize envelope
parameters and the building shape, an optimization method
based on simulation was used. An enhanced MantaRay
Foraging Optimizer coupled with the RIUSKA simulation tool
1 Department of Mechanical Engineering, Thammasat School of
Engineering, Faculty of Engineering, Thammasat University,
Pathumthani 12121, Thailand.
2 Department of Architecture, Islamic Azad University, Tehranwest
Branch, Tehran, Iran.
3 Research Unit in Sciences and Innovative Technologies for Civil
Engineering Infrastructures, Department of Civil Engineering,
Thammasat School of Engineering, Thammasat University,
Pathumthani, 12120, Thailand.
4 Research Unit in Structural and Foundation Engineering,
Department of Civil Engineering, Thammasat School of
Engineering, Thammasat University, Pathumthani, 12120,
Thailand.
*Email: jthira@engr.tu.ac.th (T. Jearsiripongkul)
Research article
Engineered Science
2  Eng. Sci., 2023, 26, 953 © Engineered Science Publisher LLC 2023
was employed to obtain the optimal values of all related
variables for the minimization of the energy usage in
residential constructions.[15,16] The applied method has
performed completely appropriately compared to the particle
swarm optimization method, approaching too close to the
optimal in less than 50 percent of the simulations.
In Ref. [17], energyefficient buildings were reviewed.
This paper was aimed at studying building optimization,
energy evaluation, and enhancements in energyefficient
buildings. This study includes the effects of various
parameters on the buildings’ energy use and the ways of
minimizing energy usage using various techniques.
In Ref. [18,19], EnergyPlus and TRNSYS have been used
effectively, also for HVAC design, due to a cognitive
estimation mechanism to decrease the number of the
simulation. In Ref. [18], a simulationassisted control
methodology has been used in a highinertia building. A
building simulation model was employed to effectively
optimize, both present and future information about the
outdoor climatic condition and the state of the building,
combining the thermal comfort and the energy consumption
indices. In Ref. [19], the authors presented Parametrized
Cognitive Adaptive Optimization which has been employed
toward the design of both modelbased and modelfree “plug
andplay” building optimization and control systems, with the
lowest human effort necessary to achieve the design.
In Ref. [20], Improved Battle Royal optimizer (IBRO) as the
optimization method and the TRNSYS simulation software
have been combined for building energy optimization to
investigate the influence of the overhangs optimization.
Giving the attainments, a development was observed in the
comfort level. Moreover, the 4.2% of cooling demand has
been reduced for Shanghai.
In Ref. [21], the costeffective energyretrofit measures
were studied. For the minimization of carbon dioxide
emissions and life cycle cost, multicriteria optimization using
a genetic algorithm (GA) has been applied in different types
of buildings for 5 various major heating systems by enhancing
the building systems and envelope. A multiplecriteria
optimization method has been proposed in Ref. [22], to study
the energy model of building envelopes. For minimizing the
original energy consumption, energyrelated global cost, and
discomfort hours, GA is coupled with EnergyPlus. The
building orientation, radiative characteristics of the plasters,
window type, setpoint temperatures, and thermophysical
features of components of the envelope are considered design
variables.
Authors in Ref. [23], presented a hierarchy of three
definition of very low energy buildings, nearly zeroenergy
building, and zeroenergy building, as the sequential energy
codes of building updated goals to 2050. Six scenarios were
provided to investigate the building’s energy use between
2025 and 2050. The results indicate an advancement in
occurrence time and a reduction in the building’s maximum
energy use.
In Ref. [24], the optimization design of lowenergy
buildings was reviewed to provide the results of former studies
and to help new researchers and architects. The performance
energy consumption and cost were the commonest objective
functions. In another review,[25] the optimizers utilized in the
energyeffective geometry and building envelope
configuration were discussed. The application of derivative
based and derivativefree techniques has been studied in this
paper. For multiobjective optimizations, decisionmaking
techniques have been assumed. Finally, the propositions and
limitations for the related future studies were resulted.
To optimize the energy efficiency of the building, different
studies have targeted a couple of energy simulation tools with
an optimization method.[2629] A comprehencive review of
these methods presented by Barber and Krarti.[30] A tailor
made thermal simulation technique was coupled with an
optimization method performed in MATLAB in Ref. [31], to
implement multiple simulations to achieve the building’s
optimum configuration. As the optimization technique, a
genetic algorithm was employed.
In Ref. [32], a couple of EnergyPlus simulation software
and NSGAII optimization method was applied to attain the
optimum result for improving the building energy
performance. The effect of several characteristics related to
the building architecture like window size, orientation, etc. has
been studied. According to the achievements, the yearly
cooling energy use was reduced between 55.8% and 76.4% in
various studied weather conditions. Nevertheless, an increase
of 1 to 4.8% was seen for the annual lighting electricity
demand. As the result, using the obtained optimum design, the
building's annual total energy consumption was decreased
between 23.8% and 42.2%.
In terms of building optimization design, there are some
mature software packages, but these take a long time to run
and require detailed input of building parameters, which
makes it very inconvenient to design a building. Optimizing
the design of buildings is relatively easy with some
optimization algorithms. Moreover, the interactions between
optimizations of energy system design and building envelope
were ignored in these techniques. In our manuscript, we
recognize the extended simulation times associated with
EnergyPlus and TRNSYS. To address this, we've employed a
new metaheuristic algorithm, the Modified Competitive
Search Algorithm (MCSA), for the optimization of office
building designs. This algorithm efficiently balances the goals
of minimizing construction costs while achieving energy
conservation targets. By optimizing parameters like window
numbers, walls, glass curtain walls, and others, we aim to
strike a harmonious balance between energy efficiency, cost
effectiveness, and overall sustainability. Still, the application
of metaheuristic algorithms to optimize building design is rare.
Therefore, the purpose is to present a model of optimization
for the office building envelope to conserve energy by a new
metaheuristic algorithm, called Modified Competitive Search
Algorithm (MCSA). In comparison to other conventional
Engineered Science
Research article
© Engineered Science Publisher LLC 2023 Eng. Sci., 2023, 26, 953  3
optimization algorithms also the studied algorithms from the
literature, a bigger ratio of glass curtain walls, quicker
convergence rate, a bigger overall area of the window, and
lesser cost by determined energy load can be achieved through
the office building envelope optimization using a modified
competitive search algorithm. That is to say, the cost of
construction and the energy load of the building per unit area
of the envelope can be decreased by the proposed method and
consequently, more efficient achievements can be acquired in
comparison with other optimization techniques in energy
saving of the building. The specific objectives of the presented
study have been given in the following:
 Developing a new metaheuristic algorithm, called Modified
Competitive Search Algorithm
 Minimizing the construction cost at the needed energy load
 Building envelope optimization to achieve a green building
design
 Energy conservation in an office building
2. Materials and methods
In the current study, the modified competitive search
algorithm (MCSA) is used to solve the optimization problem
of the envelope structure of office buildings. The material of
the roof, the number of the windows, material of the glass
curtain, the ratio of theglass curtain wall, the window’s length
and width, the material of window glass, the material of the
wall, and the width and length of the sunshade board are
considered as the first decision variables. According to the
results of optimization, the lesser envelope energy cost (
and the needed value of envelope energy load ( can be
achieved simultaneously.[11,12] For this reason, we can assume
that MCSA is an efficient approach to obtain a solution for
these types of problems. Under the assumption of ensuring the
determined , the building envelope optimization is
performed for the minimization of the. Fig. 1 represents
the architectural design of the office building.
The types of glass curtain wall material, the material of
sunshade board, the material of the roof, the material of the
wall, sunshade board, the material of window glass, and the
sunshade board’s length, window’s length and width, and
windows number are the utilized variables in this study. Fig. 2
depicts the three types of sunshade board including grid (T 1),
horizontal (T 2), and vertical (T 3), and is the width of the
window, is the window length, and defines the
sunshade board’s length. In this study, Shenzhen, Sichuan, and
Nanning in China are selected as the case study regions.
The formula of is defined as follows:
(1)
where,, ,, , and are respectively the
window glass area, wall area, glass curtain area, roof area, and
sunshade board area (m2). , , , , and
are respectively the window glass unit cost, wall unit cost,
glass curtain unit cost, roof unit cost, and sunshade board unit
cost (RMB/ m2).
(2)
where, Y_ihg=Y_ca×13.5, here, is the yearly degreehours
defined by the monthly average temperature (kh/y).
defines the building envelope’s heat loss coefficient (W/ m2K).
defines the yearly indoor heat gain (Wh/ m2y).
denotes the coefficient of insolation gain on z building
envelope orientation. is the isolation hours (Wh/ m2y).
specifies the yearly cooling airconditioning hours (h).
(3)
where, , here, refers to the increase in the
mean temperature of the room (K).
(4)
where, , , and define the wall thermal
conductivity, glass curtain thermal conductivity, and roof
thermal conductivity (W/m2 K), respectively. refers to
the overall airconditioning floor areas in the building’s
perimeter zones (13139.52) (m2).
Fig. 1 The architectural design of the office building.
(a) South orientation (b): East orientation
Research article
Engineered Science
4  Eng. Sci., 2023, 26, 953 © Engineered Science Publisher LLC 2023
Fig. 2 Three types of sunshade board for windows including (a): Grid, (b): Horizontal, and (c): Vertical.
(5)
is the depth rate of the sunshade (%). defines the type
of sunshade board. refers to the orientation. Orientations I,
II, III, and IV denote the north orientation (N), south
orientation (S), east orientation (E), and west orientation (W),
respectively.
(6)
where, is the window’s sunshade coefficient.
(7)
here, is the length of the sunshade board (m), defines
the window width (m), and denotes the window length (m).
, , and refer to the grid, horizontal, and vertical sunshade
boards. , , and are types 1, 2, and 3, respectively.
2.1 Modified Competitive Search Algorithm (MCSA)
2.1.1 The Competitive Search Algorithm (CSA)
The major structure and model of the Competitive Search
Algorithm mathematically are defined in this section after the
intellectual basis of this algorithm is introduced. Then, the rule
of this optimization algorithm is studied.
Intellectual basis; there is a difference between CSA with
other algorithms due to that this algorithm is an inspiration for
human social activities while the others are inspired by the
behaviors of animals and physical laws. A similar process is
followed by various competitive programs shown on TV like
America’s Got Talent and Pop Idol, in which a learning course
is taken by participants after being ranked from different
aspects to be used in the later step. Finally, after the evaluation
of the participants, the optimum one is chosen as the process
of optimization.[33,34] In the beginning stage, it is considered
that the program includes various competition scoring
standards which are appearance, singing, dancing, weight, and
height. Based on a comprehensive test, all competitors are
assessed, and then they were ranked based on their scores.
According to the given ranks, there will be two general and
excellent groups that have been trained for the later
competition step by various techniques. In the end, the
champion of the program is chosen after being learned and
evaluated sequentially.
The structure and mathematical model of the algorithm; the
competitions are defined based on the various rules and their
mathematical model has been developed. The rules are
including 1) based on several standards the competitors will
be evaluated and the points of each competitor are determined
subsequently, two general and excellent groups were created
based on the points of competitors; 2) competitors learn based
on their different abilities. After a while, randomly there will
be some changes in the ability of learning. A learning ability
threshold is specified by each group, and the robust learning
ability is the one with a value higher than this. Moreover, the
lesser value is assumed as the normal learning ability; 3) after
each course is completed by competitors, in the excellent
group, the more different range of learning is related to the
powerful learner than the average one. The excellent group
includes a greater range of learning, therefore, the range of
learning for the next group in the ranking is rather lesser; 4)
the learning of the competitors is defined by their capability in
the general group such that the ones with higher learning
ability aim further on their improvement. However, it is more
likely for those with the normal learning ability to be failed by
themselves; 5) if the ability of learning of a participant is
higher than a determined amount, it can be assumed as
reference behavior. According to the capabilities of the
competitors, they learn from the excellent one indicators; 6)
several competitors are removed from the competition for
different causes when each round ends and are substituted by
new ones thus the number of competitors is fixed in each
round. The indicators of major assessment and the capability
of new competitors are defined randomly.
In the simulation of competition, the virtual competitors
are accepted for the contest. The competitor’s number can be
as given below:
(8)
here, different indicators assessed for participants of the
(a) (b) (c)
Engineered Science
Research article
© Engineered Science Publisher LLC 2023 Eng. Sci., 2023, 26, 953  5
competition were denoted by , in other terms, it then
illustrates the problem. The value of fitness of each competitor
can be stated as the following formula:
(9)
here, the participants’ number is specified by , and all row
values refer to the fitness value obtained by each competitor.
After assessing all competitors in this algorithm, the ranking
of their fitness values is determined after each round of the
contest. According to the fitness value, two groups of
competitors are created including general and excellent. In the
excellent group, the maximumranked participants with higher
strong learning abilities because of the upper limitation will
progress lesser than the participants with higher strong
learning capabilities with lesser rankings. In the general group,
competitors can progress further in upper rankings with higher
powerful ability of learning. The updating of index parameters
of the excellent competitors (EC) with powerful ability of
learning and participants with maximum rankings are as
follows:
(10)
Nevertheless, the updating of parameters of each index of the
excellent competitors with the normal ability of learning and
maximum rankings are defined as given below:
(11)
here, and refer to search limit functions of competitors
with powerful ability of learning and general ability of
learning, respectively. t defines the number of present
iterations. specifies the dimensions’ number
that situated in. The amount of evaluation index of the
competitor is defined by , in other words, the place
information in the dimension. and denote the
constants; and respectively refer to the lower and higher
bounds of the function in the dimensional search limit. The
present contestant’s learning capability is defined by;
specifies the amount randomly achieved using the matrix [1,
0, 1] to show the competitors’ learning direction, i.e., if
equals 1, contestants learn in the opposing direction, if
equals 1, contestants learn in the positive direction, and if
equals 0, contestants will not learn during the current round.
denotes the value of the threshold showing the robustness
of learning capability in the excellent group which is related
to the matrix (0, 1).
According to Eqs. (10) and (11), only in and , the
competitors’ location update difference in the excellent group
is considered. Contestants with the normal ability to learn
mostly search in the range (0%%) of the available range
for searching for each dimension. Contestants with a powerful
ability to learn mostly search between () of the
available range for searching. In this regard, the search extent
becomes more inclusive. In the general group, the competitors
can investigate based on rule 4 for each evaluation round, and
each indicator’s update function can be defined as given in the
following:
(12)
here, α defines the random amount between 1 and 1, Q
denotes a random amount between 0 and 2, and F specifies a
negative factor; and refer to the matrice,
nonetheless, the components in the matrix equal 1, and the
components in have been distributed by random with 1 and
1; defines a random factor, while the competitors’ location
has been renewed, and chosen by random from the matrix [0.1,
0.2, 0.3, 0.4, 0.5]; refers to a standard normal distribution
with variance and mean equal 1 and 0, respectively.
Based on rule 5, the reference behavior is found when the
ability of learning becomes higher than a determined amount
for any competitor: the competitor can learn from the optimum
competitor as stated by their ability of learning, which can be
explained by:
(13)
here, the index amount in dimension of the optimum
competitor during iteration is specified by ;
defines the reference threshold in the range (0,1);
is the difference between the existing
competitor and the optimum competitor. The existing
competitor can go nearer to the optimum competitor by
multiplying
times the ability to learn .
Using Eq. (10) to (13), the indicators of evaluation of
competitors have been learned and renewed. Simultaneously,
based on rule 6 several competitors will always exist that
cannot continue to the later competition because of different
causes after each competition round. Then, a corresponding
amount of competitors are included randomly to have a fixed
number of competitors, and all indicators of evaluation and
abilities of learning are created randomly. According to the
abovementioned model, as shown in Fig. 3, the flowchart is
used for summarizing the original CSA procedure.
2.1.2 Modified Competitive Search Algorithm (MCSA)
The basic CSA is a new effective metaheuristic to solve the
problems of the optimization, but it might suffer some
problems such as the wrong random substitution of the worst
individual, or premature convergences that are provided by the
absence of appropriate exploitation. Consequently, some
modifications are presented herein to enhance the algorithm
efficiency.[35] The modifications are including oppositebased
learning (OBL) and sinecosine procedure as the chaotic
theory to achieve better efficiency.
To find superior candidate solutions, the OBL evaluates
opposite of possible solutions.[36]
Research article
Engineered Science
6  Eng. Sci., 2023, 26, 953 © Engineered Science Publisher LLC 2023
Fig. 3 The flowchart of the original CSA.
In the employed OBL mechanism, adjusting the Jump Rate
(JR) controls the likelihood of an opposing population.
Following each population update, a stochastic process is
employed whereby a random number is generated and
subsequently compared to the jump rate, denoted as JR. The
following formula generates the opposite population of the
present population if the random number is less than JR:
(14)
The current population and opposing population get combined
and their fitness is evaluated individually. The n solutions with
the greatest fitness are then chosen as the new current
population.[36]
In the sinecosine procedure, the individuals that define the
iterations’ worst cost, are relatively chosen to be updated and
the new location is obtained as given below:
(15)
where, , , and denote the coefficients that have
been achieved by the following formulas:
(16)
(17)
(18)
(19)
here, refers to a constant and and
respectively define the current and the highest iterations.
Algorithm1 presents the pseudocode and detailed process of
MCSA.
Algorithm 1. MCSA pseudocode
Initializing n competitors' indicators and determining their
parameters.
A= rand (1, n)
while (t < )
Evaluate competitors' fitness values and rank
for i=1: EC
End
Initializing dimension , population
size, and maximum generation
Description : L1, L3,EC,RC
Generating ability of learning (A)of
each competitor by random
Initializing the basic metrics for each
competitor
Compute the fitness of each
competitor
Ranking the fitness of all competitors Check if i<=EC 1
L=
>
)i(Check if A
1
L<)i(Check if A
Update Y (i) by Eq . (10 )
Update Y (i) by Eq. (11 )
Check if i < population size
Update Y (i) by Eq. (12)
Update Y (i) by Eq. ( 13)
3
L>)i(Check if A
Remove RC competitors and
generate RC competitors by
random
Update Y (i) by Eq. ( 13)
Check if t<maximum
generation
Start
Engineered Science
Research article
© Engineered Science Publisher LLC 2023 Eng. Sci., 2023, 26, 953  7
Use Eqs. (10) and (11) for updating the competitor's
indicators
end for
for i=1+EC: n
Equation (12) is applied to updating the competitors'
indicators
end for
for i= 1: n
Equation (13) is applied to update the competitor's
indicators
end for
Eliminate and randomly generate RC competitors
Updating competitor indicators
Generate 40% of the new population by the sinecosine
procedure based on Eq. (15)
Perform the OBL mechanism based on Eq. (14)
t=1+t
end while
2.1.3 Algorithm validation
After modeling the suggested MCSA, the efficiency of the
technique is better to be evaluated. To analyze the operation
of the suggested technique, four standard test functions are
applied for validation.[37,38] These functions are Rosenbrock,
Rastrigin, Sphere, and Ackley. Then, the results were put in
comparison with some newest algorithms including Ant Lion
Optimizer (ALO),[39] Whale Optimization Algorithm
(WOA),[40] and World Cup Optimization Algorithm
(WCOA).[41] Table 1 states the parameter values of the
investigated algorithms.
The optimization algorithms are coded in MATLAB
R2016b environment on a laptop with Intel CoreTM i52410M,
2.30 GHz CPU, and 8 GB RAM. Table 2 defines the applied
test functions.
The size of the population and the iterations’ highest
number for all optimizers are respectively 50 and 200. The
algorithms have been independently run 40 times to obtain a
proper comparison using the solutions’ standard deviation (SD)
results. To evaluate the effectiveness of the compared
algorithms, their SD and mean values are studied. Table 3
illustrates the results of the comparison of the proposed MCSA
and the optimizers.
According to the results obtained in Table 3, in comparison
to other algorithms, the proposed MCSA with the lower
amount of the mean value gives the maximum accuracy. This
better accuracy indicates higher validation of the presented
technique with appropriate values. Moreover, the lowest value
of the SD shows better reliability of the suggested method than
the comparative optimizers. The initial amounts of the
decision variables in this paper have been created randomly
from their range. If MCSA cannot achieve a more optimum
solution after several iterations, as the updated solution, the
most optimum solution is chosen for a later iteration. The
design parameters have been continuously updated until
satisfactory results have been achieved.
Table 1. The parameter value of the investigated optimizers.
Algorithm
Parameter
Value
ALO[39]
w
[2, 6]
No. search agents
50
WOA[40]
2
1
WCOA[41]
Play off
0.04
ac
0.3
2.2 Problems of optimization
The northern, southern, eastern, and western walls are
optimized wholly in this section without the optimization of
the walls in each orientation. Table 4 reports the main data of
building and original decision variables.
The range of length and width of the window is set at 1~3
m and the range of the sunshade board length is 1~2 m. The
Table 2. The definition of the applied benchmark functions.
Name
Function
Dim
Range
Rosenbrock
300
[30,30]
0
Rastrigin
300
[
5.12,5.12]
0
Sphere
300
[
100,100]
0
Ackley
300
[32,32]
0
Research article
Engineered Science
8  Eng. Sci., 2023, 26, 953 © Engineered Science Publisher LLC 2023
Table 3. The results of the comparison of the proposed MCSA and the algorithms.
Algorithm
Indicator
Rosenbrock
Rastrigin
Sphere
Ackley
ALO[39]
Mean
54.717
233.012
563.107
74.123
SD
24.911
95.009
246.204
57.413
WOA[40]
Mean
16.520
144.675
436. 198
23.178
SD
8.145
82.811
202.206
12.235
WCOA[41]
Mean
4.315
73.121
364.473
5.652
SD
3.003
52.031
186.334
2.062
CSA[42]
Mean
2.089
7.832
155.663
3.779
SD
1.971
11.240
112.905
1.543
MCSA
Mean
0.814e2
1.3512e5
1.09e7
2.095e6
SD
0.119e2
0.025e5
0.968e7
1.049e6
Table 4. The main data of building and initial decision variables.
Orientation (z)
Overall floor area ()
55130.11
Area of the building envelope ()
19010.02
N
8437.20
S
10172.55
E
9926.40
W
6793.70
Roof
14026.50
(h)
1879.72
(W/K)
6.65
(Wh/y)
25454.03
(kh/y)
16098
(K)
1.99
()
13140.08
(RMB)
14219399
(kWh/y)
48.95
materials of the glass curtain, wall, roof, and window glass
ranged from 1 to 5, 1 to 23, 1 to 19, and 1 to 58, respectively.
The numbers define the reference number of the materials. Fig.
4 shows the flowchart of the process of optimization.
Increasing the overall window area can lead to increased heat
losses, particularly during colder periods. This consideration
is indeed crucial when designing energyefficient buildings, as
heat loss through windows can have a significant impact on
the overall energy consumption and thermal comfort of the
occupants. In this study, the delicate balance between
promoting natural ventilation and minimizing heat losses is
recognized. The optimization process seeks to find an optimal
tradeoff between these competing factors. It's important to
note that this optimal tradeoff might vary depending on
factors such as local climate, building orientation, insulation
levels, and occupant behavior. The proposed methodology
takes these variables into account to ensure that the increased
window area contributes positively to natural ventilation while
mitigating potential heat loss drawbacks. In this study, the
need to strike a balance between maximizing indoor lighting
and minimizing any negative impacts on energy performance
and thermal comfort is identified. Our optimization process
takes into account factors such as local climate conditions,
building orientation, and the use of shading devices to mitigate
potential downsides of increased glass area. It's important to
emphasize that a holistic approach to design considers various
aspects of building performance, and decisions are often
influenced by a variety of practical constraints. While an ideal
scenario might involve extensive use of glass, realworld
considerations such as construction costs, energy efficiency,
and occupant comfort play a decisive role in shaping the final
design. In this paper, the interplay between building envelope
design and HVAC systems has been acknowledge. The
optimization process, which primarily focuses on building
envelope parameters, assumes a certain baseline HVAC
operation for the purpose of comparison and analysis.
Fig. 4 The detailed optimization process.
Main data of the
building
Input main data of the
building including
location and floor area
Input original design
variables
Related parameters:
gc
A,
wa
A
wg
A
gc
TH,
r
TH
w
TH
η,
sc
W,
s
Dr
hl
and C
ingz
Calculate C
and
C
Calculate env
and opening rate ,
L
env of window
Modified competitive
search algorithm
Is termination
criteria reached?
Possible design
scenario
Input onjective
C
function env
minimization
Yes NO
Engineered Science
Research article
© Engineered Science Publisher LLC 2023 Eng. Sci., 2023, 26, 953  9
3. Design of optimization process by MCSA
The optimization design process in this section is the trialand
error process. As shown in Fig. 5, the main phases have been
described in the following:
1) Initialize the algorithm parameters: in this process, the cost
function is the function that fits the needed value of the
function.
2) Set iteration=1.
3) The calculation of the population (Pop) is carried out once.
As the nature of evaluating the Pop, the probability amplitude
matrix is transformed into the binary matrix.
4) The cost function at the needed value of is
measured, and the best solution (BS) is achieved in the present
Pop.
5) Carry out a comparison of BS with the optimum best
solution of all former Pops. If BS is better than the
conventional optimum solution (OS), BS and its matching
individual substitute the OS and its matching individual
as the updated OS and . If not, the OS and stay the
same.
6) Iteration = iteration+1, repeat 16. When
iteration>maximum iteration, the termination criteria are
reached.
7) Output the OS and .
3.1 Results of MCSA
The optimization experiment is made to the architectural
design depicted in Fig. 1. MCSA optimization process curve
in comparison to CSA, GA, PSO, and NSGAII is depicted in
Fig. 6.
It is observed that the best optimum results are obtained
after 200 iterations, i.e., the minimum is achieved at this
point, which is equal to 10375281.5 RMB at the needed value
of equal to 45.7692 kWh/y. Table 5 reports the best
optimum variables’ values.
4. Comparative assessment
4.1 Comparative results of the MCSA and other
optimization algorithms
A comparison and analysis of the results of the optimization
of MCSA and several optimization algorithms are carried out,
which is reported in Table 6.
Based on the performed comparison, it is observed that
when the MCSA is applied to solve the optimization problem,
the is lower in comparison to other optimization
algorithms by choosing the proper type of material and fairly
allocating the occupied area by the glass curtain wall, walls,
and windows at the requirement of . The convergence
speed is also faster. The larger overall area of the window of
MCSA indicates that the natural ventilation is more proper. It
Table 5. The best optimum variables’ values by MCSA.
Variable
value
2
The overall area of the window (()
429.9
Number of windows ()
105
The ratio of the glass curtain wall ()
0.31
5
3
3
1
3
2.65
1.53
1.10
(RMB)
10375281.5
(kWh/y)
45.7692
Fig. 5 Design of optimization process by MCSA.
Problem of optimization:
is minimized at the
C
env
L
needed value of env
are
i
The OS and OS
achieved
Initializing Pop, maximum iteration, OS Set iteration=1
The Pop is calculated once
C
Compute the cost function values env
L
at the needed value of env
Find BS in the present Pop and
i
BS
Is the BS better than the OS?
Update the Pop by
MCSA
;
i
Renew the OS and OS
OS=BS;
i
BS=
i
OS
Is iteration>maximum iteration?
Yes
NO
Iteration=iteration+1
Yes
NO
Research article
Engineered Science
10  Eng. Sci., 2023, 26, 953 © Engineered Science Publisher LLC 2023
Fig. 6 The MCSA optimization process curve in comparison to CSA, GA, PSO, and NSGAII.
Table 6. The optimization results of MCSA and other
optimization algorithms.
Variab
le
PSO
GA
NSGAII
CSA
MCSA
415.9
375.3
410.7
427.5
429.9
121
150
108
110
105
0.30
0.29
0.28
0.28
0.31
14
11
12
7
5
4
21
15
6
3
6
10
13
4
3
31
26
17
5
1
5
1
4
2
3
2.15
1.58
1.90
2.30
2.65
1.56
1.58
1.55
1.54
1.53
1.20
1.05
1.25
1.15
1.10
11593144
.1
10877211
.4
10611275
.6
104928
35
1037528
1.5
2
3
3
2
2
Iter*
134
137
80
77
48
(*Iter: iteration number for convergence)
can be the result that the indoor lighting is more appropriate
due to the increased value of the by MCSA. Moreover,
the decreased value of the iteration number for convergence
shows a higher rate of convergence. The overall cost
decreased by 37.1% in comparison with the initial design.
To compare, the energy load and cost of the building are
normalized for all algorithms to achieve and per
unit area of envelope concerning all comparative methods.
Table 7 reports the results of the comparison.
As can be observed from Table 7, is smaller and
is lower per unit area of the envelope for the proposed
method in comparison with other algorithms, considering that
the opening rate of the window is much the same. It can be the
result that the MCSA is more efficient than the other compared
methods in energy saving of the building herein.
Table 7. The comparison results of the presented MCSA with
some other methods.
Variable
PSO
GA
NSGA
II
CSA
MCSA
per unit
area
0.0014
0.0018
0.0015
0.0010
0.0005
per unit
area
268.54
155.82
217.11
165.275
139.9218
The opening
rate of the
window

15.91%
10.07%

10.11%
5. Conclusions
A green or sustainable building is one that is resourceefficient,
environmentally responsible, healthier with lower pollution,
and has applicable space for occupants during its lifecycle of
a building. Saving energy and decreasing costs are significant
aspects of designing a green building. For the optimization of
the building design, an optimization algorithm has been used
in this paper due to the fact that simulation software such as
EnergyPlus and TRNSYS require detailed input of parameters
related to the building, and running them is a timeconsuming
process. Herein, a new metaheuristic optimizer called the
Modified Competitive Search Algorithm (MCSA) was used as
the optimum design technique for the office building envelope.
To lessen the cost of construction at the needed energy
conservation, window numbers, walls, glass curtain walls, etc.
were optimized. A comparison of the proposed algorithm with
some others from the literature was carried out. The cost is
reduced by MCSA for the optimum design of the building
structure at the needed energy load value. The number of
iterations was decreased based on the proposed method.
Moreover, the overall area of the window was increased,
which resulted in better natural ventilation. Since the ratio of
the glass curtain wall was increased, it could be concluded that
the indoor lighting was better. Per unit area of the envelope,
the value of the energy load was smaller and the total cost was
lower for the proposed method in comparison with other
Engineered Science
Research article
© Engineered Science Publisher LLC 2023 Eng. Sci., 2023, 26, 953  11
algorithms, considering that the opening rate of the window is
much the same. According to the achieved results, the total
cost decreased by 37.1% in comparison with the initial design.
It can be the result that the MCSA is more efficient than the
other compared methods in energy saving of the building
herein. Although this study is defined by the results of a
particular design, it can be used in other building designs. For
future works, the presented approach can be applied to other
building designs, various building types, and also other
weather conditions.
Acknowledgement
This study was supported by Thammasat Postdoctoral
Fellowship. Also, this work was supported by the Thailand
Science Research and Innovation Fundamental Fund fscal
year 2023.
Conflict of Interest
There is no conflict of interest.
Supporting Information
Not applicable.
References
[1] G. Aghajani, N. Ghadimi, Multiobjective energy
management in a microgrid, Energy Reports, 2018, 4, 218225,
doi: 10.1016/j.egyr.2017.10.002.
[2] X. Shi, Z. Tian, W. Chen, B. Si, X. Jin, A review on building
energy efficient design optimization Rom the perspective of
architects, Renewable and Sustainable Energy Reviews, 2016, 65,
872884, doi: 10.1016/j.rser.2016.07.050.
[3] Y. Himeur, K. Ghanem, A. Alsalemi, F. Bensaali, A. Amira,
Artificial intelligencebased anomaly detection of energy
consumption in buildings: a review, current trends and new
perspectives, Applied Energy, 2021, 287, 116601, doi:
10.1016/j.apenergy.2021.116601.
[4] Wei, Cai, Optimal bidding and offering strategies of
compressed air energy storage: a hybrid robuststochastic
approach, Renewable Energy, 2019, 143, 18, doi:
10.1016/j.renene.2019.05.008.
[5] Zhi, Yuan, A new technique for optimal estimation of the
circuitbased PEMFCs using developed Sunflower Optimization
Algorithm, Energy Reports, 2020, 6, 662671, doi:
10.1016/j.egyr.2020.03.010.
[6] Bo, Liu, Multiobjective optimization for thermal comfort
enhancement and greenhouse gas emission reduction in
residential buildings applying retrofitting measures by an
Enhanced Water Strider Optimization Algorithm: a case study,
Energy Reports, 2021, 7, 19151929, doi:
10.1016/j.egyr.2021.03.044.
[7] F. Amirifard, S. A. Sharif, F. Nasiri, Application of passive
measures for energy conservation in buildings  a review,
Advances in Building Energy Research, 2019, 13, 282315, doi:
10.1080/17512549.2018.1488617.
[8] Melika, Hamian, A framework to expedite joint energy
reserve payment cost minimization using a customdesigned
method based on Mixed Integer Genetic Algorithm, Engineering
Applications of Artificial Intelligence, 2018, 72, 203212, doi:
10.1016/j.engappai.2018.03.022.
[9] C. Z, Tian, Building energy optimization tools and their
applicability in architectural conceptual design stage, Energy
Procedia, 2015, 78, 25722577, doi:
10.1016/j.egypro.2015.11.288.
[10] Yang, Geng, Indoor environmental quality of green office
buildings in China: largescale and longterm measurement,
Building and Environment, 2019, 150, 266280, doi:
10.1016/j.buildenv.2019.01.014.
[11] Binghui, Si, Ineffectiveness of optimization algorithms in
building energy optimization and possible causes, Renewable
Energy, 2019, 134, 12951306, doi:
10.1016/j.renene.2018.09.057.
[12] Binghui, Si, Performance indices and evaluation of
algorithms in building energy efficient design optimization,
Energy, 2016, 114, 100112, doi: 10.1016/j.energy.2016.07.114.
[13] H. Xian, Li, Energy performance optimisation of building
envelope retrofit through integrated orthogonal arrays with data
envelopment analysis, Renewable Energy, 2020, 149, 14141423,
doi: 10.1016/j.renene.2019.10.143.
[14] H. Zhong, J. Wang, H. Jia, Y. Mu, S. Lv, Vector fieldbased
support vector regression for building energy consumption
prediction, Applied Energy, 2019, 242, 403414, doi:
10.1016/j.apenergy.2019.03.078.
[15] J. Liu, C. Chen, Z. Liu, K. Jermsittiparsert, N. Ghadimi, An
IGDTbased riskinvolved optimal bidding strategy for hydrogen
storagebased intelligent parking lot of electric vehicles, Journal
of Energy Storage, 2020, 27, 101057, doi:
10.1016/j.est.2019.101057.
[16] Dongmin, Yu, RETRACTED: Energy management of wind
PVstoragegrid based large electricity consumer using robust
optimization technique, Journal of Energy Storage, 2020, 27,
101054, doi: 10.1016/j.est.2019.101054.
[17] H. Samanta, R. Maity, S. Laha, P. K. Talapatra, A review on
energyefficient building. Advances in Air Conditioning and
Refrigeration. Singapore: Springer Singapore, 2020: 257268, doi:
10.1007/9789811563607_23.
[18] I. T. Michailidis, S. Baldi, M. F. Pichler, E. B.
Kosmatopoulos, J. R. Santiago, Proactive control for solar energy
exploitation: a German highinertia building case study, Applied
Energy, 2015, 155, 409420, doi:
10.1016/j.apenergy.2015.06.033.
[19] S. Baldi, I. Michailidis, C. Ravanis, E. B. Kosmatopoulos,
Modelbased and modelfree “plugandplay” building energy
efficient control, Applied Energy, 2015, 154, 829841, doi:
10.1016/j.apenergy.2015.05.081.
[20] C. Luo, L. He, S. Huang, B. Huang, C. Yang, L. Du, An
improved bald eagle algorithm based on Tent map and Levy flight
for color satellite image segmentation, Signal, Image and Video
Processing, 2023, 17, 20052013, doi: 10.1007/s11760022
02413x.
Research article
Engineered Science
12  Eng. Sci., 2023, 26, 953 © Engineered Science Publisher LLC 2023
[21] J. Hirvonen, J. Jokisalo, J. Heljo, R. Kosonen, Towards the
EU emission targets of 2050: costeffective emission reduction in
Finnish detached houses, Energies, 2019, 12, 4395, doi:
10.3390/en12224395.
[22] Fabrizio, Ascione, Building envelope design: multi
objective optimization to minimize energy consumption, global
cost and thermal discomfort. Application to different Italian
climatic zones, Energy, 2019, 174, 359374, doi:
10.1016/j.energy.2019.02.182.
[23] X. Yang, S. Zhang, W. Xu, Impact of zero energy buildings
on mediumtolong term building energy consumption in China,
Energy Policy, 2019, 129, 574586, doi:
10.1016/j.enpol.2019.02.025.
[24] S. Longo, F. Montana, E. R. Sanseverino, A review on
optimization and costoptimal methodologies in lowenergy
buildings design and environmental considerations, Sustainable
Cities and Society, 2019, 45, 87104, doi:
10.1016/j.scs.2018.11.027.
[25] F. Kheiri, A review on optimization methods applied in
energyefficient building geometry and envelope design,
Renewable and Sustainable Energy Reviews, 2018, 92, 897920,
doi: 10.1016/j.rser.2018.04.080.
[26] C. Waibel, T. Wortmann, R. Evins, J. Carmeliet, Building
energy optimization: an extensive benchmark of global search
algorithms, Energy and Buildings, 2019, 187, 218240, doi:
10.1016/j.enbuild.2019.01.048.
[27] X. Yuan, M. Ali Karbasforoushha, R. B. Y. Syah, M.
Khajehzadeh, S. Keawsawasvong, M. L. Nehdi, An effective
metaheuristic approach for building energy optimization
problems, Buildings, 2022, 13, 80, doi:
10.3390/buildings13010080.
[28] V. Monetti, E. Davin, E. Fabrizio, P. André, M. Filippi,
Calibration of building energy simulation models based on
optimization: a case study, Energy Procedia, 2015, 78, 29712976,
doi: 10.1016/j.egypro.2015.11.693.
[29] Taehoon, Hong, Automatic calibration model of a building
energy simulation using optimization algorithm, Energy Procedia,
2017, 105, 36983704, doi: 10.1016/j.egypro.2017.03.855.
[30] K. A. Barber, M. Krarti, A review of optimization based tools
for design and control of building energy systems, Renewable and
Sustainable Energy Reviews, 2022, 160, 112359, doi:
10.1016/j.rser.2022.112359.
[31] Sadik, Yigit, A simulationbased optimization method for
designing energy efficient buildings, Energy and Buildings, 2018,
178, 216227, doi: 10.1016/j.enbuild.2018.08.045.
[32] Navid, Delgarm, A novel approach for the simulationbased
optimization of the buildings energy consumption using NSGA
II: case study in Iran, Energy and Buildings, 2016, 127, 552560,
doi: 10.1016/j.enbuild.2016.05.052.
[33] N. Razmjooy, M. Ramezani, N. Ghadimi, Imperialist
competitive algorithmbased optimization of neurofuzzy system
parameters for automatic redeye removal, International Journal
of Fuzzy Systems, 2017, 19, 11441156, doi: 10.1007/s40815
01703052.
[34] A. Noori, M. J. Shahbazadeh, M. Eslami, Designing of wide
area damping controller for stability improvement in a largescale
power system in presence of wind farms and SMES compensator,
International Journal of Electrical Power & Energy Systems,
2020, 119, 105936, doi: 10.1016/j.ijepes.2020.105936.
[35] M. Khajehzadeh, M. R. Taha, M. Eslami, A new hybrid
firefly algorithm for foundation optimization, National Academy
Science Letters, 2013, 36, 279288, doi: 10.1007/s40009013
0129z.
[36] J. Xu, S. Xu, L. Zhang, C. Zhou, Z. Han, A particle swarm
optimization algorithm based on diversitydriven fusion of
opposing phase selection strategies, Complex & Intelligent
Systems, 2023, 133, doi: 10.1007/s40747023010695.
[37] D. Yu, Y. Wang, H. Liu, K. Jermsittiparsert, N. Razmjooy,
System identification of PEM fuel cells using an improved Elman
neural network and a new hybrid optimization algorithm, Energy
Reports, 2019, 5, 13651374, doi: 10.1016/j.egyr.2019.09.039.
[38] M. Khajehzadeh, M. R. Taha, M. Eslami, Oppositionbased
firefly algorithm for earth slope stability evaluation, China Ocean
Engineering, 2014, 28, 713724, doi: 10.1007/s133440140055
y.
[39] M. Mani, O. BozorgHaddad, X. Chu, Ant lion optimizer
(ALO) algorithm. Advanced Optimization by NatureInspired
Algorithms. Singapore: Springer Singapore, 2017: 105116, doi:
10.1007/9789811052217_11.
[40] S. Mirjalili, A. Lewis, The whale optimization algorithm,
Advances in Engineering Software, 2016, 95, 5167, doi:
10.1016/j.advengsoft.2016.01.008.
[41] N. Razmjooy, M. Khalilpour, M. Ramezani, A new meta
heuristic optimization algorithm inspired by FIFA world cup
competitions: theory and its application in PID designing for AVR
system, Journal of Control, Automation and Electrical Systems,
2016, 27, 419440, doi: 10.1007/s4031301602426.
[42] Y. Xu, H. Liu, S. Xie, L. Xi, M. Lu, Competitive search
algorithm: a new method for stochastic optimization, Applied
Intelligence, 2022, 52, 1213112154, doi: 10.1007/s10489021
031334.
Publisher’s Note: Engineered Science Publisher remains
neutral with regard to jurisdictional claims in published maps
and institutional affiliations.