Case Studies in Thermal Engineering 54 (2024) 104028

Contents lists available at ScienceDirect

Case Studies in Thermal Engineering

journal homepage: www.elsevier.com/locate/csite

Modeling of thermal energy saving in commercial buildings of

Australia by balanced tree growth optimizer
Xiuhong Bai a, Yasser Fouad b, *, Chalasani Chiranjeevi c, **, Salem Alkhalaf d,
Barno Sayfutdinovna Abdullaeva e, Fawaz S. Alharbi f, Laith H. Alzubaidi g, h, i,
Zuhair Jastaneyah j, Hakim AL Garalleh k
a Shandong Provincial University Laboratory for Protected Horticulture, Weifang University of Science and Technology, Weifang, 262700, China
b Department of Applied Mechanical Engineering, College of Applied Engineering, Muzahimiyah Branch, King Saud University, P.O. Box 800, Riyadh,
11421, Saudi Arabia
c School of Mechanical Engineering, Vellore Institute of Technology, Vellore, Tamilnadu, 632014, India
d Department of Computer, College of Science and Arts in Ar Rass, Qassim University, Ar Rass, Qassim, Saudi Arabia
e Vice-Rector for Scientific Affairs, Tashkent State Pedagogical University, Tashkent, Uzbekistan
f Department of Mechanical Engineering, College of Engineering, University of Hafr Al Batin, P.O. Box 1803, Hafr Al Batin, 39524, Saudi Arabia
g College of Technical Engineering, The Islamic University, Najaf, Iraq
h College of Technical Engineering, The Islamic University of Al Diwaniyah, Al Diwaniyah, Iraq
i College of Technical Engineering, The Islamic University of Babylon, Babylon, Iraq
j Department of Mechanical Engineering, College of Engineering, University of Business and Technology, Jeddah 21361, Saudi Arabia
k Department of Mathematical Science, College of Engineering, University of Business and Technology, Jeddah 21361, Saudi Arabia

ARTICLE INFO ABSTRACT

Keywords: Low-energy cases can be designed more efficiently by using analytical optimization. The non-
Simulation-based optimization linear thermal performance of buildings has led to the development of optimization techniques
Energy saving measures based on simulation. In building optimization, it is important to achieve superior solutions while
Balanced tree growth optimizer minimizing calculation expenses. This study aims to optimize an Australian office building using
Optimization algorithm benchmarking
the Balanced Tree Growth Optimizer (BTGO). It has resulted that more than 11.7 % of energy can
Commercial building design
be saved by the optimization process and also some energy-saving measures. A comparison of the
utilized algorithm with benchmark algorithms including the Nelder-Mead method, hybrid Parti-
cle Swarm Optimization, Hooke-Jeeves, and Ant Colony Optimization for continuous domain
showed that the BTGO can achieve better solutions and needs less computational time.

Nomenclature: abbreviations and symbols

BTGO Balanced Tree Growth Optimizer

IEA International Energy Agency
NSGA-II Non-dominated Sorting Genetic Algorithm
IDHVAC Daylighting, heating, ventilating, and air conditioning
ANN Artificial neural network

* Corresponding author.
** Corresponding author.
E-mail addresses: yfouad@ksu.edu.sa (Y. Fouad), chiranjeevi.c@vit.ac.in (C. Chiranjeevi).

https://doi.org/10.1016/j.csite.2024.104028
Received 11 December 2023; Received in revised form 10 January 2024; Accepted 12 January 2024
Available online 14 January 2024
2214-157X/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

BEO Building energy optimization

TGO Tree Growth Optimizer
BCA Building Code of Australia
CBD Australian Central Business Districts
VAV Variable Air Volume
PSO-HJ Particle Swarm Optimization and Hooke-Jeeves
NM Nelder-Mead algorithm
ACOR Ant Colony Optimization for continuous domain
HHO Harris Hawks optimization
PIOA Pigeon-inspired optimization algorithm
MFOA Moth-flame optimization algorithm
VT Visible Transmittance
li The lower limit of the parameter i
ui The higher limit of the parameter i
Ec The energy used for cooling the building space
Ef The energy use of the HVAC system's supply and return fans
El The energy use of lighting
Ep The energy use of pumps
Eh The energy used for heating
Em The energy use of indoor equipment and the heat rejection
f (.) The building's yearly total energy use
The global optimum at jth iteration
j
Tg
r A random parameter with a standard uniform distribution between 0 and 1
γ The declining speed of power for trees due to the decreased food nearby, aging, and high growth
di Minimum intervals
λ A random parameter with a standard uniform distribution between 0 and 1
TNe
3
The ith parameter value for the elite tree
The ith parameter amount for the jth novel individual
j
TN
3
δ A random amount between 0 and 1
x Maximum amount
x Minimum amount
( )
The generator function of the procedure
j
f Xi
dm The dimensions of the map

1. Introduction
According to the International Energy Agency (IEA), 30% of the produced energy is consumed in the building sector. A growing
population leads to an increase in energy consumption and greenhouse gas emissions. This problem can be solved by improving the
energy efficiency of buildings [1–3]. In spite of this, householders' affordability will determine the cost of installation or substitution
of technologies with further effective ones. This can have a significant impact on the future CO2 emissions and also energy expendi-
tures [4,5]. Therefore, the employment of energy efficiency actions will be effective in reducing energy consumption and the CO2
emissions. However, the complexity of the interactions of the design and environment-related parameters should be taken into ac-
count. Many multiple-criteria optimization methods have been employed; nevertheless, some were examined to solve the problems of
building design parameters. Hence, the optimizations based on simulation can be an effective design technique to answer multiple
precise needs to design buildings with higher behavior [6,7]. For the optimization of building energy effectiveness, many papers have
focused on coupling an optimization technique with building energy simulation software. In Ref. [8], a tailor-made thermal simula-
tion tool and MATLAB optimization technique have been integrated to carry out numerous simulations to achieve optimum building
arrangement. They have used a genetic algorithm as the optimization method. In this way, the time duration of the optimization is de-
creased considerably, and also adaptability matters are removed. Moreover, applicant-friendliness and the flexibility of the software
are increased by the development of this method on a single platform. In Ref. [9] the solution was found using a numerical approach.
In Ref. [10], a retrofitting plan in relation to thermal performance, energy savings, and fiscal indicators has been studied as the pri-
mary objective. In Ref. [11], the Phase Change Material applicability as a building material's thermal energy storage has been empha-
sized.
In [12], the potential use of latent thermal energy storage in the residential sector to attain nearly zero-energy buildings and how
they would interconnect within smart energy systems was discussed. In Ref. [13], an NSGA-II optimization algorithm with the Ener-
gyPlus simulation tool has been coupled to achieve the best solutions to enhance the energy behavior of the building. The impact of
some architecture-related characteristics such as the orientation of the building and the size of the window are investigated. It has re-
sulted that the yearly cooling energy usage decreases between 55.8% and 76.4 % in different climatic zones that they have studied.

2
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

However, the yearly lighting electrical power demand increments between 1% and 4.8 %. Consequently, by the achieved optimal
arrangement, the yearly overall energy use of the building reduces between 23.8 % −42.2 %. Wonuk Kim and et al. [14] combined a
surrogate model for a daylighting, heating, ventilating, and air conditioning (IDHVAC) system to forecast building energy behavior
using artificial neural network (ANN) methods and artificial light regression technique, with the utilization of a database that has
been made applying the EnergyPlus simulation tool. It has been concluded that the GA-optimized IDHVAC model saves 13.7 % of en-
ergy than the typical techniques. In Ref. [15], the optimization techniques used in the energy-effective building envelope and geome-
try design including derivative-based and derivative-free optimization techniques and their utilization have been reviewed. Also, de-
cision-making methods were considered for multiple-criteria optimizations. Then the constraints and proposals for the associated fu-
ture investigations have resulted. Fig. (1) (b) shows the geographical abstract of the proposed methodology.
As a result of this literature review, it was observed that few publications are using simulation-based optimization to design build-
ings. Thus, the findings of this paper can be useful to architects when designing energy-efficient buildings in Australia. It focuses on
presenting a new building optimization method that improves the benchmark algorithms on 3 main operation matrices, namely com-
putational expense, consistency, and solution quality. A novel building optimization technique is the focus of this paper. This method
is developed by the benchmark optimizers based on the main efficiency metrics including the calculation expense, solution feature,
and stability that controls the mathematical comparison of the novel optimizers of the building energy optimization (BEO) problems.
This helps to assess the convergence behavior (i.e. stability and velocity). Here, the Balanced Tree Growth Optimizer (BTGO) is ap-
plied as the optimization method. BTGO is a new optimization technique that in comparison to other famous ones achieves better re-
sults. To control the distance limitations a technique is combined with the BTGO. Finally, for the optimization of the representative
commercial building of the Australian cities including Gold Coast, Palmerston, Launceston, and Adelaide, this improved TGO opti-
mizer. The optimization tests are applied to assess the efficiency of the BTGO algorithm to the above-mentioned behavior metrics and
for the development of a novel way of designing a less-energy commercial building in Australian cities.
This study is novel in several ways, as highlighted by the information provided:
• Building energy efficiency through simulation-based optimization: The paper provides a detailed overview of the
effectiveness of simulation-based optimization methods for improving building energy efficiency.
• BTGO for building optimization: The introduction discusses the benefits of BTGO in terms of computational expense, quality of
solution, and stability over benchmark algorithms.
• Improved building design reduces energy consumption: Using an enhanced optimization method for building design, the
study aims to reduce energy consumption.
• An analysis of energy-saving measures in Australian cities: Using specific Australian cities as a case study to illustrate
the importance of energy-saving measures in office buildings (Gold Coast, Palmerston, Launceston, and Adelaide) adds a
geographical and practical context to the study.
As a whole, the introduction sets a solid foundation for the entire paper, providing clear reasoning for the study, introducing the
novel optimization technique (BTGO), as well as outlining the specific contributions and novelty of the study. The main contributions
of this study are as follows:
- Optimizing building design based on simulation.
- Solving building optimization problems with a Balanced Tree Growth Optimizer.
- Decreasing energy use by improving building design.
- Applying energy saving measures in office buildings in cities of Australia.

Fig. (1). The design of a ten-story building (a), the optimization framework (b).

3
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

2. Methodology
In this study, for the studied building optimization problems, regarding the higher and lower limits on variables, the possible de-
sign space is clearly described: li ∈ −∞, si , ui ∈ si , +∞ , i = 1, … , N where li is the lower ui the higher limits of the parameter i. The
( ] [ )

main objective function in this study is given below:

(1)
( )
min g (x) = min Ec (x) + Ef (x) + El (x) + Ep (x) + Eh (x) + Em (x)

Where, Ec represents the energy used for cooling the building space (MJ/m2yr), Ef denotes the energy use of the HVAC system's supply
and return fans (MJ/m2yr), El defines the energy use of lighting (MJ/m2yr), Ep describes the energy use of pumps (MJ/m2yr), Eh repre-
sents the energy used for heating the building space (MJ/m2yr), and Em shows the energy use of indoor equipment and the heat rejec-
tion which is the energy use of cooling tower fan (MJ/m2yr). Due to that, a normalization of the decision parameter inputted limita-
tion can be carried out, it can be considered that l = 0 and u = 1 without losing the generalization. The objective function, f (.), de-
fines the building's yearly total energy use that is computed using the EnergyPlus simulation tool.

2.1. Optimization algorithm: the balanced tree growth optimizer

The algorithm concept; The Tree Growth Optimization Algorithm is a bio-inspired optimizer by the competition of trees to ac-
quire light and food. The individuals are separated into 4 steps. The first group is the best group including the trees with the great
potentiality of growth desirable qualifications. This group gets adequate light and focusing on food is the major competition between
them. Because of trees' slow growth, the older trees that are taller and smoother are better. Therefore, the growth speed is reduced
by age growth and the trees focus on getting food in their roots. The next step in the competition is for light, some trees try to go to
intervals near trees with various angles to achieve the light. The other set is the eliminate and substitution group, which includes the
ones, which do not have less growth but are cut and replaced with new ones that are planted. The final group is the regeneration set,
the optimal trees, on account of their desirable growth, they start to create novel plants. To reach the mother tree, some features of
that location are inherited by the offspring. The analytical algorithm modeling is explained concisely in the following.
Phase 1: Initialization: the primary individual of trees was created randomly between the higher and lower bounds in this phase.
Phase 2: Fitness assessment: by assessment based on the assumed function, the fitness value for each tree was obtained.
Phase 3: Achieve the optimized tree: this phase obtains the minimum value for the fitness function and the reverse. Tgj denotes the
global optimum at jth iteration herein.
Phase 4: Allow N1 local better solutions: many local searches are looked for solutions as though the novel solutions achieve better
outcomes, substitute them.

1
(2)
j+1 j j
Ti = r × Ti + × Ti
𝛾

where, r denotes a random parameter with a standard uniform distribution between 0 and 1, that due to light contentment by the
trees, the roots of the trees are skilled to transmit and obtain food that has an impact on the growth of the speed of Ti units and γ ex-
j

plain the decline speed of power for trees by the decreased food nearby, aging, and high growth.
Phase 5: Transfer N2 the number of solutions to interval close to optimum solutions through various α angles. Thus, the interval of
the chosen trees and the rest should be obtained first using the following equation:

 N1 +N2  2

(3)
 
j j
di =  T N − Ti
2
i=1

So,

⎧ j j
⎪ di if TN2 ≠ Ti
di = ⎨ j j (4)
⎪ ∞if TN2 = Ti
⎩

Then, the variables x1 and x2 with minimum intervals (di are employed for achieving a linear combination of the trees. The proce-
)

dure is as follows:

y = 𝜆 × x1 + (1 − 𝜆) × x2 (5)

where λ signifies a random parameter with a standard uniform distribution between 0 and 1. To take this tree into 2 neighbor trees dis-
tance with angles of αi = U(0,1), the below formula is applied:

(6)
j j
TN = TN + 𝛼 i × y
2 2

4
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

Phase 6: Take N3 numbers of inappropriate solutions and substitute random amounts in place of them.
Phase 7: Create N numbers of the new individual, considering N = N1 + N2 + N3.
Phase 8: Create N4 numbers of the new solution and altering them by the mask operator according to the optimum solution, in
random and thus adjoin to the new individual, meaning that,

new individual = new individual + N4 (7)

Phase 9: Arranging the novel individual and choosing N numbers of them as the primary individual of the later iteration which is
implemented by tournament, roulette wheel, or the optimum solution.
Phase 10: if ending criteria are reached, stop the algorithm and show the outcomes. If not, come back to step 3.
Balanced Tree Growth Optimizer (BTGO); The Tree Growth Optimizer (TGO) is a novel and efficient optimization algorithm
that can be applied in various spaces to solve optimization problems. Nevertheless, premature convergence with less certainty
may result if an unbalanced exchange of algorithm exploiting and exploring. To find a solution to this problem, in every itera-
tion, the local searching plan is developed assuming an elite tree. Thus, the elite tree with the optimum expense is achieved in
the search space. Then, Nl number of trees are chosen. Afterward, the rest of the trees try to approach their root to the elite
tree to make novel trees in place of N3 numbers of taken trees; here, assuming 𝓇 as the radius in growth space,

(8)
j
TN = TNe × (2 × 𝛿 × 𝓇 + (1 − 𝓇)) ; i = 1, 2, … , m
3 3

where, TN3 explains the ith parameter value for the elite tree, and TN is the ith parameter amount for the jth novel individual. The novel
e j
3
individual is explained based on a row vector as follows:
[ ]
(9)
j j j j
TN = xi , x2 , … , xm ; s = 1, 2, … , Nl
3

where δ signifies a random amount between 0 and 1. Therefore, the novel achievements should be investigated and improved to en-
sure the optimum amount in the range of the maximum amount ( x ) and minimum amount ( x ) as follows:

x if xis < x
{
(10)
j
TN =
3 x if xis > x

The expense amount for the whole new trees is assessed to achieve the optimum tree with the optimum expense. The application
of the chaos mechanism is the second modification in the algorithm to resolve premature convergence. This procedure is an idea to
define the sensitive systems that are influenced by any alterations. The utilization of this procedure is incrementing regularly in the
optimization processes. This procedure can be applied to refine the algorithm about the convergence rate. The usual description of the
chaos procedure is achieved by the following equation:
( )
(11)
j j
Xi+1 = f Xi , j = 1, 2, … , dm

( )
where, f Xi shows the generator function of the procedure, and dm defines the dimension of the map. A tent map was used to mod-
j

ify the local better solutions herein as follows:

1
(12)
j+1 j j
Ti = p q × Ti + × Ti
𝛾

where,
q
1.43 × Pq , Pi < 0.7
{
Pq+1 = q (13)
3.33 × (1 − Pq ) , Pi ≥ 0.7

2.2. Building definition

An office building with 10 floors is considered as the case study herein. The suggested procedures of the Building Code of Australia
(BCA) [16] for regular larger-size commercial buildings placed in Australian Central Business Districts (CBD) are taken into account.
In this study, the arrangement, variables, and suppositions (for instance indoor load) are considered as particularized in the BCA sug-
gestions [16]. The buildings are made of heavyweight concrete construction with a gross area of 9990 m2. All characteristics of an ac-
tual building such as lighting, HVAC system, indoor loads of occupation, equipment, multi-thermal areas, and additional service
equipment are included here. DesignBuilder simulation tools are used to model Variable Air Volume (VAV) with a water-cooled
chiller (COP = 4.2) and heating and cooling sizing factors are applied to all kinds of simulation using the AHRAE standard values of
1.25 and 1.15, respectively. Tables 1 and 2 and Fig. (1) (a) represent the details of the considered case study. Based on the NABERS
[17], the lighting schedules, HVAC use time, equipment, and occupancy are determined.

5
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

Table 1
The details of the case study.

Floors Wall Windows Roof The ratio of

window to the
wall

Materials 180 mm 200 mm heavy-weight concrete R1.5 6 mm clear glass Metal deck, air gap, 150 mm HW 39 %
concrete, carpet batts, 10 mm plasterboard (absorption (SHGC = 0.818, concrete, roof space, R2.0 batts, 13 mm
2 cm coefficient (AC) = 0.6) VT = 0.88) acoustic tiles (AC = 0.6)
Total U-Value 1.345 0.550 5.81 0.228
(W/ m2K)

Table 2
The details of geometry and suppositions applied in the modeling of the case study.

Variables Values

Gross area (m2) 9990

Floors number 10
Floor-ceiling height (m) 2.8
Floor-floor height (m) 3.5
Geometry (m) 31.61 × 31.61
Equipment load 15 W/ m2
Lighting load 15 W/ m2
Lifts and additional service equipment 1 W/ m2
Set-point for temperature 21–22 °C in winter and 23–24 °C in summer
Set-back for temperature 29 °C
Occupation 0.1 Person/ m2
HVAC system Water-cooled AC, Gas boiler, VAV system, COP = 4.3‫٭‬
Infiltration 0 ACPH during HVAC hours
1 ACPH outside HVAC performing hours
(*No heat recovery and economy cycle).

3. Results
The energy operation of the studied case is presented in this part. The detailed results are illustrated in section 3.1 for four various
climate conditions in four Australian cities. Then, a comparison of the outcomes with the outcomes of literature and accessible data is
accomplished. Also, a comparison of the BTGO-optimized case for all cities with the base case and the benchmark algorithms is car-
ried out in section 3.2.

3.1. Simulation results of the base case

The simulation outcomes for the yearly energy use per unit floor area for the case study and the mean state-level energy intensity
of office buildings for different cities are represented in Fig. (2). Only in Palmerston city, simulation outcomes of yearly energy use are
not approximate to the mean state level. Moreover, the simulation outcomes of the current case study are almost the same of the cur-
rent study. The different results of Palmerston are due to the various occupation performances, high-level cooling set-points, and vari-
ous building constructions.
In Fig. (2), a comparison of the yearly energy use per unit floor area for the case study and the average state-level energy intensity
in most cities shows that simulation outcomes are consistent. As a result of varied occupation patterns, elevated cooling set-points,
and divergent types of buildings, Palmerston City stands out with divergent results. It is noteworthy that Palmerston's energy con-

Fig. (2). The yearly energy use for the case study and the average state energy use for commercial office buildings.

6
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

sumption patterns retain relatively stable simulation outcomes across cities, emphasizing the importance of local factors. As can be
seen in this figure, regional factors have a nuanced impact on the amount of energy consumed in commercial buildings.

3.2. Optimization results

The technique applied for the case study of four different cities here is the optimization based on the simulation. The climate con-
ditions of the cities include a) Gold Coast: a humid subtropical climate (Köppen climate classification Cfa), with mild winters and hot,
humid summers; b) Palmerston: average weather, with warm summer afternoon temperatures of 20–22 °C and 12 °C in winter; c) a
cool, temperate climate with four distinct seasons; and d) Adelaide: a Mediterranean climate (Köppen climate classification Csa) with
cool to mild winters with moderate rainfall and warm to hot, generally dry summers. With consideration of the information repre-
sented in Table 3, the minimization of the building's yearly energy use is the objective function.
To link the DesignBuilder to the BTGO a MATLAB code has been provided for the building optimization. Oppositely, GenOpt has
been utilized as the optimization software, and the optimization with PSO-HJ, NM, and ACOR algorithms is accomplished. For all
cities, the optimization process is run ten times for each technique. Due to that about 3000–4000 building simulations were needed
for each run, and a supercomputing cluster was applied. The needed time for the simulation of each run is about 3–4 days. To obtain
the optimized outcome, the number of simulations should be equal to develop an appropriate comparison of the various algorithms.
PSO ends following the pre-explained number of iterations (3000 simulations) in the hybrid PSO-HJ algorithm. Nevertheless, while
no enhancement is achieved, Hooke-Jeeves stops. Therefore, for each run, the simulations’ number was applied as explained subse-
quently. Firstly, to accomplish this, the hybrid PSO-HJ algorithm has been run and the number of simulations has been computed.
The achieved number has been assumed as the terminating criterion for BTGO, ACOR, and NM. However, it should be considered that
the real number of simulations will be a little different based on the specifications of each algorithm. The value of the free variables is
a significant determinant in optimization algorithm operation. Table 4 represents the variables applied in NM. Table 5 illustrates the
variables utilized in PSO-HJ and ACOR.
Also, the applied variables in BTGO are explained in Table 6. The values for variables applied in TGO are γ = 1.5 and v = 0.5.
The optimization outcomes are given in Table 7. To develop a simple evaluation of outcomes, the normalization of energy use per unit
floor area is considered. The optimum variable sets for each algorithm in all cities are represented in Table 7. For Gold Coast, Launces-
ton, and Adelaide, the optimum solutions have been achieved by BTGO (I) after 3215, 3938, and 3173 building function evaluations,
respectively. For Palmerston, the optimum solution is obtained after the 3310 building simulation by BTGO (II). Oppositely, the worst
solutions for Launceston and Palmerston are achieved by ACOR after 3600 and 3400 simulations, respectively.
Also, the optimum orientations of the building are close to 0° for Palmerston, Launceston, and Adelaide and about 10° degrees
North (clockwise) for the Gold Coast. The optimal wall with the lowest amount of solar absorptivity is achieved, and the optimum
roof with the highest amount of emissivity with the lowest amount of solar absorptivity is obtained for all cities. The best thickness for
wall insulation is equal to 0.01 when it was 0.1 before the optimization process. As described below, the lowest amount of insulation
thickness is chosen. Only during the day, the HVAC system performs and the indoor loads are properly high. The dominant mode of
performance of the HVAC system is cooling, also in the cold season, because of this integration of consumption determinants and the
climate condition of Australia. Thus, incrementing the thickness of the insulation can cause an increase in cooling loads in the winter
season, which higher than balances any decreases in the cooling load in the summer season. For instance, when the best thickness of
insulation increments 1 cm, the yearly cooling loads increment 33.2 (GJ) for Gold Coast, 11.36 (GJ) for Palmerston, 21.3 (GJ) for

Table 3
The optimization parameters and the variable ranges.

Parameters Definition Range

Sa Emissivity of Roof 0–1

Sb Solar absorptivity of roof 0.1–0.9
Sc Insulation of wall (m) 0–0.1
Sd Solar absorptivity of wall 0.2–0.9
Se Height of east window (m) 0.5–1.5
Sf Height of north window (m) 0.5–1.5
Sg Height of south window (m) 0.5–1.5
Sh Height of west window (m) 0.5–1.5
Si Depth of east overhang (m) 0–1.5
Sj Depth of north overhang (m) 0–1.5
Sk Depth of south overhang (m) 0–1.5
Sl Depth of west overhang (m) 0–1.5
Sm Set-point for heating °C 18–23
Sn Set-point for cooling °C 22–26
So Orientations for building (degree) 0–45

Table 4
Variables applied for NM.

NM variables Blocking and restarting check Certainty Step-size determinant Modify terminating criterion

Values 10 99 % 0.1 True

7
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

Table 5
Variables applied for PSO-HJ and ACOR.

Variables PSO-HJ

Particles' number 100

Social acceleration 2.05
Maximum speed gain 0.2
Ultimate inertia weight –
Primary mesh size exponent 0
The number of steps decreases 4
Topology Von Neumann
Cognitive acceleration 2.05
Constriction gain 1
Primary inertia weight –
Mesh size divider 2
Mesh size exponent increase 1

Table 6
Variables applied for BTGO (I) and BTGO (II).

Variables ACOR (I) ACOR (II)

Number of novel solutions applied in each iteration 5 5

q variable 0.0001 0.1
Convergence velocity 0.85 0.85
Archive size 50 50

Table 7
Optimization outcomes, the optimum solution for each algorithm.

Cities Algorithm Fitness function (MJ/ m2) Sa Sb Sc Sd Se Sf Sg Sh Si Sj Sk Sl Sm Sn So

Gold coast NM 644.34 0.67 0.48 0.02 0.30 0.76 0.76 0.85 0.75 0.56 0.95 0.53 0.89 21.57 27.01 19.64
PSO-HJ 642.85 0.91 0.31 0.02 0.30 0.78 0.80 0.92 0.76 1.00 0.61 0.59 0.96 18.01 26.87 4.81
ACOR (1) 642.67 0.91 0.31 0.02 0.30 0.76 0.76 0.96 0.76 0.71 0.62 0.73 0.95 18.45 27.05 2.75
ACOR (2) 642.79 0.91 0.31 0.02 0.30 0.87 0.84 0.95 0.76 0.74 0.66 0.72 1.00 19.07 26.97 10.50
BTGO (1) 642.51 0.90 0.30 0.01 0.30 0.74 0.74 0.94 0.74 1.00 0.64 0.72 1.00 18.44 26.95 10.01
BTGO (2) 642.58 0.90 0.30 0.01 0.30 0.80 0.75 0.94 0.75 1.00 0.70 0.70 1.00 18.65 26.95 11.21
Palmerston NM 780.23 0.68 0.29 0.01 0.30 0.74 0.71 0.87 0.74 1.00 0.88 0.76 0.99 18.80 27.20 16.01
PSO-HJ 780.45 0.85 0.30 0.01 0.30 0.70 0.68 0.76 0.75 1.00 0.78 0.61 0.93 22.03 27.20 25.81
ACOR (1) 779.31 0.89 0.31 0.01 0.30 0.75 1.00 0.91 0.74 1.00 0.80 0.70 1.00 20.42 26.98 13.70
ACOR (2) 779.30 0.89 0.31 0.01 0.30 0.75 0.85 0.91 0.75 1.00 0.80 0.69 0.97 20.55 26.98 2.10
BTGO (1) 779.27 0.91 0.30 0.01 0.30 0.72 0.75 0.90 0.74 1.00 1.00 0.68 1.00 21.85 26.90 1.05
BTGO (2) 779.23 0.91 0.30 0.01 0.30 0.71 0.75 0.89 0.75 1.00 1.00 0.67 1.00 21.25 26.90 0.02
Launceston NM 547.15 0.91 0.38 0.01 0.30 1.01 0.68 1.35 0.87 0.75 0.54 0.59 0.75 18.02 26.00 18.25
PSO-HJ 546.16 0.75 0.47 0.01 0.30 1.10 0.91 1.17 0.94 0.85 0.75 0.44 0.76 18.01 26.00 17.91
ACOR (1) 545.95 0.89 0.31 0.01 0.30 0.96 0.93 1.31 1.04 0.79 0.75 0.53 0.76 18.01 26.00 7.15
ACOR (2) 545.97 0.89 0.31 0.01 0.30 0.96 1.00 1.33 0.95 0.80 0.79 0.55 0.75 18.00 26.00 8.90
BTGO (1) 545.88 0.9 0.30 0.01 0.30 0.75 1.01 1.25 0.91 1.00 0.76 0.24 0.70 18.00 26.00 0.00
BTGO (2) 545.90 0.9 0.30 0.01 0.30 0.88 1.03 1.26 0.92 0.82 0.79 0.26 0.76 18.00 26.00 8.81
Adelaide NM 577.21 0.68 0.61 0.01 0.30 0.75 0.88 0.95 0.78 0.55 0.57 0.58 0.68 18.55 25.98 4.91
PSO-HJ 575.87 0.84 0.36 0.01 0.30 0.85 0.81 1.02 0.81 0.75 0.70 0.44 0.70 18.25 26.00 13.12
ACOR (1) 575.61 0.91 0.31 0.01 0.30 0.79 0.82 1.05 0.74 0.69 0.95 0.30 0.99 18.49 26.00 9.65
ACOR (2) 575.66 0.91 0.31 0.01 0.30 0.74 0.74 1.05 0.80 0.69 1.00 0.29 0.73 18.50 26.00 10.00
BTGO (1) 575.56 0.89 0.30 0.01 0.30 0.74 0.86 1.00 0.85 1.00 0.75 0.38 0.75 18.31 26.00 0.00
BTGO (2) 575.57 0.90 0.30 0.01 0.30 0.84 0.88 1.01 0.78 0.78 0.76 0.37 0.95 18.30 26.00 15.01

Launceston, and 16.1 (GJ) for Adelaide, whereas the heating loads reduce 3.2 (GJ) only for Launceston and Adelaide. Due to the
trade-off among heating, cooling, and lighting loads, there is a dependency on the values of the best overhangs and windows in the di-
rection of the building and city. Also, the obtained outcomes can be applied to calculate the best amounts for the ratio of the window
to the wall. For instance, this value is equal to 37.3 % for the South, 32.8 % for the North, 31.9 % for the West, and 27.8 % for the
East building side in Adelaide. For all cases, the lowest value for heating set-points and the highest value for cooling set-points have
been chosen. This is anticipated if the building energy use is decreased. Notably, the optimization solutions represented in Table 7
might not be efficient for some parameters in building design due to that they have been determined with decimal points that are
valuable concerning solutions feature of optimization algorithms. As represented in Table 7, the optimization results show a small dif-
ference between the cost functions achieved using BTGO (e.g. 642.51 MJ/ m2 (Gold Coast)) and ACOR (e.g. 642.67 MJ/ m2 (Gold
Coast)). Also, based on these results, it is observed that the different groups of variables were achieved using different algorithms,
meaning that the case study fitness function is significantly multiple-modal. Therefore, designers have more alternatives to choose

8
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

from and fewer energy cases. Because of higher calculation expense, fewer optimization runs will be applied in practical optimization
problems. Thus, an algorithm that causes better solutions usually is desirable. A lesser average value with a lesser extension or varia-
tion in outcomes proposes a further dependable algorithm for achieving better solutions in each run. The distribution of optimization
outcomes of yearly energy use (MJ/ m2) for each case, by ten runs, is shown by the box-and-whisker plot. The optimum performance
of all cases is obtained by BTGO (II) and BTGO (1), respectively, based on the comparison of the median values depicted in Fig. (3).
Even though the difference between these values is very small, BTGO (I) gives a bigger variation compared to BTGO (II) which
makes BTGO (II) more reliable. Oppositely, the extension of optimization outcomes in NM in all cases is much higher in comparison to
others. Moreover, this value for NM is much bigger in comparison to the others, only in Palmerston it is different due to the highest
value of PSO-HJ. The Mann–Whitney U test has been utilized to know the statistical importance of diversity in the operation of the al-
gorithms. This test is a non-parametric test that is applied to evaluate whether the distributions of observations achieved between two
different algorithms on dependent parameters systematically vary from one another. Here, a lesser probability that the outcomes have
been achieved in a random is shown by lesser p-values whereas a considerable probability that shows no difference among the algo-
rithm operations is demonstrated by higher p-values. A significant difference between both BTGO algorithms and NM, ACOR also
PSO-HJ is represented in Table 8 for all cases.
Nevertheless, this difference is very small for BTGO (I) and BTGO (2). The speed of convergence is the next effective metric for op-
timization algorithms. The assessment of fitness function needs more time in building optimization problems. Thus, it is important to
keep the number of function assessments at a minimum value. The comparison of the convergence rate of optimization algorithms is
specifically effective if the total operation is much near the objective value. An example of an optimization run for a solution approxi-
mate to the median for the Gold Coast is depicted in Fig. (4).
It is clear that in comparison to other optimization algorithms, converging to the final solutions by BTGO (I) and BTGO (II) is very
quick. NM operation is more proper than PSO-HJ and rapidly converges to a solution in primitive iterations. Nevertheless, the
achieved solution has a significant difference in the optimum solution.
Also in the hybrid PSO-HJ algorithm, after 3100 building simulations, the PSO ended and then the PSO outcomes were refined by
HJ. Figs. (5) and (6) depict the total convergence rate of optimization algorithms after running ten times. The comparison of the con-
vergence rate in the last steps of the optimization process when algorithms converge to a solution much approximated to the optimal
for Gold Coast and Palmerston is shown in Fig. (5). It can be observed that the greatly conflicting outcomes are generated by NM. Both
BTGO (I) and BTGO (II) are between 2 and 4.5 times quicker than NM, PSO-HJ, and ACOR comparing median values.
The comparison of the convergence rate in the primary steps of the optimization process when algorithms converge to a solution
approximate to the final for Launceston and Adelaide is represented in Fig. (6). BTGO (I) and BTGO (II) displayed a little quicker con-
vergence speeds in comparison to NM and quicker operation than PSO-HJ and ACOR. Comparing the median values BTGO (I) is about
8 times quicker than PSO-HJ in Adelaide, and even though NM has a potentially quick convergence speed, this speed is conflicting
and the solutions obtained have considerably more energy use than the BTGO solutions. The building's yearly energy use and break-
down of energy use before and after sensitivity analysis and after optimization are depicted in Fig. (7).

4. Discussions
Across the studied cities, discrepancies are evident between Palmerston and the other studied cities in terms of yearly energy use
per unit floor area, owing to differences in occupancy, cooling set-points, and construction. Simulation-based optimization using the
Balanced Tree Growth Optimizer (BTGO) algorithm provides better solutions, lower variation, and faster convergence than other al-
gorithms (NM, PSO-HJ, and ACOR). According to the study, computational expense is an important consideration in building opti-
mization, and BTGO (II) has superior accuracy, rapid convergence, and consistent results. It is evident that the optimization process
has succeeded, despite the implementation of energy efficiency actions and sensitivity analysis, since the energy consumption de-
creased by 11.71 %–14.15%. Optimization outcomes are presented in tables and figures to demonstrate how variable sets are opti-
mized for each algorithm across different cities. It is noteworthy that BTGO (I) and BTGO (II) perform optimally at Gold Coast,
Launceston, and Adelaide, consuming the least amount of energy per square meter. The discussion also emphasizes the role of vari-
ables such as building orientation, solar absorption, emissivity, and insulation thickness in determining energy consumption. It con-
cludes with an analysis of BTGO's efficiency and reliability in achieving energy-efficient building designs with a comprehensive evalu-
ation of optimization outcomes.
Considering the details provided, potential limitations of the used method can be inferred as follows:
- Building Homogeneity: The study examines a specific 10-floor office building in accordance with the Building Code of
Australia (BCA). In this case, building characteristics should be homogeneous. Depending on the building type and size, the
method may be inapplicable.
- Occupancy Representation Static: A static representation of occupancy can be constructed based on lighting schedules, HVAC
use time, equipment, and occupancy determined by NABERS. Users' behaviors and occupancy patterns may vary in the real world.
- HVAC System Standardization: An HVAC system that follows ASHRAE standard values, such as Variable Air Volume (VAV)
with a water-cooled chiller, does not reflect the variety of HVAC systems found in real buildings. It is unclear whether the method
can be used to optimize different HVAC configurations.
- Design Space Simplification: It is possible that the study oversimplifies the complex interactions between design parameters,
despite describing the design space based on BCA recommendations. Building energy performance may not be fully accounted for
by the representation of the design space.

9
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

Fig. (3). The comparison of the algorithm with Box-Whisker plots.

10
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

Table 8
The outcomes of the Wilcoxon rank-sum test. The p-values are in bold which less than the common 0.05 importance level is.

(p-values) Gold coast Palmerston Launceston Adelaide

BTGO (II) VS NM (0.0002) (0.0002) (0.0002) (0.0002)

BTGO (II) VS PSO-HJ (0.0002) (0.0002) (0.0002) (0.0002)
BTGO (II) VS ACOR (2) (0.0032) (0.0002) (0.0002) (0.0004)
BTGO (II) VS ACOR (1) (0.0001) (0.0002) (0.0002) (0.0003)
BTGO (II) VS BTGO (1) (0.2710) (0.1400) (0.1400) (0.4250)
BTGO (I) VS NM (0.0003) (0.0002) (0.0002) (0.0002)
BTGO (I) VS PSO-HJ (0.0200) (0.0002) (0.0032) (0.0200)
BTGO (I) VS ACOR (2) (0.0003) (0.0002) (0.0002) (0.0004)
BTGO (I) VS ACOR (1) (0.0003) (0.0002) (0.0002) (0.0002)

Fig. (4). The convergence rate for the solution approximates the median on the Gold Coast.

Fig. (5). Building simulation numbers needed to converge to within 0.1 % of the optimum solution for each algorithm.

11
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

Fig. (6). Building simulation numbers needed to converge to within 1 % of the optimum solution for each algorithm.

Fig. (7). Building yearly energy use; (a): before optimization, (b): after optimization.

- Assumption of BCA compliance: In the case that compliance with specific standards is assumed, it may overlook potential
innovations or alternative strategies that are not covered by the guidelines.
- A limited range of alternative designs: The study's focus on one particular building design may prevent it from exploring a
broader range of alternatives. Energy efficiency can be affected by different architectural and technological choices.
- Extent of Generalizability: The study does not explicitly address its generalizability to buildings in different climate zones.
There might be significant differences in energy consumption due to optimized designs in climates that are not considered here.
- Specifications of energy tools: Assumptions and modeling capabilities of the EnergyPlus simulation tool may introduce
limitations to calculating yearly total energy use. Other simulation tools may not be able to use the method.
Further work could address these limitations, incorporate more diverse building types and configurations, explore alternative
HVAC systems, consider a wider range of design parameters, and validate results against real-world data to ensure broader applicabil-
ity of the study.

12
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

4.1. BTGO verification

To verify the algorithm, some standard benchmark functions are applied here. Also, a comparison of the outcomes with some
novel well-explained algorithms for further describing the benefits of the proposed BTGO in comparison to others. These functions in-
clude a) the Sphere function, b) the Ackley function, c) the Rastrigin function, and d) the Rosenbrock function that has been defined
in the following subsections.
a) Sphere function
The definition of this function is as follows:

d
(14)
∑
f1 (x) = xi2
i=1

where, the decision parameters, i.e. xi are between −5.12 and 5.12. The optimum solution for this function is equal to 0.
b) Ackley function
This function has further complexity. The equation is as follows:
  
  
1  d
1
d
 
f3 (x) = −20 exp −0.2 xi2  - exp cos 2𝜋xi + 20 + exp (1) (15)
 D i=1  D i=1
 

where the decision amount is between −32 and 32 with 0 being the optimum solution.
c) Rastrigin function
This function analytically is modeled as follows:

d
(16)
∑ ( 2 ( ) )
f2 (x) = xi − 10 cos 2𝜋xi + 10
i=1

where the decision parameters are between −1.28 and 1.28. Also, the optimum solution for this function is equal to 0.
d) Rosenbrock function
The analytical formula of this function is as follows:

d−1 
(17)
 2  2 
100 xi+1 − xi2 + xi − 1

f4 (x) =
i=1

where the decision amounts are between −10 and 10. For this function, the optimum solution is 0. The size of the whole artificial
landscape is assumed to be 50. The 2-dimensional diagram for the selected fitness functions is represented in Fig. (8).
Some well-explained optimizers were selected for validation including Harris Hawks optimization [18], the Pigeon-inspired opti-
mization algorithm [19], the Moth-flame optimization algorithm [20], and the Basic TGO [21]. Each algorithm separately is run 30
times to achieve a certain outcome. The simulation outcomes of the algorithms used in artificial landscapes are represented in Table
9.
The effectiveness of the suggested BTGO has better outcomes in comparison to the HHO, PIOA, MEOA, and the basic TGO as seen
in Table 9. Based on the maximum, average, and minimum values of the algorithms, it has resulted that the suggested BTGO gives the
best certainty. Equally, compared to other algorithms, the lowest amount of the standard deviation for the BTGO presents higher ro-
bustness for the suggested algorithm. Fig. (9) depicts the convergence diagram of the compared algorithms.
As can be observed from the convergence diagrams, the suggested algorithm gives the quickest convergence with precise solu-
tions, not premature convergence based on the outcomes of the whole artificial landscape. Moreover, in convergence and accuracy,
the basic TGO has the second rank.

4.2. Sensitivity analysis

In this section, the modification of the case study is studied to be compared to optimized outcomes. Therefore, to enhance the en-
ergy effectiveness, four usual enhancements are considered including a) extra (0.4 m) overhangs at the top of the windows; b) replac-
ing the one-paned windows with two-paned one (U = 2.6 W/ m2 K), SHGC equal to 0.421 and VT equal to 0.308; c) applying day-
lighting conduct for each region using a reference point with 320 lx set-point 0.8 (m) above the floor and continued lighting conduct
(lowest electrical power and light produced equal to zero), and d) eliminating temperature set back. It has resulted in a yearly energy
decrease of 32.5 %, 30.2 %, 35.2 %, and 38.1 % for Gold Coast, Palmerston, Launceston, and Adelaide, respectively. Table 10 also
shows that cooling loads after sensitivity analysis decreased by 48.6 %, 39.7 %, 62.8 %, and 61.3 % for Gold Coast, Palmerston,
Launceston, and Adelaide, respectively. The yearly energy use has been decreased by 14.15 %, 12.95 %, 13.19 %, and 11.71%s for
Gold Coast, Palmerston, Launceston, and Adelaide, respectively, after employing simulation-based optimization. The comparison of

13
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

Fig. (8). The selected standard benchmark functions' diagrams.

Table 9
The simulation outcomes of the algorithms used in artificial landscapes.

Algorithm Indicator f1 f2 f3 f4

HHO [12] Minimum 9.8672 0.0019 0.0011 9.1439

Maximum 2.9477e+3 1.6767e+5 1.8762e+2 1.0512e+3
Average 1.2249e+3 1.2347e+4 3.9670e+2 31.1922
STD 2.4444e+4 2.2136e+4 1.9128e+2 10.9342
PIOA [13] Minimum 19.6888 2.9904 0.2018 2.9265
Maximum 401.7232 3.9865 1.9334 34.4768
Average 102.9518 4.6545 0.8876 13.9876
STD 71.6509 0.9954 0.1789 3.6908
MFOA [14] Minimum 4.2716 3.7654e-5 2.3763e-5 0.8927
Maximum 4.6651e+2 0.0029 0.0321 0.9563
Average 94.5021 0.0053 0.0091 1.0036
STD 210.9666 2.1234e-5 0.0182 0.8069
TGO [22] Minimum 1.4456 2.4567e-11 6.6545e-21 3.4987e-15
Maximum 34.8694 2.8765e-10 2.3998e-19 1.8768e-14
Average 4.7742 1.6785e-10 1.5678e-19 0.8790e-14
STD 2.7764 6.3498e-11 0.8876e-19 4.4459e-15
BTGO Minimum 0.1322 4.07658e-22 1.5876e-22 2.6897e-17
Maximum 32.6872 0.2346e-10 1.3106 e−20 1.2891e-14
Averages 2.4610 1.1879e-10 0.2911 e−20 0.0763 e−14
STD 2.4329 1.9882e-12 0.0786 e−20 0.1576 e−15

the energy breakdown for the modified case with optimized building displays that optimization has considerably decreased the cool-
ing loads and fan (fan energy use become lower 53.40 %, 43.31 %, 61.29 %, and 53.18 % for Gold Coast, Palmerston, Launceston,
and Adelaide, respectively). The most energy decrease is equal to 34.68 MJ/ m2 is seen for the optimized case in Palmerston. Signifi-
cantly, cooling loads, which have important influences on the building peak load, have been decreased by 35.9 %, 24.93 %, 52.05 %,

14
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

15
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

Fig. (9). The compared algorithms' convergence diagram.

◀

Table 10
Building yearly energy use after sensitivity analysis.

Gold coast Palmerston Launceston Adelaide

Heating 0.00 0.00 0.00 0.00

Cooling 190 280 65 100
Fans 720 830 600 610
Heat rejection 770 875 615 625
Indoor equipment 630 785 575 580
Indoor lighting 340 435 230 235
Pumps 750 850 610 615

and 39.54 % for Gold Coast, Palmerston, Launceston, and Adelaide, respectively. The highest decrease of a cooling load equal to
75.95 MJ/ m2 is seen for Palmerston and the lowest decrease is equal to 42.85 MJ/ m2 is accomplished by Launceston. Notably, de-
spite the application of daylighting conduct, lighting loads nearly stay constant between the modified case and the optimized out-
come. It should be noted that the decrease of the cooling loads is prioritized by an optimization algorithm, which is not strange in
Australia because the minimization of the lighting and cooling loads are inconsistent targets. Due to that, the optimization searches
the optimum equivalence among different building loads, it is more probable that an effort to reduce the cooling or lighting load will
cause an equivalent increment of the same or higher value in the other.

5. Conclusions
In conclusion, the application of simulation-based optimization techniques to improve the energy efficiency of a designed com-
mercial building in various Australian cities was investigated, with a particular focus on the effectiveness of the Balanced Tree
Growth Optimizer (BTGO) algorithm. The conclusion of the study was outlined through key points as follows:
• Application of Simulation-Based Optimization: The paper focuses on optimizing a designed commercial building in four
Australian cities using simulation-based optimization techniques.
• Introduction of Balanced TGO Algorithm: The study introduces the Improved Tree Growth Optimizer (BTGO) algorithm for
solving building optimization problems, optimizing 15 parameters.
• Comparison with Existing Algorithms: BTGO is compared to NM, PSO-HJ, and ACOR algorithms, demonstrating superior
performance in terms of obtaining better solutions.
• The superiority of BTGO in Outcomes' Spread and Convergence Speed: BTGO, especially BTGO (II), exhibits lower variation
and higher consistency in outcomes, with faster convergence speed compared to other algorithms.
• Recommendation for BTGO (II): The study suggests using BTGO (II) for building optimization due to its more accurate
solutions, faster convergence speed, and greater outcome consistency.
• Energy Efficiency Achievements: The optimization process, even after implementing energy efficiency actions and sensitivity
analysis, results in energy decreases ranging from 11.71 % to 14.15 %.
• Identification of Energy-Saving Solutions: The simulation-based optimization identifies potential energy-saving solutions and
determines optimum values for design parameters.
• Implications for Building Codes: The findings can be valuable for future designers in arranging higher-performance building
codes, contributing to sustainable and energy-efficient building practices.
• Consideration of Future Works: Future works are suggested to consider the uncertainty of variables in real building problems
and explore multi-objective optimization problems beyond the scope of this study.
• Overall Contribution and Recommendations: The paper concludes by highlighting the significance of the BTGO algorithm
in addressing building optimization challenges, achieving energy reductions, and providing valuable insights for future research
and design practices.

CRediT authorship contribution statement

Xiuhong Bai: Supervision, Methodology. Yasser Fouad: Project administration, Funding acquisition. Chalasani Chiranjeevi:
Formal analysis, Resources. Salem Alkhalaf: Data curation, Writing – original draft. Barno Sayfutdinovna Abdullaeva: Resources,
Investigation. Fawaz S. Alharbi: Investigation, Writing – original draft. Laith H. Alzubaidi: Investigation, Conceptualization.
Zuhair Jastaneyah: Formal analysis, Writing – review & editing. Hakim AL. Garalleh: Validation, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

16
X. Bai et al. Case Studies in Thermal Engineering 54 (2024) 104028

Data availability
The authors do not have permission to share data.

Acknowledgement
The authors extend their appreciation to Deputyship for research and innovation, “Ministry of Education” in Saudi Arabia for
funding this research (IFKSUOR3-273-6).

References
[1] A. Kamal, S.G. Al-Ghamdi, M. Koç, Role of energy efficiency policies on energy consumption and CO2 emissions for building stock in Qatar, J. Clean. Prod. 235
(2019) 1409–1424.
[2] Z. Xiao, H. Li, H. Jiang, Y. Li, M. Alazab, Y. Zhu, S. Dustdar, Predicting urban region heat via learning arrive-stay-leave behaviors of private cars, IEEE Trans.
Intell. Transport. Syst. 24 (10) (2023) 10843–10856.
[3] H. Jiang, Z. Xiao, Z. Li, J. Xu, F. Zeng, An energy-efficient framework for internet of things underlaying heterogeneous small cell networks, IEEE Trans. Mobile
Comput. 21 (1) (2022) 31–43.
[4] B. Tan, et al., Optimal selection of energy efficiency measures for energy sustainability of existing buildings, Comput. Oper. Res. 66 (2016) 258–271.
[5] X. Li, F. Wang, M. Al-Razgan, E. Mahrous Awwad, S. Zilola Abduvaxitovna, Z. Li, J. Li, Race to environmental sustainability: can structural change, economic
expansion and natural resource consumption effect environmental sustainability? A novel dynamic ARDL simulations approach, Resour. Pol. 86 (2023) 104044.
[6] M. Hamdy, A.-T. Nguyen, J.L. Hensen, A performance comparison of multi-objective optimization algorithms for solving nearly-zero-energy-building design
problems, Energy Build. 121 (2016) 57–71.
[7] R. Zhao, X. Huang, J. Xue, A practical simulation of carbon sink calculation for urban buildings: a case study of Zhengzhou in China, Sustain. Cities Soc. 99
(2023) 104980.
[8] S. Yigit, B. Ozorhon, A simulation-based optimization method for designing energy efficient buildings, Energy Build. 178 (2018) 216–227.
[9] M. Hassan, et al., Thermal energy and mass transport of shear thinning fluid under effects of low to high shear rate viscosity, International Journal of
Thermofluids 15 (2022) 100176.
[10] A. Jia, et al., Energy efficiency measures in existing buildings by a multiple-objective optimization with a solar panel system using Marine Predators
Optimization Algorithm, Sol. Energy 267 (2024) 112208.
[11] P.K.S. Rathore, et al., Thermal performance of the building envelope integrated with phase change material for thermal energy storage: an updated review,
Sustain. Cities Soc. 79 (2022) 103690.
[12] E. Zavrl, U. Stritih, Improved thermal energy storage for nearly zero energy buildings with PCM integration, Sol. Energy 190 (2019) 420–426.
[13] N. Delgarm, et al., A novel approach for the simulation-based optimization of the buildings energy consumption using NSGA-II: case study in Iran, Energy Build.
127 (2016) 552–560.
[14] W. Kim, Y. Jeon, Y. Kim, Simulation-based optimization of an integrated daylighting and HVAC system using the design of experiments method, Appl. Energy
162 (2016) 666–674.
[15] F. Kheiri, A review on optimization methods applied in energy-efficient building geometry and envelope design, Renew. Sustain. Energy Rev. 92 (2018)
897–920.
[16] A.B.C. Board, Building Code of Australia, CCH Australia, 1996.
[17] R. Fay, R. Vale, P. Bannister, The National Australian Built Environment Rating System (NABERS), Environment Design Guide, 2004, pp. 1–6.
[18] A.A. Heidari, et al., Harris hawks optimization: algorithm and applications, Future Generat. Comput. Syst. 97 (2019) 849–872.
[19] Z. Cui, et al., A pigeon-inspired optimization algorithm for many-objective optimization problems, Sci. China Inf. Sci. 62 (7) (2019) 70212:1-70212:3.
[20] M. Abd El Aziz, A.A. Ewees, A.E. Hassanien, Whale optimization algorithm and moth-flame optimization for multilevel thresholding image segmentation, Expert
Syst. Appl. 83 (2017) 242–256.
[21] A. Cheraghalipour, M. Hajiaghaei-Keshteli, M.M. Paydar, Tree Growth Algorithm (TGA): a novel approach for solving optimization problems, Eng. Appl. Artif.
Intell. 72 (2018) 393–414.
[22] A. Faramarzi, et al., Marine Predators Algorithm: A Nature-Inspired Metaheuristic, Expert Systems with Applications, 2020 113377.

17