This article has been accepted for publication in IEEE Access.

This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2022.Doi Number

Optimal Placement and Capacity of Battery

Energy Storage System in Distribution
Networks Integrated with PV and EVs Using
Metaheuristic Algorithms
NATSAWAT POMPERN1, (Member, IEEE), SUTTICHAI PREMRUDEEPREECHACHARN1,
APIRAT SIRITARATIWAT2, AND SIROTE KHUNKITTI1
1
Department of Electrical Engineering, Faculty of Engineering, Chiang Mai University, Chiang Mai 50200, Thailand
2
Department of Electrical Engineering, Faculty of Engineering, Khon Kaen University, Khon Kaen 40002, Thailand
Corresponding author: Sirote Khunkitti ([email protected])
This research work is supported and funded by Provincial Electricity Authority – Graduate School, Chiang Mai University program.

ABSTRACT In this research, the optimal placement and capacity of battery energy storage systems
(BESS) in distribution networks integrated with photovoltaics (PV) and electric vehicles (EVs) have been
proposed. The main objective function is to minimize the system costs including installation, replacement,
and operation and maintenance costs of the BESS. The replacement cost has been considered over 20 years
while the operation and maintenance costs are the costs incurred by transmission line loss, voltage regulation,
and peak demand. To solve the problem, three metaheuristic algorithms, namely particle swarm optimization
(PSO), african vultures optimization algorithm (AVOA), and salp swarm algorithm (SSA), are employed.
The proposed approach is evaluated on the IEEE 33- and 69-bus distribution systems integrated with PV and
EVs. The results provided by the considered algorithms are compared in terms of the objective function,
system efficiency enhancement, payback period, and statistical analysis. The simulation results show that
after the BESS installation, the voltage profile can be improved, transmission loss is reduced, and peak
demand is decreased where PSO provides the best objective values and AVOA achieves the fastest payback
periods in both systems.

INDEX TERMS Battery energy storage systems, photovoltaics, electric vehicles, metaheuristic algorithms,
particle swarm optimization, african vultures optimization algorithm, salp swarm algorithm

I. INTRODUCTION distortion (THD), etc. [7], [8]. So that, to support the

At present, energy consumption tends to continuously increase of electricity consumption from EVs and load
increase because of economic growth and advanced growth, the efficiency of the distribution systems integrated
industrial development technology making electric power with PV is aimed to be enhanced by using an energy storage
an important factor in driving the economy and the quality system (ESS) for storing energy from PV and grid when
of life of people in many countries [1]–[3]. Moreover, demand is low and supply it back to the grid during high
electric vehicle (EV) and solar power generation industries, demand period.
which significantly impact the economy and electricity The technology of ESS for installation in distribution
generation, have been continuously growing [4]–[6]. The systems is chosen depending on several factors such as
increasing number of EVs needs a greater number of EV energy capacity, efficiency, cost, advantages, disadvantages,
charging stations affecting distribution systems to be and credibility of the ESS [9]. In addition, the ESS selected
vulnerable. Photovoltaic (PV) is becoming commonplace to be installed in the distribution systems should take a
in our daily lives and has been integrated into electric moderate amount of time to discharge (minutes-hours),
power distribution systems supporting distribution systems because this discharging time can respond to daily loads in a
to supply loads and improve voltage profile and harmonic timely manner [10].

VOLUME XX, 2023 1

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

By installing BESSs in distribution systems, the efficiency alone microgrid 17-bus system connected with residential
of the distribution systems can be improved depending on and EV loads by using teaching learning-based optimization
the placement and capacity of the BESS. Finding the best (TLBO). The results showed that losses in the system could
appropriate placement and capacity of the BESS is thus an be decreased, and voltage quality is improved. Y. Zheng et
important task that must be studied and considered to al. indicated the optimal number, placement, and sizes of the
improve distribution system performance and reliability of BESSs in radial distribution systems to control voltage
power generation and also minimize system costs. Several regulation and minimize life cycle costs by using a
works have studied many methods to determine the best hierarchical planning model and natural aggregation
appropriate placement and capacity of the BESS. In [11] , P. algorithm (NAA) [18]. Thus, several methods have been
Boonluk et al. presented the optimal placement and capacity used to determine the optimal placement and capacity of the
of the BESS together with renewable energy sources (RESs) BESS in the distribution systems in order to reduce system
by using genetic algorithm (GA) and particle swarm costs such as installation, replacement, transmission loss,
optimization (PSO) to determine the lowest system costs in voltage regulation, and peak demand costs. However, most
the IEEE 33-bus distribution network, and it was found that of them have only considered some of the system costs
PSO is more effective than GA. In [12], A. Fathy. proposed although all the system costs are important, and traditional
a novel artificial hummingbird algorithm (AHA) to algorithms have been used to solve the problems in most
determine the optimal locations and sizes of biomass-based studies.
distributed generators (DGs) in radial distribution networks So, several works presented the optimal placement and
in order to minimize active power loss and voltage deviation. capacity of the BESS for distribution systems by considering
It was found that the proposed algorithm could significantly all of the system costs, and some new efficient optimization
enhance network performance in terms of power loss and algorithms were employed. In [19], S. Khunkitti et al.
voltage deviation reductions. In [13] M. Tolba et al. proposed the optimal placement and sizing of the BESS for
introduced an effective optimization method called modified distribution networks connected with DGs by using GA,
forensic-based investigation (mFBI) optimizer to determine PSO, and SSA algorithms to minimize system costs
the optimal location and capacity of DGs in the IEEE 118- including transmission loss, voltage regulation, and peak
bus system and real distribution system in Delta-Egypt. The demand costs in the IEEE 33- and 69-bus distribution
optimization process was based on a multi-objective function systems. The results presented the BESS could mitigate the
aiming to minimize power losses, overall node voltage system costs, improve the voltage profile, decrease power
deviation and total annual operational costs, and maximize losses, and decrease peak demand. However, this research
voltage stability margin. The results demonstrate that the only considered the price of operation and maintenance of
mFBI optimizer outperforms other existing techniques, distribution systems, which did not include the costs of
providing superior outcomes in terms of the specified battery installation and replacement. In addition, V.Janamala
objectives. N. Jayasekara et al. introduced the optimal et al. proposed an approach to optimally integrate interline-
installation of the BESS together with PVs and winds PV (I-PV) system under different EV load penetration levels
turbines (WTs) by using the interior point method for by aiming at real power loss reduction and voltage profile
controlling operation together with data management system enhancement [20]. The coyote optimization algorithm
(DMS) and energy management system (EMS) in the IEEE (COA), PSO, and grey wolf optimizer (GWO) were used to
33-bus distribution system to minimize distribution system determine the optimal solutions. The simulation results
costs and battery cycling cost [14]. The results showed that presented that I-PV could be used to improve the voltage
the BESS can improve the efficiency of the distribution profile, decrease transmission loss, and adapt to real-world
system evaluated by examining voltage regulation situations; however, this paper only considered I-PV
enhancement, decreasing loss, and decreasing of peak installations where PV and BESS were installed at the same
demand. A. Mazza et al. considered finding the optimal placement, and the BESS installation costs were not
placement and capacity of the BESS connected with EVs and included. In [21], D. Sadeghi presented the optimal
DGs by using GA and greedy algorithm to minimize installation of BESS together with DGs and EVs by using
investment and operation and maintenance (O&M) costs in multi-objective PSO (MOPSO) and Monte Carlo simulation
a rural 22-bus distribution system which is a low voltage grid (MCS) to find optimal life cycle costs (LCCs) consisting of
[15]. In [16], A. Ahmadian presented the optimal BESS initial cost (IC), maintenance cost (MC), and replacement
installation by using a fuzzy model to predict the uncertainty cost (RC), and the results showed that the LCC could be
of load profile, and tabu search (TS) and simulated annealing decreased.
(SA) algorithms were applied to find the power of DGs, the Although various works have studied the optimal
number of battery units, power rating of BESS, and capacity placement and capacity of the BESS installation, most of
and locating of the BESS in the multi-objective problems them have not considered battery installation and
where the objectives included costs of electrical energy replacement costs as in [19], [20], and some of them used
purchased by distribution utility and power loss in the traditional optimization algorithms [21]. Thus, this work
distribution network. In [17], M. Khalid et al. introduced the presents the optimal placement and capacity of the BESS in
optimal placement and capacity of DGs and BESS in a stand- distribution networks connected with PV and EVs by

2 VOLUME XX, 2023

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

P.Natsawat et al.: Preparation of Papers for IEEE Access (February

2017)

considering system costs as the objective function. A can be increased by carefully controlling heat dissipation at
traditional efficient optimization algorithm which is PSO the optimal temperature which is around 15–35 degrees
[22] and new efficient algorithms which are AVOA [23] and Celsius, avoiding charging and discharging frequently.
SSA [24] are employed to determine the optimal solutions. Frequent charging and discharging should be avoided. The
This work's primary contributions can be summarized as: DOD of the Li-ion is recommended at 80% of the total
1) The study aims to determine the optimal placement and capacity. The rates of charge and discharge should not be too
capacity of the BESS in the IEEE 33- and 69-bus distribution high because it will cause high temperatures to the BESS,
systems connected with PV and EV penetrations by which will result in a shorter service life [28], [29].
considering overall system costs, which include investment,
replacement, operation, and maintenance costs as the B. BATTERY ENERGY STORAGE SYSTEM SIMULATION
objective function to be minimized. The simulation of the BESS considers the charging and
2) The efficiency of the distribution systems is enhanced by discharging of the BESS at equal intervals within the period
decreasing transmission line loss, decreasing peak demand, of 24 hours [11], [14], [19]. The period is divided equally
and improving voltage profile after the BESS installation. into 1 hour 30 minutes or 15 minutes, so BESS can support
3) Three mentioned optimization algorithms are utilized to the battery charging and discharging rate at a value of 24, 48,
find solutions to the problems, and their efficiency is or 96 respectively. The charging and discharging rates in any
evaluated through a comparative analysis. period are CiT, which can be determined according to the
4) The break-even point for investing in the BESS following equation:
installation is calculated for investment decisions, and the
statistical analysis is considered to validate the performance  EB (1) 
of each algorithm. CiT =  

(1)
The content has been arranged in topical order as follows:
Section 2 presents the ESS installation in a distribution  EB (m) 
system. In Section 3, the problem is defined and described.
The outlines of the metaheuristic algorithms and their where EB(t) represents the energy in the BESS (MWh) at
implementation are explained in Section 4. Section 5 depicts time t = 1, 2, 3, …, m.
the results and discussions. The work is finally concluded in The Fourier series is used to find the energy in the BESS
Section 6. by using the Fourier coefficient vector (CiF) provided by the
optimization process. The Fourier series is adopted in this
II. ENERGY STORAGE SYSTEMS IN A DISTRIBUTION work because it can be utilized to find the energy in the BESS
SYSTEM expressed in finely dispersed periodic patterns. By using the
ESS are terms used to describe systems or equipment Fourier series, the periodic pattern can be divided into a
installed in a system in order to convert electrical energy into group of sinusoidal components in time domain enabling a
other energy types such as electrochemical, mechanical, thorough analysis in the energy of the BESS [30]. The
chemical, thermochemical, and thermal energies that can be method is to start at random with 16 Fourier coefficient
stored and converted back into electrical energy again when values and use the Fourier transform to predict the BESS's
needed [10], [25]–[27]. In this research, BESS which, electrical energy (EB) every hour . The Fourier series is then
applies the energy conversion between electrical energy and used to represent the state of energy (SOE) over the entire
electrochemical energy, is used to be installed in the considered period. The energy in the BESS is determined by
distribution systems because of its suitable discharging time using the presented equations [14], [30].
that can respond to daily loads in a timely manner. The BESS
and its simulation are described in the following subsection.  a1 , b1 
CiF =  

(2)
A. BATTERY ENERGY STORAGE SYSTEMS (BESSs)
The BESSs take a moderate amount of time to discharge  an , bn 
(about minutes to hours) with a ratio between 1 and 10,
meaning that there is a capacity of 1–10 kWh for a system  2 t   2 t 
EB (t ) = a0 + a1 cos   + b1 sin   + ...
with a power of 1 kW. Examples of the BESSs are lead acid  T   T  (3)
batteries (LA), lithium-ion batteries (Li-ion), redox flow
 2 nt   2 nt 
batteries (RFB), and sodium sulphur batteries (NaS). The Li- + an cos   + bn sin  
ion battery has been chosen in this study because of its  T   T 
several advantages such as high efficiency of more than 90
%, high energy density (90-190 Wh/kg), up to 10,000 cycle where CiF, a0, an, bn, n, t, and T are the Fourier coefficient
times, and suitable cost. vector, constant Fourier coefficient, Fourier cosine
There are several factors affecting the service life of the coefficient, Fourier sine coefficient, number of Fourier
Li-ion batteries in the BESS, which should be considered, coefficients which is set to 8, time, and period, respectively.
including temperature, the operation cycle numbers of the The Fourier cosine and sine coefficients are the optimization
BESS, and depth of discharge (DOD) [11]. The BESS life variables for the problem.

VOLUME XX, 2023 3

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

The energy in the BESS, EB(t) is obtained by substituting where Cycles, D = 285 days [14], Cycle Life = 3,000 cycles
CiF from (2) into (3). From (3), a0 is not required because the [14], and Q are the number of daily BESS cycles, number of
constant a0 has no impact on the charging and discharging of BESS operating days, number of cycles in the nominal cycle
the BESS and has no impact on the daily energy cost life of Li-ion battery, and actual service life cycle in years,
coefficient. So, it can be set after an optimization process to respectively.
ensure that the BESS power curve is not below the minimum
value required to perform the DOD requirements. III. PROBLEM FORMULATION
The energies’ variation in the BESS at two consecutive To determine the optimal placement and capacity of the
times are computed by (4) in order to find the power of the BESS in distribution networks connected with PV and EV
BESS as in (5) and (6). The power of the BESS can then be charging stations, the costs of installation, replacement,
used to indicate the state of the BESS. When the BESS is in transmission loss, voltage regulation, and peak demand in
a charging state, the BESS power is positive, indicating the the distribution systems are considered as the objective
energy being added to the BESS. Conversely, when the function to be minimized while the constraints are the
BESS is in a discharging state, the BESS power is negative, voltages of all buses and BESS power and energy.
indicating the energy being released from the BESS.

EB = EB (t ) − EB (t − 1) (4)
A. OBJECTIVE FUNCTION
The objective function of this research is to decrease the total
PB (t ) = EB / ( t c ) , PB (t )  0 (5) costs of BESS installation (Csystem) including investment cost
(CIC), replacement cost (CRC), and operation and
PB (t ) = ( EB d ) / t , PB (t )  0 (6)
maintenance costs (CMC) [17]-[19]. The size of the BESS will
affect the investment cost while the replacement cost varies
where ΔEB is the energies’ variation in the BESS at two
depending on the size and life cycle time of the BESS.
consecutive times, ηc = ηd =√𝜼𝒃𝒂𝒕 , ηbat = 0.9 is the BESS Moreover, the operation and maintenance costs of the
efficiency in a cycle, and ηc, ηd, PB, and Δt are the BESS distribution systems include voltage regulation cost (CCVR),
charging efficiency, BESS discharging efficiency, BESS transmission loss cost (Closs), and peak demand cost (Cp).
power, and sampling interval time, respectively. Equation (10) represents the objective function of this work,
which can be obtained from (11) – (17).
C. OPTIMAL CAPACITY OF THE BESS
To find the optimal capacity of the BESS, the power capacity f (CiF ) = min ( Csystem ) (10)
and energy capacity should be considered to decrease the
total costs and maintain the quality and reliability of Csystem = CIC + CMC + CRC (11)
distribution systems. Moreover, the number of cycles and the
SOC should also be considered because they are two main CIC = N bat   IC (12)
factors that affect the BESS life [9]. By improving the CMC = CCVR + Closs + C p (13)

(  )
BESS's charging and discharging cycles, the daily SOC
Vi − Vref   VR
T N
swings can be decreased, which helps increase the efficiency CCVR = t =1 i =1
(14)
of the BESS's life cycle. The size of the BESS can be

= (  Lineloss )  
calculated by the difference between the maximum and T M
minimum energy in the BESS divided by the maximum Closs t =1 i =1 loss (15)
DOD as shown below.

EBmax − EBmin
C p = Pmax  t   p (16)
Battery size(kWh) = (7)
t year
DODmax
CRC = Nbat   IC  (17)
where DODmax = 0.8, EBmax and EBmin are the maximum and Cycles
minimum values of energy in the BESS, respectively. where Nbat, N, T, 𝑉𝑖 , 𝑉𝑟𝑒𝑓 , M, Lineloss, Pmax, 𝛾𝑉𝑅 , 𝛾𝑙𝑜𝑠𝑠 , 𝛾𝑝 ,
The BESS daily cycle and service life cycle can be
𝛾𝐼𝐶 , and tyear are the BESS size (kWh), total number of buses,
estimated from (8) and (9), respectively.
period, voltage at bus i (p.u.), reference voltage which is 1
1   t =1 EB (t ) − EB (t − 1) 
T p.u., total number of branches, actual power loss in each line,
Cycles =   (8) maximum power demand, rate of voltage regulator cost (𝛾𝑉𝑅
2  DODmax  Battery size  = 0.142 $/p.u. [14], [19]), rate of transmission loss cost
  (𝛾𝑙𝑜𝑠𝑠 = 0.284 $/kWh [14], [19]), rate of maximum energy
CyclesLife demand cost (𝛾𝑝 = 200 $/kWh/year [14], [19]), rate of BESS
Q( years) = (9) installation cost (𝛾𝐼𝐶 = 100 $/kWh [31]), and duration of this
Cycles  D
study (tyear = 20 years), respectively.

4 VOLUME XX, 2023

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

P.Natsawat et al.: Preparation of Papers for IEEE Access (February

2017)

B. CONSTRAINTS Each particle i then keeps the best answer in each iteration
1) INEQUALITY CONSTRAINTS called personal best (pbest). All particles share their best
Voltage conditions of each bus must be within the range of answer obtained so far, such that a single particle has access
±10% of the reference voltage as follows: to the most effective solution among all particles called
global best (gbest) in order to adjust its position and speed to
Vmin  Vi t  Vmax (18) the best solution. The velocity and position of the ith particle
are updated by the following equations.
where Vmin and Vmax are the minimum and maximum voltages
of each bus, respectively, and 𝑉𝑖𝑡 is the voltage at bus i at vik +1 = wik  vik + c1r1 ( pbest
k
, i − xi ) + c2 r2 ( g best − xi ) (22)
k k k

each time t.
xik +1 = xik + vik +1 (23)
In addition, the power and capacity of the BESS are
limited to prevent the BESS from being harmed during where the velocities of the ith particle at iterations k+1 and k
charging and discharging, which can be expressed by the are denoted by 𝑣𝑖𝑘+1 and 𝑣𝑖𝑘 , respectively. 𝑥𝑖𝑘+1 and 𝑥𝑖𝑘
following equations. represent the positions of the ith particle at iterations k+1 and
PBmin  Pcha
t t
, Pdis  PBmax (19) k, respectively, c1 and c2 are both positive acceleration
coefficients set to 2. r1 and r2 are two uniformly randomly
EBmin  EB  EBmax t
(20) 𝑘
produced numbers between [0,1]. 𝑝𝑏𝑒𝑠𝑡,𝑖 is the personal best
where 𝑃𝐵𝑚𝑖𝑛 and 𝑃𝐵𝑚𝑎𝑥 represent the minimum and position of the i particle at iteration k, g 𝑘𝑏𝑒𝑠𝑡 refers to the
th

maximum powers of the BESS, respectively. Recent global best position of all particles during the kth iteration,
charging and discharging powers of the BESS at time t are and w is the inertia weight at the kth iteration calculated by
𝑡
represented by 𝑃𝑐ℎ𝑎 𝑡
and 𝑃𝑑𝑖𝑠 , respectively, and the minimum the equation below.
and maximum capacities of the BESS are 𝐸𝐵min and 𝐸𝐵max , wmax − wmin
respectively. w = wmax − (  iter ) (24)
itermax
2) EQUALITY CONSTRAINT
For the equality constraint, the power balance is controlled where wmax is set to 0.9, wmin is set to 0.4, iter is the current
in each bus of a distribution system as shown in the equation iteration number, and itermax is the maximum iteration
below [20]. number.
Pgrid (t ) = PD (t ) − Ppv (t )  PB (t ) − PL (t ) (21)
B. AFICANS VULTURES OPTIMIZATION ALGORITHM
where Pgrid, PD, Ppv, PB, and PL are grid power, load demand (AVOA)
power, PV power, BESS power, and transmission line power AVOA is a meta-heuristic algorithm inspired by African
loss, respectively. vultures that migrate in flocks to locate food and cohabit
where the food location indicates the best solution. The
IV. METAHEURISTIC ALGORITHMS AND AVOA begins with a random sampling of the vultures' initial
IMPLEMENTATION positions in the problem area, and the best two vultures are
In this research, to determine the optimal placement and evaluated by the following equation.
capacity of the BESS in distribution systems connected with
 BestVulture1 if pi = L1 (25)
PV and EVs, three metaheuristic algorithms including PSO, R(i) = 
AVOA, and SSA are adopted. The effectiveness of these  BestVulture2 if pi = L2
algorithms is compared in terms of voltage deviation index Fi (26)
pi =
 i =1 Fi
n
(VDI), power losses, and peak demand enhancement. The
implementation of the whole process is explained in the
following subsections. where R(i) is one of the best vultures selected, pi is the
probability of choosing the best solution, n is the number of
A. PARTICLE SWARM OPTIMIZATION (PSO) group’s vulture, L1 and L2 are the indicators calculated before
PSO is one of the most widely used techniques for solving the searching process, with the values in the range of 0 and
optimization problems because of its simplicity and high 1, and the sum of them is equal to 1. The rates of starvation
efficiency in searching for solutions. Although PSO is a of both vultures are then calculated by the following
traditional method, it could overcome new proposed equations.
optimization algorithms in some works [19], [32], [33]. So,
 iter 
it is possible that PSO can be better than modern algorithms F = ( 2  rand1 + 1)  z  1 − +t
(27)
in this work. PSO was proposed by Kennedy and Eberhart  iter max 
inspired by the social behavior of a group of fishes or birds  w  iter   iter  
t = h   sin   (28)
looking for food [22]. In the PSO simulation, firstly, the  + cos    − 1
position and speed of each particle are randomly initialized.   2 itermax   2 itermax  

VOLUME XX, 2023 5

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

where F indicates the rate of starvation, rand1 has a random BestVulture2 (i)  P(i) (38)
A2 = BestVulture2 (i ) − F
value that varies between 0 and 1, iter is the current iteration, BestVulture2 (i) − P(i)2
itermax is the maximum number of iterations, z is a randomly A1 + A2 (39)
P(i + 1) =
generated number in the range of [-1,1] regenerating in each 2
iteration, t is adopted to improve the search process by P(i + 1) = R(i) − d (t )  F  Levy (d ) (40)
avoiding the local optima, h is a number randomly chosen
between -2 and 2, w is a parameter used to balance the where A1 and A2 are great deals of competition for food that
exploration and exploitation phases [23]. When z falls below may accumulate various species of vultures on one food
zero in the range of [-1,0), the vulture is starved, and when it source, BestVulture1(i) is the best vulture of the first group in
rises in the range of [0,1], the vultures are satiated. iteration i, BestVulture2(i) is the best vulture of the second
If the vultures have a value of F ≥ 1, then the vultures are group in iteration i, Levy(d) is a Levy flight used to increase
satiated. Vultures will explore in search of food at random the randomness of the AVOA computed by the given
distances from one of the two groups and update the position equation.
by using the presented equations.
1
   (41)
D(i ) = X  R(i ) − P(i ) (29) u   (1 +  )  sin( 2 ) 
Levy ( d ) = 0.01 , = 
1
 (1 + 2  )    2(
 − 1 
P(i + 1) = R(i ) − D(i )  F (30) v
 2 
)

P(i + 1) = R(i) − F + rand 2  ( ( ub − lb )  rand3lb ) (31)

where u and v are a random number in the range of [0,1], and
where D(i) is a parameter used to update the position of the β is equal to 1.5.
best vultures in two groups, X is the vultures move randomly
to prevent food from others, P(i) is the vulture position C. SALP SWARM ALGORITHM (SSA)
vector in iteration i, P(i+1) is the vulture position vector in SSA is a metaheuristic algorithm inspired by the chain
iteration i+1, rand2 and rand3 are random number generated motion of the salps in the deep ocean. By classifying salps
between [0,1], and lb and ub illustrate the lower and upper into two groups, leaders and followers, the leaders are at the
limits of the variables. front of the chain and the rest of the salps are followers. The
Vultures are quite full and very active if 0.5≤F<1. As a leader guides them to approach the food source, represented
result, there will be violent disputes between vultures, with by the variable F in the n-dimensional search area, which is
aggressive vultures refusing to share their meal. Weak the number of variables in each problem. To update the
vultures receive food scraps from stronger vultures by position of the leader, the following equation is applied.
updating the position of the vulture by using the given
 Fj + c1 ((ub j − lb j )c2 + lb j ), c3  0
equations. x1j =  (42)
 Fj − c1 ((ub j − lb j )c2 + lb j ), c3  0
d (t ) = R (i ) − P (i ) (32)
P(i + 1) = D(i)  ( F + rand 4 ) − d (t ) (33) where xj1 represents the leader’s position of the swarm in the
jth dimension, Fj represents the food source in the jth
 rand5  P(i )  (34)
S1 = R(i)     cos( P(i)) dimension, ubj is the upper bound of the jth dimension, lbj is
 2 
the lower bound of the jth dimension, and c2, and c3 are
 rand6  P(i )  (35)
S2 = R(i )     sin( P(i)) randomly generated numbers.
 2  The parameter c1 is an important parameter in SSA used to
P(i + 1) = R(i ) − ( S1 + S 2 ) (36) balance the exploration and exploitation phases which can be
calculated by the given equation.
where d(t) is the distance between the vulture and one of the
4 iter 2
best vultures from the two groups, S1 and S2 are a spiral −( )
equation generated between all vultures and one of the two c1 = 2e itermax
(43)
best vultures, and rand4, rand5, and rand6 are random
numbers in the range of [0,1]. where iter is the current iteration and itermax is the maximum
When the vulture's hunger level is F < 0.5, it assumes a iteration number.
situation in which the number of vultures is more than the The parameters c2 and c3 are random numbers uniformly
number of foods resulting in conflicts between vultures. The randomly generated in the range of [0,1] that determine the
food is usually in one place, and all the vultures tend to fly next position of the salp swarm.
there. Different equations are applied to update the vulture To update the location of the follower group, Newton's
position, and the best foraging response is then produced as laws of motion are applied where the time in optimization is
shown in the equations below. the iteration, and the discrepancy between iterations is equal
to 1 as expressed in the following equation.
BestVulture1 (i)  P(i) (37)
A1 = BestVulture1 (i ) − F 1 i
BestVulture1 (i) − P(i) 2 xij = ( x j + xij−1 ), i  2 (44)
2

6 VOLUME XX, 2023

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

P.Natsawat et al.: Preparation of Papers for IEEE Access (February

2017)

where i≥2, xij is the position of the ith follower in the jth BESS in distribution systems connected with PV and EV
dimension, 𝑥𝑗𝑖−1 is the position of the followers in the jth charging stations. The implementation of the proposed
dimension. method is depicted as a flowchart in Fig. 1, and the tested
systems and EV modeling are described in the following
D. DISTRIBUTION SYSTEM EFFICIENCY EVALUATION subsections.
In this research, distribution system efficiency is investigated
after the BESS installation. The efficiency is investigated in
terms of transmission loss, VDI, and peak demand, which are
described in the following subsections.
1) VOLTAGE DEVIATION INDEX (VDI)
VDI is used to indicate the efficiency of the distribution
system evaluated by examining voltage profile enhancement
when installing the BESS. The VDI can be calculated by the
percentage of the difference between the reference voltage
and the actual voltage for the period as presented in the
following equation.
Vref − Vi
%VDI i = max T 100,
Vref (45)
Nbus
%VDI =  %VDI i
i =1

where %VDIi is the maximum percentage of VDI at bus i for

the period time T, Vref is the reference bus voltage, Vi is the
voltage at bus i, %VDI is the total percentage of VDI of the
distribution system, and Nbus is the total number of buses.
2) TRANSMISSION LOSSES
The transmission losses consist of active power, reactive
power, and apparent power losses, which are used to
compare the efficiency of the system after the BESS
installation calculated by the following equations.
T Nl
Ploss =  Pl t ,
t =1 l
T Nl
Qloss =  Qlt , (46)
t =1 l

Sloss = Ploss 2 + Qloss 2

where Ploss, Qloss, and Sloss are the active power, reactive
power, and apparent power losses for the period time T,
respectively, Plt, Qlt are the active and reactive power
losses of line l for the period time T.
3) PEAK DEMAND
Peak demand represents the maximum active power demand
during the considered time interval which is 24 hours in this
study. To further evaluate the efficiency enhancement of the
distribution systems, the peak demand enhancement
indicated by the peak shaving is investigated after the BESS
installation.

E. IMPLEMENTATION
To determine the optimal solution, three metaheuristic
algorithms consisting of PSO, SSA, and AVOA are
employed to find the optimal placement and capacity of the FIGURE 1. Implementation of the proposed method.

VOLUME XX, 2023 7

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

1) SYSTEM MODELLING 10 0.937 0.757 22 0.911 0

The IEEE 33-bus and IEEE 69-bus distribution systems are 11 0.941 0.904 23 0.889 0
adopted to determine the optimal placement and capacity of 12 0.933 0.979 24 0.867 0
the BESS. The single-line diagrams of these systems are
shown in Fig. 2 and Fig. 3, respectively [34], [35]. The base
power is 10 MVA and the base voltage is 12.66 kV for both
systems. The PV which is 5000 kW in this work was installed
on the 6th bus which is the best bus for the PV installation as
presented in [20]. The maximum load demand for the IEEE
33-bus and 69-bus systems are 3,715 kW and 3,802 kW
respectively. The load demands are presented as the 24 hours
in p.u. of the base power at 2,070.86 kW as in Table I. Fig. 4
presents the relationship between time and the amount of the
original load demand, and the solar power generation
capacity within a day of the IEEE 33-bus and IEEE 69-bus
distribution systems.

FIGURE 4. Load Demand and PV load profile of the considered

distribution systems [20].
2) MODELLING OF CHARGING STATIONS FOR EVs
To model the charging stations for EVs, EV load demand on
all buses is assumed to be increased by using an AC/DC
converter or charging port since the BESS and grid are used
to supply power to the EVs by the penetration load [20], [36].
Equations (47) and (48) specify additional real and reactive
power loads of EVs, respectively.
Pev0 ( n ) = ev  PL0( n ) (47)
 tan(n ( c ) )
FIGURE 2. Single-line diagram of the IEEE 33-bus distribution system
[34]. Q 0
ev ( n ) =P 0
ev ( n )
(48)
0 0
where 𝑃𝑒𝑣(𝑛) and 𝑄𝑒𝑣(𝑛) are active and reactive powers of the
additional load due to the integration of EVs at the nth bus,
𝜆𝑒𝑣 is a scale factor representing the amount of EV load
compared to the actual power load at the considered
placement, P0L(n) is the actual nominal power of the load at
the nth bus, P0ev(n) is the added EV load of the nth bus, and
𝜑𝑛(𝑐) is the AC/DC converter's power factor.
The summation of the real and reactive load powers after
considering the EV penetration at different placements can
be presented by the expressed equations.
  ev
 V(tn )    V(tn )   (49)
Pdt ( n ) = PL0( n )   0  +  Pev0 ( n )   0  
V  V 
 ( n )    ( n )  
FIGURE 3. Single-line diagram of the IEEE 69-bus distribution system
 ev
[35].  V(tn )    V(tn )    (50)
Qdt ( n ) = QL0( n )   0  + Qev0 ( n )   0  
V  V 
TABLE I  ( n )    ( n )   
Hourly load profile and PV generation profile in a typical day [20].
Hr
Load PVp
Hr
Load PVp Where Ptd(n) and Qtd(n) are the active and reactive load
(p.u.) (p.u.) (p.u.) (p.u.)
1 0.867 0 13 0.911 1.000 powers at the nth bus after consideration in conjunction with
2 0.852 0 14 0.904 0.943 the EV load, respectively, Q0L(n) is the reactive nominal
3 0.837 0 15 0.907 0.820 power of the load at the nth bus, respectively, V0(n) and Vt(n)
4 0.830 0 16 0.911 0.634 are the initial and time nominal voltages, respectively, 𝛼 =
5 0.830 0 17 0.915 0.387
0 and 𝛽 = 0 are the real and reactive power exponents of the
6 0.859 0 18 0.919 0.144
7 0.889 0 19 1.000 0 load demand, respectively, and 𝛼𝑒𝑣 = 2.59 and 𝛽𝑒𝑣 = 4.06
8 0.926 0.216 20 1.000 0 are the real and reactive vehicle load exponents, respectively
9 0.933 0.531 21 0.963 0

8 VOLUME XX, 2023

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

P.Natsawat et al.: Preparation of Papers for IEEE Access (February

2017)

[36]. Fig. 5 and Fig. 6 show the load demand when the EV
penetration is at 20%.

A. IEEE 33-BUS DISTRIBUTION SYSTEM

The optimal placement and capacity of the BESS was firstly
determined to install in the IEEE 33-bus distribution system.
The load data of this system was obtained from Table 1, and
the loads on all buses were increased according to EV
penetration by using (46)-(49). The efficiency of each
algorithm is investigated after the BESS installation in terms
of the minimum system costs which include the investment
cost, replacement cost, and operation and maintenance costs
of installing the BESS. Moreover, the efficiency
enhancement of the distribution system provided by all
considered algorithms was compared in terms of the VDI,
FIGURE 5. Load demand, PV profile, and EV penetration at 20% of the transmission line loss, and maximum peak demand after the
IEEE 33-bus distribution system [20]. BESS installation.
1) OPTIMAL PLACEMENT AND CAPACITY OF THE BESS
The optimal placement and capacity of the BESS were
evaluated in the IEEE 33-bus distribution system in order to
minimize system costs which comprise investment,
replacement, and operating costs by applying PSO, AVOA,
and SSA. The optimized values of the decision variables,
Fourier coefficients, calculated by three considered
algorithms are provided in Table II. The best Fourier
coefficients obtained by each algorithm were used to
determine the SOE of the BESS within 24 hours as presented
in Fig. 7. The optimal placement and capacity of the BESS
together with the power of the BESS, lifetime of the BESS,
and system cost were then found by all considered
FIGURE 6. Load demand, PV profile, and EV penetration at 20% of the algorithms as shown in Table III. It was noticed that the best
IEEE 69-bus distribution system [35].
placements of the BESS provided by all algorithms were at
bus 6 where PSO required the largest BESS size of 5.3342
V. SIMULATION RESULTS AND DISCUSSION
MWh, followed by SSA at 5.2113 MWh, and AVOA
The optimal placement and capacity of the BESS were
requires the smallest size of the BESS at 4.4386 MWh,
simulated in the IEEE 33- and 69-bus distribution systems
respectively. In term of the objective function, which is the
connected with PV and EVs in order to find the minimum
system costs, PSO generated the best objective value
costs of installation, replacement, transmission loss, voltage
followed by SSA and AVOA, respectively.
regulation, and peak demand. MATLAB and MATPOWER
were utilized to run the simulation to simulate the power TABLE II
The best Fourier coefficients provided by each algorithm in the IEEE 33-
flow. The computational complexity of the proposed bus distribution system
techniques depends on the Fourier series applied to predict
the energy of the BESS in a distribution system. The Fourier Algorithms Fourier Coefficients
series then primarily depends on the number of Fourier a1 b1 a2 b2
coefficients considered in the series. The more number of 0.10322 -1.74857 -0.70902 0.15369
Fourier coefficients provides more accuracy of the generated a3 b3 a4 b4
solutions but also requires higher computational effort. -0.03417 0.18476 0.10263 0.03452
PSO a5 b5 a6 b6
Moreover, to compare distribution system efficiency
0.08428 0.02162 -0.03887 -0.04304
enhancement before and after the BESS installation, three
a7 b7 a8 b8
metaheuristic algorithms, including PSO, AVOA, and SSA
-0.00464 -0.02699 -0.01250 0.03469
were employed. The number of the populations and
a1 b1 a2 b2
maximum iterations of algorithm operation were imposed to -0.25026 -0.88360 -0.49761 0.06689
60 and 250, respectively. The simulation results are a3 b3 a4 b4
presented as follows: 0.00184 0.17462 0.11483 0.01912
AVOA a5 b5 a6 b6
0.01037 0.01036 -0.02604 -0.01841
a7 b7 a8 b8
-0.00089 -0.01067 0.00674 -0.00182

VOLUME XX, 2023 9

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

a1 b1 a2 b2 TABLE IV
0.04825 -1069711 -0.62566 0.19301 Efficiency comparison of the BESS installation in the IEEE 33-bus
distribution system
a3 b3 a4 b4
-0.08193 0.07924 0.12065 0.14032 Real Reactive Apparent Maximum
SSA power power power peak
a5 b5 a6 b6 Algorithms VDI (%)
loss loss loss Demand
0.9653 -0.88260 -0.10149 0.02082
(MW) (MVAr) (MVA) (MW)
a7 b7 a8 b8
Base 200.2029 4.4743 3.0297 5.4036 4.6484
0.04179 -0.01961 -0.01447 0.02059
PSO 172.1783 4.2803 2.9046 5.1728 3.8197
AVOA 176.5046 4.2925 2.9125 5.1873 3.9286
SSA 173.2407 4.2934 2.9130 5.1883 3.8288

The enhancement of the 24-hour voltage profile at the 18th

bus, which is the weakest bus before and after the BESS
installation by each algorithm is shown in Fig. 8. It is
observed that the voltage was lower than the lower voltage
constraint from 7:00 p.m. to 8:00 p.m. in the base case and
could be increased to be within the constraints after installing
the BESS by all considered algorithms. Fig. 9 presents the
voltage profile of all buses in the system at 8:00 p.m., which
was the time when the peak demand was required in a day.
It can be seen that the voltage was the lowest in the 18th bus
because it was the farthest bus from the reference bus, and
after installing the BESS, it can be seen that the voltage in
the 18th bus was improved to be within the constraints.
FIGURE 7. Stage of Energy (SOE) of the BESS provided by each
algorithm for the IEEE 33-bus distribution system.

TABLE III
Optimal placement and capacity of the BESS in the IEEE 33-bus
distribution system
Power
Best BESS
of Lifetime System Costs
Algorithms place size
BESS of BESS ($)
ment (MWh)
(MW)
PSO 6 5.3342 0.9439 8.8247 25,933,282.71
AVOA 6 4.4386 0.9426 8.7514 26,110,137.49
SSA 6 5.2113 1.0463 8.8211 25,957,185.51

2) ALGORITHM EFFICIENCY COMPARISON

The efficiency enhancement of the IEEE 33-bus distribution
system was evaluated by examining the VDI, transmission
line loss, and maximum peak demand before and after the
BESS installation by using PSO, AVOA, and SSA as
presented in Table IV. It was found that after the BESS was FIGURE 8. Voltage profile comparison of the 18th bus for the IEEE 33-
bus system.
installed in the system, all three algorithms achieved decreases
in the VDI, transmission losses, and peak demand, which
improved the system efficiency. PSO provided the best VDI
value, followed by SSA, and AVOA, respectively. For the
transmission line losses, PSO obtained the largest decrease of
loss followed by AVOA and SSA, respectively. The
maximum peak demands after the BESS installation were
decreased with the largest reduction by PSO followed by SSA
and AVOA, respectively.

10 VOLUME XX, 2023

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

P.Natsawat et al.: Preparation of Papers for IEEE Access (February

2017)

than the base case because the BESS was supplied together
with the power from the grid. From 8:00 a.m. to 6:00 p.m.,
the peak demand was higher than the base case because
BESS was in the charging state to store power for supply
during the peak time.

FIGURE 9. Voltage profile comparison of all buses at 8.00 p.m. for the
IEEE 33-bus.

To investigate the transmission loss at each hour after

installing the BESS by all algorithms, real power loss with a
day in this system is plotted in Fig. 10. The loss slightly FIGURE 11. Peak demand comparison for the IEEE 33-bus system.
decreased when compared to the base case from 1:00 a.m. to
3) COMPARISONS OF ALGORITHM PERFORMANCE
8:00 a.m. and increased more than the base case from 8:00
a.m. to 6:00 p.m. because the BESS was in the charging state To further investigate and validate the efficiency after the optimal
to reserve power for peak times. The loss was significantly placement and capacity of the BESS installation obtained from
reduced from 6:00 p.m. to 0:00 a.m. because the BESS the algorithms, the statistical analysis, operation times
discharged power to supply the load demand together with convergence curves of the minimum costs and payback period are
the grid. So, the overall 24-hour transmission line loss of the presented. The statistical results including best value, worst value,
distribution system was decreased after the BESS installation mean, median and standard deviation of the objective values
compared to the base case. generated by each algorithm together with operation times are
expressed in Table V. It is noticed that PSO gave the best values
in all aspects except for standard deviation while AVOA
generated the best standard deviation. However, SSA consumed
the shortest operation time followed by PSO and AVOA. The
comparison results of system costs converging for 250 iterations
of the considered algorithms are shown in Fig. 12. When the
BESS was installed on the 6th bus, PSO generated the most
effective system costs, followed by SSA and AVOA,
respectively, and PSO was also the fastest one to converge to the
minimum objective value. After that, the obtained information
was used to calculate the break-even point for investment
consideration. The break-even point was calculated by the
installation costs of the BESS divided by the difference between
the O&M costs of the system before and after the BESS
installation. The results are presented in Table VI, where BESS
installed by using AVOA showed the fastest payback period,
FIGURE 10. Real power loss comparison for the IEEE 33-bus system. followed by SSA and PSO, respectively, because AVOA has the
smallest size of the BESS compared to the other algorithms
Fig. 11 expresses the comparison of peak demand for the according to Table III, resulting in BESS installation using a
IEEE 33-bus distribution system to evaluate the peak lower budget.
demand enhancement after the BESS installation. Peak
demand for the base case was at its highest at 8:00 p.m., as
well as from 1 a.m. to 8:00 a.m. and from 6:00 p.m. to 0:00
a.m. It can be seen that the peak demand after installing the
BESS by all considered algorithms was improved to be less

VOLUME XX, 2017 11

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

TABLE V are the Fourier coefficients, generated by three considered

Statistical results of each algorithm in the IEEE 33-bus distribution system
algorithms are found as shown in Table VII. The battery
energy in the 24-hour period was then calculated by using
Algorithms PSO AVOA SSA
the Fourier coefficients, and the states of energy provided by
each considered algorithm were plotted as in Fig. 13. The
Best 25,933,282.71 26,110,137.49 25,957,185.51
placement of the BESS installation provided by PSO was at
Worst 26,200,842.01 26,393,674.66 26,633,224.01 the 54th bus and at the 55th bus for both AVOA and SSA. The
Mean 26,040,010.76 26,209,077.01 26,236,034.25 placement, capacity, power, lifetime, and system costs of the
Median 26,014,139.99 26,186,509.77 26,157,091.50
BESS provided by each algorithm are presented in Table
VIII. It was found that PSO required the largest BESS size
Std. 87,562.67 79,907.06 203,617.23
of 3.7692 MWh, followed by SSA at 3.5166 MWh and
Time (s) 23,242.32 23,392.12 23,227.05 AVOA at 3.2089 MWh, respectively. For the system costs,
PSO obtained the best costs followed by AVOA and SSA,
TABLE VI respectively.
System cost and payback period comparison of the IEEE 33-bus
distribution system TABLE VII
Operation and The best Fourier coefficients provided by each algorithm in the IEEE 69-
Payback bus distribution system
Algorithms System costs ($) maintenance costs
(year)
for 1 day ($)
Algorithms Fourier Coefficients
Base - 3,823.00 -
a1 b1 a2 b2
PSO 25,933,282.71 3,313.82 2.8702 -0.09183 -1.25233 -0.69867 0.07528
AVOA 26,110,137.49 3,376.97 2.7264 a3 b3 a4 b4
-0.00346 0.21159 0.10899 0.00227
SSA 25,957,185.51 3,322.53 2.8529 PSO a5 b5 a6 b6
0.00572 0.06526 0.06811 -0.05842
a7 b7 a8 b8
-0.05844 -0.04646 -0.00705 0.05010
a1 b1 a2 b2
-0.04958 -0.97433 -0.60675 0.00536
a3 b3 a4 b4
-0.03994 0.23192 0.13353 0.04209
AVOA a5 b5 a6 b6
-0.00052 -0.00928 0.02007 -0.03136
a7 b7 a8 b8
-0.01360 -0.02457 -0.00842 0.02156
a1 b1 a2 b2
-0.32422 -1.11871 -0.54196 0.19744
a3 b3 a4 b4
-0.03062 0.08279 0.13438 0.06685
SSA
a5 b5 a6 b6
-0.00445 -0.03240 -0.01543 -0.00902
a7 b7 a8 b8
0.00661 0.02062 -0.00521 -0.00550

FIGURE 12. Convergence curves of the system costs of each algorithm

for the IEEE 33-bus system.

B. IEEE 69-BUS DISTRIBUTION SYSTEM

To confirm the efficiency of the proposed method in a larger
system, the IEEE-69 bus distribution system was tested to
determine the optimal placement and capacity of the BESS
in the distribution system. The load data of this system is
provided in Table I. The simulation results of this system are
presented in the subsequent sections.
1) OPTIMAL PLACEMENT AND CAPACITY OF THE BESS
To determine the optimal capacity and placement of the
BESS with minimum system costs in the IEEE 69-bus
system, the optimized values of the decision variables, which

12 VOLUME XX, 2023

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

P.Natsawat et al.: Preparation of Papers for IEEE Access (February

2017)

TABLE IX
Efficiency comparison of the BESS installation in the IEEE 69-bus
distribution system
Real Reactive Apparent Maximum
power power power peak
Algorithms VDI (%)
loss loss loss Demand
(MW) (MVAr) (MVA) (MW)
Base 216.4787 5.8152 2.6273 6.3812 4.7834
PSO 199.6260 5.8202 2.6286 6.3863 4.0415
AVOA 199.0248 5.8213 2.6294 6.3875 4.0974
SSA 197.4168 5.8195 2.6283 6.3855 4.0836

The voltage profile of the weakest bus, which is the 65th

bus over 24 hours is displayed in Fig. 14 to illustrate the
enhancement in the voltage profile after the BESS
installation by all considered algorithms. It is noticeable that
between 7:00 p.m. and 8:00 p.m., the voltage for a base case
FIGURE 13. SOE of the BESS provided by each algorithm for the IEEE
69-bus distribution system.
was lower than the lower voltage constraint; however, it can
be improved within the constraints once BESS was installed.
TABLE VIII Fig. 15 presents the voltage profile of all buses to observe the
Optimal placement and capacity of the BESS in the IEEE 69-bus overall voltage enhancement of the system when the
distribution system
majority of a load demand was required at 8:00 p.m. It is
Power
Best BESS
of
Lifetim
System Costs noticed that the 65th bus had the lowest voltage which was
Algorithms place size e of below the voltage constraint because it was the farthest bus
BESS ($)
ment (MWh) BESS
(MW) from the reference bus. However, after installing the BESS
PSO 54 3.7692 0.9769 8.2254 29,572,175.59 in the optimal placement, the voltage profile at the 65th bus
AVOA 55 3.2089 0.7901 8.2456 29,603,862.21 increased more than the base case and within the constraints.
SSA 55 3.5166 0.6611 7.9646 29,680,415.60

2) ALGORITHM EFFICIENCY COMPARISON

The efficiency of all algorithms in the IEEE 69-bus
distribution system is evaluated by examining the VDI,
transmission losses, and maximum peak demand before and
after the installation of the BESS as presented in Table IX.
By evaluating the comparison of the VDI, it was observed
that when BESS was installed in the distribution system, all
three algorithms resulted in a VDI reduction, which was
advantageous to the distribution system. The best VDI was
obtained by SSA followed by AVOA, and PSO, respectively.
It was observed that the BESS installation cannot decrease
transmission losses in the IEEE 69-bus distribution systems.
This is because, during the period between 1:00 a.m. to 6:00
a.m. and 8:00 a.m. to 6:00 p.m., the BESS was charging to
reserve power to decrease peak demand during the peak
demand period from 6:00 p.m. to 0:00 a.m. In term of peak FIGURE 14. Voltage profile comparison of the 65th bus for the IEEE 69-
demand reduction comparison, the highest reduction of peak bus system.
demand was provided by PSO, followed by SSA and AVOA,
respectively.

VOLUME XX, 2017 13

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

backup power to decrease peak demand during the peak load.

So, peak demand when installing the BESS was significantly
improved from that of the base case between 6:00 p.m. to
0:00 a.m. and slightly enhanced from the base case between
6:00 to 8:00 a.m. Consequently, the overall 24-hour peak
demand after installing the BESS by all algorithms could be
decreased.

FIGURE 15. Voltage profile comparison of all buses at 8.00 p.m. for the
IEEE 69-bus system.

The transmission real power loss in the IEEE 69-bus

distribution system is shown in Fig. 16 to notice the line loss
at each hour. It is noticed that from 1:00 a.m. to 6:00 a.m.
and 8:00 a.m. to 6:00 p.m., the transmission line loss
increased more than the base case since the BESS was in the
charging state to reserve power for peak demand period.
However, the loss was significantly reduced from 6:00 p.m.
to 0:00 a.m. because the BESS discharged power to help FIGURE 17. Peak demand comparison for the IEEE 69-bus system.
supply the system together with the grid. As a result, the 3) COMPARISONS OF ALGORITHM PERFORMANCE
overall 24-hour transmission line loss of the distribution To compare the efficiency of the considered algorithms to
system was not decreased compared with that of the base determine the optimal placement and capacity of the BESS in the
case. IEEE 69-bus system, the statistical results and operation times of
all algorithms are presented in Table X, the convergence curves
presenting the minimum system costs over 20 years of each
algorithm are shown in Fig. 18, and the payback period is also
compared. From the statistical results in Table X, PSO generated
the best values in best, worst, mean, and median values followed
by AVOA and SSA while AVOA obtained the best standard
deviation value followed by PSO and SSA. SSA used the fastest
operation time followed by PSO while AVOA consumed the
longest time. For the convergence curves in Fig. 18, it is seen that
PSO was the first algorithm to converge to the optimal solution
followed by AVOA and SSA, respectively. The break-even point
for investment consideration was investigated as presented in
Table XI, where BESS installed using AVOA showed the fastest
payback period, followed by SSA and PSO, respectively, because
AVOA has the smallest size of the BESS compared to the other
algorithms according to Table VIII. As a result, PSO provided the
best costs in terms of technical price and AVOA gave the best
cost in terms of economic price.
FIGURE 16. Real power loss comparison for the IEEE 69-bus system.

Peak demands for the IEEE 69-bus distribution system

before and after the BESS installation by the considered
algorithms are compared in Fig. 17. It is seen that the base
case had the highest peak demand at 8:00 p.m. It can be
observed that from 1:00 a.m. to 6:00 a.m. and 8:00 a.m. to
6:00 p.m., peak demand after the BESS installation was
higher than that of the base case because BESS charged

14 VOLUME XX, 2023

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

P.Natsawat et al.: Preparation of Papers for IEEE Access (February

2017)

TABLE X AVOA, and SSA are used to solve the optimization problem,
Statistical results of each algorithm in the IEEE 69-bus distribution system
and the IEEE 33- and 69-bus distribution systems are tested.
Algorithms PSO AVOA SSA
In the IEEE 33-bus system, the results show that PSO could
obtain the optimal placement and capacity of the BESS with
Best 29,572,175.60 29,603,862.21 29,680,415.59 the best system costs followed by AVOA and SSA,
respectively. PSO also provided the best VDI, decreasing of
Worst 29,957,728.13 30,046,106.10 32,959,028.38
transmission loss, and decreasing of peak demand after the
Mean 29,721,951.40 29,780,709.30 30,592,240.12
BESS installation followed by AVOA and SSA in term of
Median 29,671,100.87 29,756,369.56 30,348,051.77 loss reduction and SSA and AVOA in term of peak demand
Std. 130,722.10 111,396.71 992,083.15 reduction while AVOA achieved the fastest payback period
Time (s) 59,299.55 60,141.59 56,469.76 followed by SSA and PSO. The statistical results of PSO are
the best ones compared to the other algorithms except for
standard deviation which AVOA provided the best while
TABLE XI
System cost and payback period comparison of the IEEE 69-bus
SSA used the fastest operation time. It is also found that
distribution system installing the BESS by all considered algorithms could move
Operation and the voltage of all buses from the base case to be within the
Payback
Algorithms System costs ($) maintenance costs constraints. The simulation results of the IEEE 69-bus
(year)
for 1 day ($)
system express that PSO provided the best system costs for
Base - 4,278.9010 -
giving the optimal placement and capacity of the BESS
PSO 29,572,175.59 3,872.9910 2.5440 followed by AVOA and SSA, respectively. The efficiency
AVOA 29,603,862.21 3,904.7884 2.3493 comparison in this system found that SSA obtained the least
SSA 29,680,415.60 3,896.6692 2.5206 VDI followed by AVOA and PSO while PSO achieved the
best maximum peak demand reduction followed by SSA and
AVOA, and AVOA provided the fastest payback period
followed by SSA and PSO. PSO also generated the best
values in all aspects of the statistical results except for
standard deviation while AVOA generated the best standard
deviation and SSA consumed the shortest operation time.
The out-of-range bus voltages of the base case could also
move to be within the constraints by the BESS installation
by each algorithm. However, the overall transmission line
losses after the BESS installation provided by all considered
algorithms were close to those of the base because the IEEE
69-bus system is too large for only one installation BESS.
The results show that it is not always necessary to install
BESS on the same bus as the PV when the distribution
system is larger. In future work, the optimal placement and
capacity of more than one BESSs will be considered for
installation in distribution systems to evaluate system
efficiency enhancement and system costs.

FIGURE 18. Convergence curves of the system costs of each algorithm

for the IEEE 69-bus system.
REFERENCES
[1] A. Raihan and A. Tuspekova, “Dynamic impacts of
VI. CONCLUSION economic growth, energy use, urbanization, tourism,
This study presents an approach for determining the optimal agricultural value-added, and forested area on
placement and capacity of the BESS to minimize the system carbon dioxide emissions in Brazil,” J. Environ.
Stud. Sci., vol. 12, no. 4, pp. 794–814, 2022, doi:
costs and improve the efficiency of the distribution systems
10.1007/s13412-022-00782-w.
integrated with PV and EVs. The main objective function is
[2] M. S. Hassan, H. Mahmood, and A. Javaid, “The
to minimize system costs, including installation, impact of electric power consumption on economic
replacement, transmission loss, voltage regulation, and peak growth: a case study of Portugal, France, and
demand costs while ensuring optimal control by considering Finland,” Environ. Sci. Pollut. Res., vol. 29, no. 30,
the voltage of all buses, battery power, and energy battery as pp. 45204–45220, 2022, doi: 10.1007/s11356-022-
constraints. Three metaheuristic algorithms comprising PSO, 19097-y.

VOLUME XX, 2017 15

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

[3] M. Kongkuah, H. Yao, and V. Yilanci, “The [14] N. Jayasekara, M. A. S. Masoum, and P. J. Wolfs,
relationship between energy consumption, economic “Optimal operation of distributed energy storage
growth, and CO2 emissions in China: the role of systems to improve distribution network load and
urbanisation and international trade,” Environ. Dev. generation hosting capability,” IEEE Trans. Sustain.
Sustain., vol. 24, no. 4, pp. 4684–4708, 2022, doi: Energy, vol. 7, no. 1J, pp. 250–261, 2016, doi:
10.1007/s10668-021-01628-1. 10.1109/TSTE.2015.2487360.
[4] Q. Dai, J. Liu, and Q. Wei, “Optimal [15] A. Mazza, H. Mirtaheri, G. Chicco, A. Russo, and
photovoltaic/battery energy storage/electric vehicle M. Fantino, “Location and sizing of battery energy
charging station design based onmulti-agent particle storage units in low voltage distribution networks,”
swarm optimization algorithm,” Sustain., vol. 11, Energies, vol. 13, no. 1, pp. 2–5, 2019, doi:
no. 7, 2019, doi: 10.3390/su11071973. 10.3390/en13010052.
[5] M. Motalleb, E. Reihani, and R. Ghorbani, “Optimal [16] A. Ahmadian, M. Sedghi, and M. Aliakbar-Golkar,
placement and sizing of the storage supporting “Fuzzy load modeling of plug-in electric vehicles for
transmission and distribution networks,” Renew. optimal storage and dg planning in active
Energy, vol. 94, pp. 651–659, 2016, doi: distribution network,” IEEE Trans. Veh. Technol.,
10.1016/j.renene.2016.03.101. vol. 66, no. 5, pp. 3622–3631, 2017, doi:
[6] S. A. El-Batawy and W. G. Morsi, “Optimal design 10.1109/TVT.2016.2609038.
of community battery energy storage systems with [17] M. Khalid, U. Akram, and S. Shafiq, “Optimal
prosumers owning electric vehicles,” IEEE Trans. planning of multiple distributed generating units and
Ind. Informatics, vol. 14, no. 5, pp. 1920–1931, storage in active distribution networks,” IEEE
2018, doi: 10.1109/TII.2017.2752464. Access, vol. 6, pp. 55234–55244, 2018, doi:
[7] D. M. Tovilović and N. L. J. Rajaković, “The 10.1109/ACCESS.2018.2872788.
simultaneous impact of photovoltaic systems and [18] Y. Zheng, Y. Song, A. Huang, and D. J. Hill,
plug-in electric vehicles on the daily load and “Hierarchical Optimal Allocation of Battery Energy
voltage profiles and the harmonic voltage distortions Storage Systems for Multiple Services in
in urban distribution systems,” Renew. Energy, vol. Distribution Systems,” IEEE Trans. Sustain. Energy,
76, pp. 454–464, 2015, doi: vol. 11, no. 3, pp. 1911–1921, 2020, doi:
10.1016/j.renene.2014.11.065. 10.1109/TSTE.2019.2946371.
[8] O. Babacan, W. Torre, and J. Kleissl, “Siting and [19] S. Khunkitti, P. Boonluk, and A. Siritaratiwat,
sizing of distributed energy storage to mitigate “Optimal Location and Sizing of BESS for
voltage impact by solar PV in distribution systems,” Performance Improvement of Distribution Systems
Sol. Energy, vol. 146, pp. 199–208, 2017, doi: with High DG Penetration,” Int. Trans. Electr.
10.1016/j.solener.2017.02.047. Energy Syst., vol. 2022, pp. 1–16, 2022, doi:
[9] C. K. Das, O. Bass, G. Kothapalli, T. S. Mahmoud, 10.1155/2022/6361243.
and D. Habibi, “Overview of energy storage systems [20] V. Janamala and D. Sreenivasulu Reddy, “Coyote
in distribution networks: Placement, sizing, optimization algorithm for optimal allocation of
operation, and power quality,” Renew. Sustain. interline –Photovoltaic battery storage system in
Energy Rev., vol. 91, no. November 2016, pp. 1205– islanded electrical distribution network considering
1230, 2018, doi: 10.1016/j.rser.2018.03.068. EV load penetration,” J. Energy Storage, vol. 41, no.
[10] B. G. Desai, Electrical Energy Storage., vol. 1, no. July, 2021, doi: 10.1016/j.est.2021.102981.
3. 1981. doi: 10.1049/pbpo167e_ch7. [21] D. Sadeghi, A. Hesami Naghshbandy, and S.
[11] P. Boonluk, A. Siritaratiwat, P. Fuangfoo, and S. Bahramara, “Optimal sizing of hybrid renewable
Khunkitti, “Optimal siting and sizing of battery energy systems in presence of electric vehicles using
energy storage systems for distribution network of multi-objective particle swarm optimization,”
distribution system operators,” Batteries, vol. 6, no. Energy, vol. 209, p. 118471, 2020, doi:
4, pp. 1–16, 2020, doi: 10.3390/batteries6040056. 10.1016/j.energy.2020.118471.
[12] A. Fathy, “A novel artificial hummingbird algorithm [22] R. Eberhart and J. Kennedy, “A new optimizer using
for integrating renewable based biomass distributed particle swarm theory,” in MHS’95. Proceedings of
generators in radial distribution systems,” Appl. the Sixth International Symposium on Micro
Energy, vol. 323, no. July, p. 119605, 2022, doi: Machine and Human Science, 1995, pp. 39–43. doi:
10.1016/j.apenergy.2022.119605. 10.1109/MHS.1995.494215.
[13] M. A. Tolba, E. H. Houssein, A. A. Eisa, and F. A. [23] B. Abdollahzadeh, F. S. Gharehchopogh, and S.
Hashim, Optimizing the distributed generators Mirjalili, “African vultures optimization algorithm:
integration in electrical distribution networks: A new nature-inspired metaheuristic algorithm for
efficient modified forensic-based investigation, vol. global optimization problems,” Comput. Ind. Eng.,
35, no. 11. Springer London, 2022. doi: vol. 158, no. January, 2021, doi:
10.1007/s00521-022-08103-6. 10.1016/j.cie.2021.107408.

16 VOLUME XX, 2023

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

P.Natsawat et al.: Preparation of Papers for IEEE Access (February

2017)

[24] S. Mirjalili, A. H. Gandomi, S. Z. Mirjalili, S. AUPEC 2011, 2011.

Saremi, H. Faris, and S. M. Mirjalili, “Salp Swarm [31] N. Ghorbani, A. Kasaeian, A. Toopshekan, L.
Algorithm: A bio-inspired optimizer for engineering Bahrami, and A. Maghami, “Optimizing a hybrid
design problems,” Adv. Eng. Softw., vol. 114, pp. wind-PV-battery system using GA-PSO and
163–191, 2017, doi: MOPSO for reducing cost and increasing
10.1016/j.advengsoft.2017.07.002. reliability,” Energy, vol. 154, pp. 581–591, 2018,
[25] F. R. McLarnon and E. J. Cairns, “Energy Storage | doi: 10.1016/j.energy.2017.12.057.
Energy Storage,” Annu. Rev. 1nc. All rights Reserv., [32] L. Abualigah and M. Altalhi, “A novel generalized
vol. 14, no. February, pp. 241–271, 1989. normal distribution arithmetic optimization
[26] T. Kousksou, P. Bruel, A. Jamil, T. El Rhafiki, and algorithm for global optimization and data clustering
Y. Zeraouli, “Energy storage: Applications and problems,” J. Ambient Intell. Humaniz. Comput., no.
challenges,” Sol. Energy Mater. Sol. Cells, vol. 120, 0123456789, 2022, doi: 10.1007/s12652-022-
no. PART A, pp. 59–80, 2014, doi: 03898-7.
10.1016/j.solmat.2013.08.015. [33] X. Chen, K. Ding, J. Zhang, Z. Yang, Y. Liu, and H.
[27] M. Yekini Suberu, M. Wazir Mustafa, and N. Bashir, Yang, “A two-stage method for model parameter
“Energy storage systems for renewable energy identification based on the maximum power
power sector integration and mitigation of matching and improved flow direction algorithm,”
intermittency,” Renew. Sustain. Energy Rev., vol. Energy Convers. Manag., vol. 278, no. January, p.
35, pp. 499–514, 2014, doi: 116712, 2023, doi:
10.1016/j.rser.2014.04.009. 10.1016/j.enconman.2023.116712.
[28] M. Stecca, L. R. Elizondo, T. B. Soeiro, P. Bauer, [34] M. E. Baran and F. F. Wu, “Network reconfiguration
and P. Palensky, “A comprehensive review of the in distribution systems for loss reduction and load
integration of battery energy storage systems into balancing,” IEEE Power Eng. Rev., vol. 9, no. 4, pp.
distribution networks,” IEEE Open J. Ind. Electron. 101–102, 1989, doi: 10.1109/MPER.1989.4310642.
Soc., vol. 1, no. 1, pp. 46–65, 2020, doi: [35] M. E. Baran and F. F. Wu, “Optimal capacitor
10.1109/OJIES.2020.2981832. placement on radial distribution systems,” IEEE
[29] P. Boonluk, S. Khunkitti, P. Fuangfoo, and A. Trans. Power Deliv., vol. 4, no. 1, pp. 725–734,
Siritaratiwat, “Optimal siting and sizing of battery 1989, doi: 10.1109/61.19265.
energy storage: Case study seventh feeder at nakhon [36] S. Satyanarayana, T. Ramana, S. Sivanagaraju, and
phanom substation in Thailand,” Energies, vol. 14, G. K. Rao, “An efficient load flow solution for radial
no. 5, 2021, doi: 10.3390/en14051458. distribution network including voltage dependent
[30] P. Wolfs, N. Jayasekera, and S. Subawickrama, “A load models,” Electr. Power Components Syst., vol.
Fourier series based approach to the periodic 35, no. 5, pp. 539–551, 2007, doi:
optimisation of finely dispersed battery storage,” 10.1080/15325000601078179.
2011 21st Australas. Univ. Power Eng. Conf.

NATSAWAT POMPERN was born in SUTTICHAI

Phitsanulok, Thailand, on December 8, 1993. This PREMRUDEEPREECHACHARN
author became a Member of IEEE in 2022. He received the B.Eng. degree in electrical
received B.Eng. degrees in electrical engineering engineering from Chiang Mai University,
from King Mongkut’s Institute of Technology Thailand, in 1988, and the M.S. and Ph.D. degrees
Ladkrabang University (KMITL) in 2015. He is in electric power engineering from Rensselaer
currently an Electrical Engineer in the Provincial Polytechnic Institute, Troy, NY, USA, in 1992 and
Electricity Authority, Thailand. His research 1997, respectively. Currently, he is an Associate
interests include battery energy storage systems, Professor in the Department of Electrical
electric vehicles, optimization, metaheuristic Engineering, Chiang Mai University. His research
algorithms, temperature control, ventilation interests include renewable energy, power quality,
system, and programmable logic controller. high-quality utility interface, power electronics, and artificial intelligence
applied to power systems.

VOLUME XX, 2017 17

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
 This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2023.3291590

APIRAT SIRITARATIWAT was born in Maha

Sarakham Province, Thailand, on July 4, 1970.
He received the B. Eng. degree from Khon Kaen
University, Thailand, in 1992, and the Ph.D.
degree from the University of Manchester, UK,
in 1998, both in electrical engineering. Since
then, he has carried on with his research on GMR
films and magnetic recording heads at HDD,
Thailand. He is currently the Director of
Industry/University Cooperation Research
Center, HDD.

SIROTE KHUNKITTI was born in Khon Kaen,

Thailand, on January 30, 1993. He received
B.Eng. and Ph.D. degrees in electrical
engineering from Khon Kaen University (KKU)
in 2014 and 2018. He is currently an Assistant
Professor in the Department of Electrical
Engineering, Chiang Mai University. His
research interests include battery energy storage
systems, electric vehicles, power system
planning and optimization, and metaheuristic
algorithms.

18 VOLUME XX, 2023

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4