Allocation of EV Public Charging Station in

Renewable based Distribution Network using HHO

Considering Uncertainties and Tra c Congestion
Arnab Pal 
National Institute of Technology Agartala
Aniruddha Bhattacharya  (  bhatta.aniruddha@gmail.com )
National Institute of Technology Durgapur https://orcid.org/0000-0002-4059-1090
Ajoy Kumar Chakraborty 
National Institute of Technology Agartala

Research Article

Keywords: Distributed generation, EV charging station, HHO, optimal allocation, 2m PEM, uncertainty

Posted Date: May 7th, 2021

DOI: https://doi.org/10.21203/rs.3.rs-172433/v1

License:   This work is licensed under a Creative Commons Attribution 4.0 International License.  
Read Full License
 1

Allocation of EV Public Charging Station in

Renewable based Distribution Network using HHO
Considering Uncertainties and Traffic Congestion
Arnab Pal1, Aniruddha Bhattacharya2,*, Ajoy Kumar Chakraborty1
1
Department of Electrical Engineering, National Institute of Technology Agartala, India
2
Department of Electrical Engineering, National Institute of Technology Durgapur, India
arnabpal1994@gmail.com; bhatta.aniruddha@gmail.com; akcall58@gmail.com
* Corresponding author
Abstract—Electric vehicle (EV) is the growing vehicular The suitable locations are selected for CS in (Zhang et al.
technology for sustainable development to reduce carbon emission 2019a), based on service risk parameter using upgraded whale
and to save fossil fuel. The charging station (CS) is necessary at optimization technique. Authors in (Wang et al. 2013),
appropriate locations to facilitate the EV owners to charge their
allocated the CS in a distribution network overlapped with
vehicle as well as to keep the distribution system parameters
within permissible limits. Besides that, the selection of a charging traffic network where the objectives are minimization of the
station is also a significant task for the EV user to reduce battery power loss and voltage deviation. The fast-charging stations are
energy wastage while reaching the EV charging station. This paper allocated in a test system to minimize the various costs
presents a realistic solution for the allocation of public fast- including losses in (Sadeghi-Barzani et al. 2014). The total cost
charging stations (PFCS) along with solar distributed generation is minimized in (Liu et al. 2013) by optimal siting and sizing of
(SDG). A 33 node radial distribution network is superimposed
the CSs in IEEE 123 distribution system with reduced power
with the corresponding traffic network to allocate PFCSs and
SDGs. Two interconnected stages of optimization are used in this loss and improved voltage profile. In (Awasthi et al. 2017), the
work. The first part deals with the optimization of PFCS’s optimal locations are identified in the distribution network of
locations and SDG’s locations with sizes, to minimize the energy Allahabad city for electric vehicle charging station (EVCS) to
loss and to improve voltage profile using harris hawk optimization minimize the active power loss and development cost. Real
(HHO) and few other soft computing techniques. The second part power loss is minimized in (Pal et al. 2019; Ponnam and
handles the proper assignment of EVs to the PFCSs with less
Swarnasri 2020) to CS in radial distribution network using
consumption of the EV’s energy considering the road distances
with traffic congestion using linear programming (LP), where the several soft-computing techniques. In (Zhang et al. 2019b), the
shortest paths are decided by Dijkstra's algorithm. The 2m point cost for the CS is minimized with maximum service capability
estimation method (2m PEM) is employed to handle the by the optimal allocation in Beijing city. Optimal battery
uncertainties associated with EVs and SDGs. The robustness of swapping stations are planned in IEEE 15 bus distribution
solutions are tested using wilcoxon signed rank test and quade test. system in (Zheng et al. 2014). In (Davidov and Pantoš 2019),
the placement cost of the CS is minimized considering the
Index Terms—Distributed generation, EV charging station,
HHO, optimal allocation, 2m PEM, uncertainty.
power system reliability constraints. Authors in (Liu et al. 2020)
considered annual profit maximization to allocate DG and
I. INTRODUCTION EVCS in 33 and 69 distribution system. In (Shaaban et al. 2019;
Luo et al. 2020) CS and DG are optimally allocated in a micro-
Use of EV is the way to replace the conventional fossil fuel- grid and distribution network respectively considering the cost
based vehicle and it helps to reduce the carbon emission and minimization. Simultaneous allocation of CS and DG is
fossil fuel consumption (Wei et al. 2019). Nowadays, the proposed in (Atat et al. 2020) with charging coordination.
number of users is rapidly increasing who are motivated Location and capacity of CSs are optimized considering user
towards the EV. However, the availability of public charging satisfaction in (Yi et al. 2019). Charging stations’ locations are
station is one of the important hesitations to adopt EV. found out in (Gong et al. 2019) based on millage and
Therefore, the fast charging station is the greatest solution for distribution of EVs. CSs are allotted with different charging
the public charging infrastructure (Sadeghi-Barzani et al. 2014; level and various associated costs are minimized in (Liu and Bie
Motoaki 2019), which needs to be allocated optimally. The 2019). The uncertain driving pattern is taken during allocation
PFCS creates enormous load to the power system network of CS with minimum cost in (Andrade et al. 2020). The
which leads to huge power loss and makes the voltage profile locations of CS are found out in IEEE 69 distribution system
weaken. Therefore, appropriate allocation of PFCSs is very with optimal grid to vehicle (G2V) and vehicle to grid (V2G)
crucial to keep the losses minimum with a healthy voltage scheduling in (Hadian et al. 2020). (Xiong et al. 2018; Dong et
profile. Moreover, the improper selection of PFCS by the EV al. 2019; Hosseini and Sarder 2019; Kong et al. 2019) provide
owners can cause extra energy consumption of batteries to good solutions to allocate the CS in geographical map and road
reach at CS, which affects the energy demand from the power network. In (Spieker et al. 2017), multi-objective is solved
system.
using a modified genetic algorithm to serve maximum number network with power distribution system and traffic network, 3)
of EV by placing EVCS. The optimal charging facilities are Zone wise allocation of PFCS to make it distributed in the area,
selected in (Wu et al. 2020) considering uncertainties. 4) Simultaneous allocation of SDGs to reduce the energy loss
Uncertain vehicle rent is taken care for maximum profit of and to improve voltage profile, 5) Proper assignment of EVs at
suppliers in (Long et al. 2019). PFCSs considering the shortest distance with traffic
Research work on charging station allocation are still limited. congestions, 6) Consideration of all possible uncertainties
The major shortfalls from most of the available articles are, associated with EV, SDG and traffic congestion using 2m PEM.
either CSs are allocated in power system network or in road The statistical hypothesis tests i.e. wilcoxon signed rank test
network. Moreover, a very few tackled the uncertainties of EV and quade test, are performed to check robustness of the
flow and state of charge (SOC) requirement. The EVs need to solutions and to confirm the hypothesis.
be assigned at right CS by using minimum battery energy to The rest of this article is arranged as follows. Section II
travel up to CS, to reduce the energy demand from the power shows the modelling and formulation for both the stages with
system. As per best knowledge of the authors, the apt selection constraints. Section III presents the solution methodologies to
of the CS by the EVs considering road congestion is not solve the problem, i.e. HHO, 2m PEM and algorithm of the
considered in the existing literatures while dealing with PFCS entire procedure. The results and discussion are demonstrated
allocation. However, the CSs also should be distributed in the in section IV. This paper is concluded in section V with a
area to serve all the EV users effectively. Furthermore, the EV’s roundup and future possibility.
motivation regarding air pollution reduction would not be
satisfied if the EV takes power, generated by conventional II. MATHEMATICAL FORMULATION
power plants, during charging process. Therefore, the power The entire problem is formulated in two stages interconnected
system should be renewable supported installed with optimal with each other. In the first stage, the energy loss is minimized
sizes and locations. Again, the EV flow, PV output and loads by optimal allocation of PFCSs and SDGs. In the second stage,
are having hourly variations. Therefore, power loss the EVs reach to apt PFCS by spending minimum energy. The
minimization as considered in most of the papers is not proper SOC arrival and first trip distance in the first stage depend on
for a sustainable solution because energy loss minimization is the second stage after assigning the EV at apt PFCS.
more justified in a dynamic environment.
In this present work, the energy loss minimization is adopted A. Stage I: Allocation of PFCS and SDG
as objective to decide the optimal locations of PFCS and The main objective in this stage is to keep the energy loss
locations along with sizes of SDG in the distribution network. minimum by allocating the PFCS at optimal places. Moreover,
The solar-based DGs are incorporated to reduce the energy loss the optimal locations and sizes of the SDG need to find out
and to improve the voltages with less carbon emission. based on the same objective to reduce more loss. The objective
A superimposed network with a distribution system and road function, constraints, load modelling and energy requirements
map is taken as the study area. Three zones are considered in are discussed below.
the study area to allocate the PFCSs in distributed pattern which 1) Objective Function-1 (𝐹1 )
2 (𝑡) 2 (𝑡)
will help to serve all users evenly. The solution methodology is 𝑇 𝑁𝑏𝑟 𝑃𝑏𝑟 + 𝑄𝑏𝑟
𝐹1 = 𝑚𝑖𝑛 {∑ ∑ 𝑅𝑏𝑟 ∙ 2 } (1)
proposed in two interconnected stages. In the first stage, the 𝑡=1 𝑏𝑟=1 𝑉𝑏𝑟
PFCSs are allocated and SDGs are located with their optimal where 𝑃𝑏𝑟 (𝑡) and 𝑄𝑏𝑟 (𝑡) are real and reactive power flow
sizes simultaneously. The second stage works to assign the EVs through 𝑏𝑟 𝑡ℎ branch respectively at 𝑡 𝑡ℎ time, 𝑉𝑏𝑟 is voltage of
at apt PFCS and helps to calculate the load and energy demand the sending node of 𝑏𝑟 𝑡ℎ branch, 𝑅𝑏𝑟 is resistance of branch
at respective PFCSs. HHO and Grey wolf optimizer (GWO) are 𝑏𝑟. 𝑇 and 𝑁𝑏𝑟 are total time interval and total number of branch
used in first stage and the energy loss is minimized. Finally, respectively.
eight other optimization techniques are also used to validate the 2) Constraints
solutions. In the second stage, the integer linear programming The voltage of every node at any time interval should be
(ILP) is clubbed to find out the suitable PFCS for the EVs to within permissible limits as presented in (2), where 𝑉𝑚𝑖𝑛 and
minimize their energy requirement. The traffic congestion and 𝑉𝑚𝑎𝑥 are the lower and upper limits respectively.
distance which affect the SOC of the EV to reach up to the 𝑉𝑚𝑖𝑛 ≤ 𝑉𝑏𝑟 (𝑡) ≤ 𝑉𝑚𝑎𝑥 , ∀ 𝑏𝑟, 𝑡 (2)
PFCS, both are taken into account to choose the apt CS for the The current flowing through every branch (𝐼𝑏𝑟 ), should be
EVs with minimum energy requirement. The shortest routes are less than the allowable upper limit (𝐼𝑚𝑎𝑥 ).
found out by the Dijkstra's algorithm. The annual average 24 |𝐼𝑏𝑟 | ≤ 𝐼𝑚𝑎𝑥 , ∀ 𝑏𝑟 (3)
hours’ variations of EV flow, PV profile and conventional load The total installed capacity of the SDGs should not be more
profile are taken in this work. Moreover, instead of fixed than the DG penetration level (𝜆𝑝 ) as follows.
random values, the uncertainties related to SOC requirements, 𝑁𝑑𝑔
∑ 𝐼𝐶𝑑𝑔 ≤ 𝜆𝑝 ∙ 𝑙𝑜𝑎𝑑𝑝 (4)
EV locations, traffic congestions and solar irradiance, are 𝑑𝑔=1
handled by a powerful statistical tool, i.e. 2m PEM. where 𝑙𝑜𝑎𝑑𝑝 is the peak load of the system and 𝐼𝐶𝑑𝑔 is the
In brief, the key contributions of the paper are: 1) Energy loss installed capacity of the 𝑑𝑔𝑡ℎ SDG.
minimization instead of power loss minimization considering The power generated from SDGs (𝑃𝑠𝑑𝑔 ) and power supplied
dynamic event, 2) Placement of PFCSs in a superimposed from the utility (𝑃𝑢𝑡 ) should be equal to total load (𝐿 𝑇 ) and
 3

total loss (𝐿𝑜𝑠𝑠𝑇 ) at any hour 𝑡 𝑡ℎ . Therefore, the load 𝑑𝑖𝑓 (ℎ) 𝜁𝑐 (𝑔, ℎ)
𝐴 𝑠𝑜𝑐 (ℎ) = 1 − − , where, 𝐴𝐸𝑅
balancing constraint is written as: 𝐴𝐸𝑅(ℎ) 𝐵𝑐(ℎ) (13)
𝑃𝑠𝑑𝑔 (𝑡) ∪ 𝑃𝑢𝑡 (𝑡) ∈ 𝐿 𝑇 (𝑡) ∪ 𝐿𝑜𝑠𝑠𝑇 (𝑡), ∀ 𝑡 (5) ∈ 𝑓𝑛(ℎ)
Branch thermal limit constraint is provided bellow. 𝐷 𝑠𝑜𝑐 (ℎ) = (𝑆𝑇𝐷(ℎ)⁄𝐴𝐸𝑅(ℎ)) + 0.25 (14)
|𝑆𝑏𝑟 | ≤ |𝑆𝑏𝑟𝑚𝑎𝑥
|, ∀ 𝑏𝑟 (6) where 𝜁𝑐 (𝑔, ℎ) is consumed energy by ℎ𝑡ℎ EV to reach 𝑔𝑡ℎ
where 𝑆𝑏𝑟 is apparent power of branch 𝑏𝑟 and 𝑆𝑏𝑟 𝑚𝑎𝑥
is the PFCS considering traffic congestion, which is calculated as (31)
maximum apparent power limit of 𝑏𝑟 𝑡ℎ branch. in the second stage. 𝑑𝑖𝑓 (ℎ) is the initial first trip distance,
Zone constraint is taken in consideration to make the public 𝐵𝑐(ℎ) is battery capacity, 𝐴𝐸𝑅(ℎ) is all-electric range and
charging station (𝑔) distributed over the city. This will bound 𝑆𝑇𝐷(ℎ) is the subsequent trip distance of ℎ𝑡ℎ vehicle which is
one PFCS in each zone. obtained as (Mehta et al. 2018), by
𝑔𝑖 ∈ ℤ𝑖 , 𝑖 ∈ 𝑛𝑜. 𝑜𝑓 𝑧𝑜𝑛𝑒 (7) 𝑆𝑇𝐷(ℎ) = 𝑑𝑚(ℎ) − 𝑑𝑓𝑓 (ℎ) (15)
where 𝑆𝑖 is the 𝑖 𝑡ℎ EVCS and ℤ𝑖 is the 𝑖 𝑡ℎ zone. where 𝑑𝑓𝑓 (ℎ) is the final first trip distance, which is obtained
The maximum SOC level (𝑆𝑂𝐶𝑚𝑎𝑥 ) and minimum SOC level in the second stage as (30). 𝑑𝑚(ℎ) is daily mileage of ℎ𝑡ℎ
(𝑆𝑂𝐶𝑚𝑖𝑛 ) should be maintained to keep the battery healthy. vehicle in km. Departure time (𝐷𝑡𝑖𝑚𝑒 (ℎ)) of ℎ𝑡ℎ vehicle is
𝑆𝑂𝐶𝑚𝑖𝑛 ≤ 𝑆𝑂𝐶ℎ ≤ 𝑆𝑂𝐶𝑚𝑎𝑥 , ∀ ℎ (8) found out using
where 𝑆𝑂𝐶ℎ is the SOC level of the ℎ𝑡ℎ EV. 𝐷𝑡𝑖𝑚𝑒 (ℎ) = 𝐴𝑡𝑖𝑚𝑒 (ℎ) + 𝑅𝑡𝑖𝑚𝑒 (ℎ) (16)
𝑡𝑖𝑚𝑒
3) Load Modelling where 𝑅 (ℎ) is the required time to attain the desire charge
The total load at 𝑏𝑢𝑡ℎ bus (𝐿(𝑏𝑢, 𝑡)) at any 𝑡 𝑡ℎ time interval level of 𝐸𝑉ℎ , 𝐴𝑡𝑖𝑚𝑒 (ℎ) is the arrival time of ℎ𝑡ℎ vehicle at
is written as follows: PFCS. Required time for charging of 𝐸𝑉ℎ with 𝐶𝑟 charging rate
𝐿(𝑏𝑢, 𝑡) = 𝐶𝐿(𝑏𝑢, 𝑡) + 𝐶𝑆𝐿(𝑏𝑢, 𝑡) is calculated as:
(9)
− 𝑆𝐷𝐺(𝑏𝑢, 𝑡), ∀ 𝑏𝑢, 𝑡 𝑅𝑡𝑖𝑚𝑒 (ℎ) = 𝑅𝑒𝑛𝑔 (ℎ)⁄𝐶𝑟 (17)
where 𝐶𝑆𝐿(𝑏𝑢, 𝑡) is the load due to PFCS at bus 𝑏𝑢, if it is 5) SDG modelling
connected to 𝑏𝑢𝑡ℎ bus, otherwise zero. On the other hand, The main stochastic variable for PV is solar irradiance
𝑆𝐷𝐺(𝑏𝑢, 𝑡) is the generation from PV at bus 𝑏𝑢 , if it is (𝑘𝑊/𝑚2 ). PV cell temperature is also uncertain but it depends
connected to 𝑏𝑢𝑡ℎ bus, otherwise zero. on irradiance only. PV power output at 𝑡 𝑡ℎ hour (𝑃𝑤𝑝𝑣 (𝑡))
The load at 𝑔𝑡ℎ CS at 𝑡 𝑡ℎ time (𝐶𝑆𝐿(𝑔, 𝑡)), which may be will be added in (9) as a negative load and is expressed as
connected to any 𝑏𝑢𝑡ℎ bus, is calculated as follows. (Sultana and Roy 2015):
𝐸𝑉
𝐶𝑆𝐿(𝑔, 𝑡) = ∑ 𝐸𝑉𝐿(𝑔, ℎ, 𝑡), ∀ 𝑡 (10) 𝑆𝐷𝐺(𝑡) = 𝑃𝑤𝑝𝑣 (𝑡) = 𝑃𝑣𝑣𝑎𝑟𝑖 (𝑡) × 𝑃𝑤𝑝𝑣 , ∀ 𝑡 (18)
ℎ=1
𝑡ℎ 𝑡ℎ
where 𝑃𝑣𝑣𝑎𝑟𝑖 is the 24 hours variation of solar irradiance in p.u.
where 𝐸𝑉𝐿(𝑔, ℎ, 𝑡) is the load due to 𝐸𝑉ℎ at 𝑔 CS at 𝑡 hour is followed as (Sultana and Roy 2015). Therefore, at the
when 𝐴𝑡𝑖𝑚𝑒 (𝑔, ℎ) < 𝑡 < 𝐷 𝑡𝑖𝑚𝑒 (𝑔, ℎ) . 𝐴𝑡𝑖𝑚𝑒 (𝑔, ℎ) and endpoint PV power (𝑃𝑤𝑝𝑣 ) is calculated from solar irradiance
𝐷𝑡𝑖𝑚𝑒 (𝑔, ℎ) are the arrival time and departure time respectively (Soroudi et al. 2012) by
of the 𝐸𝑉ℎ at 𝐶𝑆𝑔 . 𝑃𝑤𝑝𝑣 = 𝑁𝑝 ∙ 𝐹𝑓 ∙ 𝑉𝑖𝑟 ∙ 𝐼𝑖𝑟 (19)
4) Energy Requirement of EVs where 𝑁𝑝 is total number of PV module, 𝐹𝑓 is the fill factor is
Arrival time, first trip distance and daily mileage are the
presented as:
uncertain variables for all the EV. Arrival SOC, departure SOC 𝑉𝑚𝑝 ∙ 𝐼𝑚𝑝
and departure time of EV are also uncertain which are 𝐹𝑓 = (20)
calculated based on pre-mentioned uncertainties. Energy 𝑉𝑜𝑐 ∙ 𝐼𝑠𝑐
requirement by an EV depends on the arrival SOC (𝐴𝑠𝑜𝑐 (ℎ)) 𝑉𝑖𝑟 = 𝑉𝑜𝑐 − 𝐶𝑣 × 𝑡𝑐𝑒𝑙𝑙 (21)
and departure SOC(𝐷 𝑠𝑜𝑐 (ℎ)) of the ℎ𝑡ℎ vehicle. The arrival 𝐼𝑖𝑟 = 𝑖𝑟𝑎𝑣𝑔 {𝐼𝑠𝑐 + 𝐶𝑖 (𝑡𝑐𝑒𝑙𝑙 − 25)} (22)
SOC is related to final first trip distance (𝑑𝑓𝑓 (ℎ)) in km, 𝑡𝑐𝑒𝑙𝑙 = 𝑡𝑎𝑚𝑏 + (𝑖𝑟𝑎𝑣𝑔 ∙ (𝑡𝑛𝑜𝑚 − 20)⁄0.8) (23)
whereas departure SOC is associated with subsequent trip where 𝑉𝑚𝑝 and 𝐼𝑚𝑝 are maximum power point voltage and
distance (𝑆𝑇𝐷(ℎ)) in km. Energy requirement (kWh) current respectively. 𝑉𝑜𝑐 and 𝐼𝑠𝑐 are open circuit voltage and
(𝑅𝑒𝑛𝑔 (ℎ)) by ℎ𝑡ℎ EV is evaluated as follows (Mehta et al. short circuit current. 𝑖𝑟𝑎𝑣𝑔 is average solar irradiance, 𝑡𝑐𝑒𝑙𝑙 is
2018). the cell temperature in °C. 𝐶𝑣 and 𝐶𝑖 are voltage and current
𝑅 𝑒𝑛𝑔 (ℎ) = (𝑅 𝑠𝑜𝑐 (ℎ) ∙ 𝐵𝑐(ℎ))⁄𝜂𝑐ℎ , ∀ ℎ, where 𝐵𝑐 temperature coefficient. 𝑡𝑛𝑜𝑚 is nominal operating temperature
(11)
∈ 𝑓𝑛(ℎ) of PV cell. 𝑡𝑎𝑚𝑏 is ambient temperature. The solar irradiance
𝑠𝑜𝑐 (ℎ)
where 𝑅 and 𝐵𝑐(ℎ) indicate required SOC and battery follows beta distribution (Soroudi et al. 2012), which is
capacity of ℎ𝑡ℎ vehicle respectively, 𝜂𝑐ℎ is charging efficiency. demonstrated as follows.
𝑅𝑒𝑞 𝑠𝑜𝑐 is calculated in percentage (Mehta et al. 2018) as 𝑓(𝑖𝑟)
Γ(𝛼𝑖𝑟 + 𝛽𝑖𝑟 )
follows. × 𝑖𝑟 (𝛼𝑖𝑟 −1) × (1 − 𝑖𝑟)(𝛽𝑖𝑟−1) ,
𝑅 𝑠𝑜𝑐
(ℎ) = Γ(𝛼𝑖𝑟 )Γ(𝛽𝑖𝑟 ) (30)
1 − 𝐴𝑠𝑜𝑐 (ℎ), 𝑖𝑓 𝐷 𝑠𝑜𝑐 (ℎ) > 1 𝑖𝑓 0 ≤ 𝑖𝑟 ≤ 1; 𝛼𝑖𝑟 ≥ 0; 𝛽𝑖𝑟 ≥ 0
𝑠𝑜𝑐 𝑠𝑜𝑐
= {(𝐷 (ℎ) − 𝐴 (ℎ), 𝑖𝑓 𝐴𝑠𝑜𝑐 (ℎ) < (𝐷 𝑠𝑜𝑐 (ℎ) < 1 (12) { 0, 𝑒𝑙𝑠𝑒
0, 𝑖𝑓 𝐴𝑣𝑠𝑜𝑐 = (𝐷 𝑠𝑜𝑐 (ℎ) where 𝛼𝑖𝑟 and 𝛽𝑖𝑟 are beta shape parameters. 𝜇𝑖𝑟 and 𝜎𝑖𝑟 are
Arrival SOC (Rezaee et al. 2013) and departure SOC (Mehta the mean and SD of solar irradiance, which is determined as:
et al. 2018) are calculated as (13) and (14) respectively. 𝜇𝑖𝑟 = 𝛼𝑖𝑟 ⁄(𝛼𝑖𝑟 + 𝛽𝑖𝑟 ) (24)
 2 (1
Again, from the equation of Greenshields model (Shlayan et
𝜎𝑖𝑟 = √𝜇𝑖𝑟 + 𝜇𝑖𝑟 )⁄(𝛼𝑖𝑟 + 𝜇𝑖𝑟 ) (25)
al. 2018), the modified velocity of the EV due to traffic
congestion is presented as:
B. Stage 2: PFCS Selection by EVs
𝓋𝑚 (𝑔, ℎ, 𝑡) = 𝑣𝑓 (𝑔, ℎ) (1 − 𝑦(𝑔, ℎ, 𝑡) ∙ 𝜏(𝑔, ℎ, 𝑡)
The main goal is to assign each EV to right PFCS by (32)
consuming less quantity of energy while reaching its suggested ∙ ℂ𝑗 (𝑔, ℎ) ∕ 𝕕(𝑔, ℎ))
PFCS. This is determined using integer linear programming where 𝑣𝑓 (𝑔, ℎ) is the free flow velocity and the required time
considering congestions and shortest distances. The shortest 𝜏(𝑔, ℎ, 𝑡) for covering the distance (𝕕(𝑔, ℎ)) with vehicle flow
paths and distances may be identified using Dijkstra's algorithm
(Amaliah et al. 2016). The formulation for optimal section of 𝑦(𝑔, ℎ, 𝑡) by the EV is expressed as (Mkahl et al. 2017):
PFCS by the ILP is discussed below. 𝑣𝑓 (𝑔, ℎ) ∙ 𝑦(𝑔, ℎ, 𝑡)
𝜏(𝑔, ℎ, 𝑡) = 𝑣𝑓 (𝑔, ℎ)/2 ∙ (33)
1) Assignment Matrix for ILP ℂ𝑗 (𝑔, ℎ) ∙ 𝑑(𝑔, ℎ)
The linear programming problem (ILPP) is formulated by an
assignment matrix 𝜓(𝑔, ℎ) (Das et al. 2020) where ℎ >> 𝑔 III. SOLUTION METHODOLOGY
and each EV is assigned to a PFCS. This matrix is grounded on The HHO technique, 2m PEM and the algorithm for this
the renowned assignment problem (Aktel et al. 2017) with the proposed work, are presented below.
combination of traffic flow structure (Schrieber et al. 2017).
Since, traffic flow is a dynamic event, hence, it is expressed as A. Harris Hawks Optimization (HHO)
𝜓(𝑔, ℎ, 𝑡), where, t ∈ 𝑡𝑖𝑚𝑒. The matrix is shown in the Table 1. In 2019, A.A. Heidari et al. introduced the HHO (Heidari et
TABLE 1
al. 2019). This optimization is based on chasing tactic of Harris’
STRUCTURE OF EV ASSIGNMENT MATRIX TO THE PFCS
𝐸𝑉1 𝐸𝑉2 𝐸𝑉3 … 𝐸𝑉ℎ hawks to its target. The key steps are instructed as follows
𝐶𝑆1 𝜓(1,1, 𝑡) 𝜓(1,2, 𝑡) 𝜓(1, 3, 𝑡) … 𝜓(1, ℎ, 𝑡) (Heidari et al. 2019).
𝐶𝑆2 𝜓(2,1, 𝑡) 𝜓(2,2, 𝑡) 𝜓(2,3, 𝑡) … 𝜓(2, ℎ, 𝑡) 1) Initialization
⋮ ⋮ ⋮ ⋮ ⋮ ⋮ Decision variables (𝐾ij ) , preliminary energy (𝐸𝑛0 ) and
𝐶𝑆𝑔 𝐴(𝑔, 1, 𝑡) 𝜓(𝑔, 2, 𝑡) 𝜓(𝑔, 3, 𝑡) … 𝜓(𝑔, ℎ, 𝑡)
jump strength (𝐽𝑠) are initialized as follows.
2) Objective Function-2 (𝐹2 ) 𝐾𝑖𝑗 = 𝐾𝑗𝑚𝑖𝑛 + 𝑟𝑎𝑛𝑑(𝐾𝑗𝑚𝑎𝑥 − 𝐾𝑗𝑚𝑖𝑛 ), 𝑤ℎ𝑒𝑟𝑒 𝑖
(34)
In this formulation, the main objective is to keep the battery = 1,2,3 … 𝑆𝑝 ; j = 1,2,3 … 𝐷𝑛
energy consumption minimum by the EV to reach at PFCS 𝐸𝑛0 = 2 ∙ 𝑟𝑎𝑛𝑑(−1, 1) − 1 (35)
considering traffic congestions at time 𝑡, i.e. 𝐽𝑠 = 2(1 − 𝑟𝑎𝑛𝑑(0, 1)) (36)
𝐶𝑆 𝐸𝑉
2) Update part
𝐹2 = 𝑚𝑖𝑛 {∑ ∑(𝜁𝑐 (𝑔, ℎ, 𝑡) ∙ 𝜒(𝑔, ℎ, 𝑡))} (26) The energy (𝐸𝑛) is updated in every iteration (𝑡) as:
𝑔=1 ℎ=1 𝑡
𝜒(𝑔, ℎ, 𝑡) 𝐸𝑛 = 2 ∙ 𝐸𝑛0 (1 − ) (37)
𝑇
1, if 𝐸𝑉ℎ 𝑖s assigned to 𝐶𝑆𝑔 at time 𝑡 (27) where number of iteration is 𝑇. Decision vector (𝐾 (𝑍+1) ) is
={
0, otherwise updated according to various situations as follows.
3) Constraints
𝑐𝑠  Exploration phase (if |En|≥1)
∑ 𝜒(𝑔, ℎ, 𝑡) = 1, ∀ 𝐸𝑉ℎ 𝐾 (𝑍+1)
(28) (𝑍) (𝑍)
𝑔=1 𝐾𝑟𝑎𝑛𝑑 − 𝑚1 |𝐾𝑟𝑎𝑛𝑑 − 2𝑚2 𝐾 (𝑍) | , 𝑖𝑓 𝑞 ≥ 0.5 (38)
={
(𝑔, ℎ) ≤ 𝑑𝑓𝑓 (ℎ) (29) (𝑍)
𝐾𝑟𝑎𝑏𝑏𝑖𝑡 −
(𝑍)
𝐾𝑚 − 𝑚3 (𝐿𝐿 + 𝑚4 (𝑈𝐿 − 𝐿𝐿)), 𝑖𝑓 𝑞 < 0
𝑑𝑓𝑓 (ℎ) ∈ 𝑑𝑖𝑓 (ℎ) ∪ 𝕕(𝑔, ℎ) (30) (𝑍)
where 𝑍 is iteration number, 𝐾𝑟𝑎𝑏𝑏𝑖𝑡 is the position of the
where in (28), 1 ≤ ℎ ≤ 𝐸𝑉 and each EV is assigned to only one (𝑍)
PFCS at a time. In (29), the final first trip distance should not rabbit. 𝐾𝑟𝑎𝑛𝑑 is randomly selected hawk position.
be the lower than the distance of 𝐸𝑉ℎ from 𝐶𝑆𝑔 charging 𝑚1 , 𝑚2 , 𝑚3 , 𝑚4 and 𝑞 is random number ∈ [0, 1]. 𝑈𝐿 is upper
stations 𝕕(𝑔, ℎ). limit and 𝐿𝐿 is lower limit of the decision variables.
4) Energy modelling of EVs  Exploitation phase (𝑖𝑓 |𝐸𝑛| < 1)
In order to model the energy consumption by the EV The different conditions in this phase as follows, where 𝑟 is
considering traffic congestions and distance, various factors the possibility of a rabbit can run away.
such as the distance between the location of the vehicle and the  Soft besiege (𝑖𝑓 𝑟 ≥ 0.5 𝑎𝑛𝑑 |𝐸𝑛| ≥ 0.5)
(𝑍)
charging station, vehicles speed considering congestion 𝐾 (𝑍+1) = ∆𝐾 (𝑍) − 𝐸|𝐽𝑠 ∙ 𝐾𝑟𝑎𝑏𝑏𝑖𝑡 − 𝐾 (𝑍) | (39)
(𝓋𝑚 (𝑔, ℎ, 𝑡)) , all electric range (AER) of the individual  Hard besiege (𝑖𝑓 𝑟 ≥ 0.5 𝑎𝑛𝑑 |𝐸𝑛| < 0.5)
vehicles, battery capacity of the EV, vehicles flow (𝑦(𝑔, ℎ, 𝑡)) (𝑍)
𝐾 (𝑍+1) = 𝐾𝑟𝑎𝑏𝑏𝑖𝑡 − 𝐸|∆𝐾 (𝑍) | (40)
and jam coefficient (ℂ𝑗 (𝑔, ℎ)) are considered [25]. The  Soft besiege with progressive rapid dives (𝑖𝑓 𝑟 <
consumed energy by 𝐸𝑉ℎ considering the vehicles flow and 0.5 𝑎𝑛𝑑 |𝐸𝑛| ≥ 0.5)
traffic congestions it is written as (Mkahl 2015): 𝐴, 𝑖𝑓 𝐹(𝐴) < 𝐹(𝐾 (𝑍) )
𝐵𝑐(ℎ) ∙ 𝜏(𝑔, ℎ, 𝑡) ∙ 𝓋𝑚 (𝑔, ℎ, 𝑡) 𝐾 (𝑍+1) = { (41)
𝜁𝑐 (𝑔, ℎ) = (31) 𝐵, 𝑖𝑓 𝐹(𝐵) < 𝐹(𝐾 (𝑍) )
𝐴𝐸𝑅 ∙ ℂ𝑗 (𝑔, ℎ)
 5

 Hard besiege with progressive rapid dives (𝑖𝑓 𝑟 < is calculated using
𝑁𝑂
0.5 𝑎𝑛𝑑 |𝐸𝑛| < 0.5) 𝑘 𝑘
𝑋, 𝑖𝑓 𝐹(𝐴) < 𝐹(𝐾 (𝑍) ) 𝔼 [(𝑅𝑙 − 𝜇𝑅𝑙 ) ] = ∑(𝑅𝑙,𝑗 − 𝜇𝑅𝑙 ) ∙ ℙ(𝑅𝑙,𝑗 ),
(53)
𝐾 (𝑍+1) = { (42) 𝑗=1
𝑌, 𝑖𝑓 𝐹(𝐵) < 𝐹(𝐾 (𝑍) ) 𝑘 ∈ 3,4
Detail information about the variables used in above steps are where 𝑅𝑙,𝑗 is the value produced from distribution function at
available in (Heidari et al. 2019). 𝑗𝑡ℎ observation, 𝑁𝑂 is the number of inspection of 𝑅𝑙 . ℙ(𝑅𝑙,𝑗 )
B. Hong’s 2m-Point estimation Method signifies the probability density of 𝑅𝑙,𝑗 . 𝑅𝑙 is generated by (54)
for normal distribution.
Hong’s 2m PEM (Hong 1998) is applied in this work which
𝑅𝑙 = 𝜇𝑢𝑣 + (𝕫 × 𝜎𝑢𝑣 ), 𝑙 ∈ 1,2,3, … , 𝑚 (54)
is an effective statistical technique to handle multiple
where 𝜇𝑢𝑣 and 𝜎𝑢𝑣 is the mean and SD of the uncertain variable
uncertainties. The objective functions (𝐹1 , 𝐹2 ) are modified to
𝑢𝑣 respectively, 𝕫 is the standard normal random value. The
mean values (𝜇𝐹1 , 𝜇𝐹2 ) according to 2m PEM as follows. probability density of 𝑅𝑙 is obtainable by
𝑀𝑖𝑛(𝐹) = 𝑀𝑖𝑛(𝜇𝐹), 𝑤ℎ𝑒𝑟𝑒 𝐹 ∈ 𝐹1 𝑜𝑟 𝐹2 (43) 2 2
−[(𝑅𝑙 −𝜇𝑅𝑙 ) ⁄2(𝜎𝑅𝑙 ) ]
𝜇𝐹 = 𝐹(1) (44) ℙ(𝑅𝑙 ) = (𝑒 ) /√2𝜋𝜎𝑅𝑙 2 (55)
Standard deviation (SD) of objective function is calculated by
𝜎𝐹 = √(𝐹(2) − (F(1))2 ) (45) C. Algorithm of Proposed Work
where 𝐹(1) is the first moment and 𝐹(2) is the and second Start
moment of the objective function (𝐹), which are obtained as Initialization
follows. 1. Initialize the decision variables (PFCS locations, SDG locations,
𝑚 2 SDG sizes).
ℎ Objective Function Calculation
𝐹(ℎ) = ∑ ∑ (𝑤𝑙,𝑝𝑜 (𝐹𝑙,𝑝𝑜 ) ) , ℎ ∈ 1,2 (46)
2. Set time interval, 𝑡 = 0.
𝑙=1 𝑝𝑜=1
(−1)𝑝𝑜 𝜉𝑅𝑙,(3−𝑝𝑜) 3. Find out the shortest path and distance to the PFCSs from the
𝑤𝑙,𝑝𝑜 = ∙ , 𝑝𝑜 ∈ 1,2 (47) EVs at 𝑡 𝑡ℎ time interval using Dijkstra's algorithm
𝑚 𝜉𝑅𝑙 ,1 − 𝜉𝑅𝑙 ,2 4. Find out the energy consumption by the EVs to reach PFCSs
𝐹𝑙,𝑝𝑜 = 𝑓(𝐷𝑣 , 𝜇𝑅1 , 𝜇𝑅2 , … , 𝑧𝑙,𝑝𝑜 , … , 𝜇𝑅𝑚 ), considering congestions using (31).
(48)
Under 2m PEM

𝑝𝑜 ∈ 1,2; 𝑙 ∈ 1,2,3, … , 𝑚 5. Assign the EVs to apt PFCS by consuming the least energy
𝑡ℎ using ILP.
where 𝐹(ℎ) is the ℎ moment of 𝐹 . 𝑝𝑜 ∈ 1, 2 generate two
6. Calculate the final energy requirement by the EVs using (11).
weighting factors (𝑤𝑙,𝑝𝑜 ) of 𝑧𝑙 . 𝑚 represents the number of 7. Calculate the total load for every PFCS at 𝑡 𝑡ℎ time interval.
uncertain variable. 𝜉𝑅𝑙 is the standard location of 𝑅𝑙 . 𝐷𝑣 is 8. Find the SDG generations at 𝑡 𝑡ℎ time interval.
deterministic variable, 𝜇𝑅𝑙 is the mean of the 𝑙 𝑡ℎ uncertain 9. Run the backward forward sweep load for to calculate power
loss at 𝑡 𝑡ℎ time interval.
variable. 𝑧𝑙,𝑝𝑜 is definite positions of the input variable. 10. 𝑡 = 𝑡 + 1.
2 × 𝑚 number of sets are created using (48). The objective 11. Go to next step if 𝑡 = 𝑇, otherwise go to step 3.
functions 𝐹1 and 𝐹2 are calculated for all stochastic set clubbed 12. Calculate the energy loss.
with a fixed deterministic set. In this present work, the sizes of 13. Calculate the mean and SD of the objective function
the SDG and locations of the PFCS and SDG are the Update
14. Update the decision variables using optimization techniques.
deterministic variables (𝐷). The PEM sets are expressed as:
15. Go to next step if optimization is converged, otherwise go to
𝑓(𝐷𝑣 , 𝜇𝑅1 , 𝜇𝑅2 , … , 𝑧𝑙,𝑝𝑜 , … , 𝜇𝑅𝑚 ) step 2.
𝐷1 𝐷2 … 𝐷𝑛 𝑧1,1 𝜇𝑅2 𝜇𝑅3 … 𝜇𝑅𝑚 16. Display the optimized results.
𝐷1 𝐷2 … 𝐷𝑛 𝑧1,1 𝜇𝑅2 𝜇𝑅2 … 𝜇𝑅𝑚 end
𝐷1 𝐷2 … 𝐷𝑛 𝜇𝑅1 𝑧2,1 𝜇𝑅3 … 𝜇𝑅𝑚
= 𝐷1 𝐷2 … 𝐷𝑛 𝜇𝑅1 𝑧2,2 𝜇𝑅3 … 𝜇𝑅𝑚 (49) IV. SIMULATION STUDY
⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮
𝐷1 𝐷2 … 𝐷𝑛 𝜇𝑅1 𝜇𝑅2 𝜇𝑅3 … 𝑧𝑚,1 A. Input Data
[⏟𝐷1 𝐷2 … 𝐷𝑛 ⏟ 𝜇𝑅1 𝜇𝑅2 𝜇𝑅3 … 𝑧𝑚,2 ]
𝐷𝑒𝑡𝑒𝑟𝑚𝑖𝑛𝑖𝑠𝑡𝑖𝑐𝑠 𝑆𝑡𝑜𝑐ℎ𝑎𝑠𝑡𝑖𝑐 𝑃𝑎𝑟𝑡
𝑃𝑎𝑟𝑡
where 𝑛 is the number of deterministic variable.
𝑧𝑙,𝑝𝑜 = 𝜇𝑅𝑙 + (𝜉𝑅𝑙 ,𝑝𝑜 × 𝜎𝑅𝑙 ), 𝑝𝑜 = 1,2 (50)
where 𝜎𝑅𝑙 is the standard deviation of the uncertain variable.
𝜆𝑅𝑙 ,3
𝜉𝑅𝑙 ,𝑝𝑜 = + (−1)3−𝑝𝑜 √𝜆𝑅𝑙 ,4 − 3𝜆2𝑅𝑙 ,3 ⁄4 ,
2 (51)
𝑝𝑜 ∈ 1,2
where 𝜉𝑅𝑙 ,𝑝𝑜 is the 𝑝𝑜 𝑡ℎ standard location of 𝑅𝑙 . 𝜆𝑅𝑙 ,3 and 𝜆𝑅𝑙 ,4
are the coefficient of third central moment//skewness and
coefficient of fourth central moment/kurtosis of 𝑅𝑙 Fig. 1. Superimposed study area with power distribution and traffic network.
respectively.
𝑘 𝑘
𝜆𝑅𝑙,𝑘 = 𝔼 [(𝑅𝑙 − 𝜇𝑅𝑙 ) ]⁄(𝜎𝑅𝑙 ) , 𝑘 ∈ 3,4 (52) Figure 1 shows the area, where a 33 node distribution
𝑘 network (Amiri et al. 2018) is overlapped with a traffic
where 𝔼 is the notation for expected value and 𝔼 [(𝑅𝑙 − 𝜇𝑅𝑙 ) ]
network. Zone wise allocation of PFCS is necessary for well 2014; Bodo et al. 2017). The arrival SOC is limited to 25 – 90%
accessibility of charging station to all EV users in a city. (Shafie-khah et al. 2016; Pal et al. 2021).
Therefore, three zones are identified based on the vehicle TABLE 4
ELECTRIC VEHICLE TYPES AND SPECIFICATIONS (MEHTA ET AL. 2018)
density caused by market place, housing area, land accessibility Type Battery AER Presence (%)
and road arrangement. of EV Capacity (kWh) (km) in the Area
Three different types of roads are taken here, where the T1 13.8 48.27 60
busiest road is type 1, next to it is the type 2 and the least busy T2 18.4 64.36 30
T3 24 117 10
roads are type 3. Based on one year’s vehicle behaviours, it is
observed that the peak vehicle flow varies 500-1000 vehicle/hr Different types of vehicle are present in a city with different
for different sections of the roads and figure 2 shows the hourly battery capacities. Hence, three types of EV are taken in this
profile of vehicle flow of the three types of road. Table 2 present work and shown in Table 4. Total 200 EVs are considered in
the jam coefficients (Mkahl et al. 2017) for different types of the proposed study area.
the road.
B. Simulation Case Study
In this work, the simulations are performed in MATLAB
coding platform in a computer with Intel Core i5, 8 GB ram.
Three case studies are conducted to check the performances of
the proposed allocation strategy considering the uncertainties
by 2m PEM. The case studies are defined as follows:
1) Case-1: In this study, the PFCSs are allocated without the
presence of SDG and the EVs are not assigned at apt PFCSs
using ILP, rather it is coming to CS whichever is nearest by
Dijkstra's algorithm.
Fig. 2. Annual average vehicle flow profile of three types of road.
2) Case-2: This case is also for PFCS allocation without SDG
TABLE 2 but EVs are assigned at proper PFCSs considering the traffic
ELECTRIC VEHICLE TYPES AND SPECIFICATION congestions and distances by ILP and Dijkstra's algorithm.
Road Type 1 Road Type 2 Road Type 3 3) Case-3: In this study, the optimized sizes and locations of
Jam Coefficient 250 200 150 SDG are achieved along with PFCS locations, where apt choice
of PFCSs by the EVs is also taken into account.
The distribution, mean value, standard deviation and Table 5 and Table 6 present the solutions for the case-1 and
maximum, minimum values of different variables associated case-2 respectively for 100% EV penetration, where the optimal
with EVs are presented in Table 3 and the uncertain data related locations of PFCS using GWO and HHO are shown with
to solar PV are taken as (Soroudi et al. 2012). Annual average
minimized means of the energy loss and obtained standard
PV profile and conventional load profile are shown in Figure 3
deviations.
as (Pal et al. 2020). TABLE 5
TABLE 3
ALLOCATED LOCATIONS OF PFCS AND ENERGY LOSS FOR CASE-1 WITH 100%
DISTRIBUTION DETAILS FOR UNCERTAIN PROPERTIES OF EV
EV PENETRATION
Variable Unit Distribution Mean Std. dev. Min Max
Optimization PFCS Location Mean Energy Loss SD Energy loss
Truncated Gaussian Technique (Node) (kWh) (kWh)
Arrival time hour (Shafie-khah et al. 12 5 4 24
GWO 2 10 28 4940.9903 26.5687
2016)
HHO 2 10 28 4940.9903 26.5687
Normal (Mehta et al.
Daily mileage km 55 10 0 -
2018)
First trip Normal (Mehta et al. TABLE 6
km 19.25 7.95 0 - ALLOCATED LOCATIONS OF PFCS AND ENERGY LOSS FOR CASE-2 WITH 100%
distance 2018)
EV PENETRATION
Optimization PFCS Location Mean Energy Loss SD Energy loss
Technique (Node) (kWh) (kWh)
GWO 2 16 28 4336.6315 24.3021
HHO 2 16 28 4336.6315 24.3021

In table 5, the optimal locations of PFCS for case-1 are 2, 10

and 28, and Table 6 shows the locations are 2, 16 and 28 for
case-2. It is observed from Table 5 and Table 6 that one PFCS’s
location is changed from node 10 to 16 in case-2 and the energy
loss is reduced compared to case-1. In case-1, the locations are
Fig. 3. Annual average vehicle flow profile of three types of road (Pal et al.
2020). closest to the substation in each zone to keep the system loss
minimum. As, in this case, the EVs are assigned to nearest
In this work, level 2 (fast) charging facility is considered with PFCS only and their energy loss due to travel was not the
22 kW (Awasthi et al. 2017; Collin et al. 2019; Andrade et al. concern. Therefore, the PFCS locations are only motivated to
2020). Charging efficiency is considered 95% (Mohamed et al. reduce the power system energy loss, not the energy demand by
 7

the EVs. However, in case-2, the energy requirement by the the allocation of PFCSs at apt locations considering proper
EVs to travel up to PFCSs are also taken care, which helps to assigned EVs to the PFCSs by ILP. For case-3, the voltage
reduce the SOC requirement of the EVs. Hence, for the profile is far improved due to the allocations of SDGs.
locations of node 2, 16 and 28, the energy loss is less due to less
C. Sensitivity Analysis
energy demand by the EVs compared to case-1.
The allocation results of PFCSs and SDGs are shown in The same three cases are performed with 50% EV
Table 7 with 100% EV penetration. In this case, HHO obtains penetration and Table 8 shows the respective results. It shows
less energy loss by allocating SDG at proper locations with apt that the locations of the PFCS and SDG with 50% EV
sizes w.r.t. GWO. It is seen from Table 6 and Table 7 that one penetration are same as locations in case of 100% EV
location of PFCS is changed from node 28 to 31 in case-3 for penetration for all the three cases, which confirm the robustness
the simultaneous allocation of PFCSs and SDGs. Significant of the solutions. Also, both the optimization techniques are
energy loss reduction is achieved in case-3 due to the optimal reached to the same solutions for all the cases with 50% EV
presence of SDGs. penetration scenario. Like 100% EV penetration, the energy
TABLE 7 loss is more for case-1, and reduced in case-2. Case-3 offers the
LOCATIONS OF PFCS AND LOCATIONS WITH SIZES OF SDG FOR CASE-3 WITH best among all with least energy loss. However, the energy
100% EV PENETRATION losses are less compared to 100% EV penetration.
Optimization Technique GWO HHO
TABLE 8
PFCS Location (Node) 2 16 31 2 16 31 SOLUTIONS FOR CASE-1 TO CASE-3 WITH 50% EV PENETRATION
SDG Location (Node) 3 14 30 24 14 30
Optimization Technique GWO HHO
SDG size (MW) 2.2354 1.1075 1.3128 1.4046 1.1417 1.3904 PFCS Location (Node) 2 10 28 2 10 28
Mean Energy Loss (kWh) 2961.0982 2918.5838 Case-1 Mean Energy Loss (kWh) 4573.6032 4573.6032
SD Energy loss (kWh) 28.4861 28.8568 SD Energy loss (kWh) 18.2569 18.2569
PFCS Location (Node) 2 16 28 2 16 28
Again it is noticed from Table 5-7, that both the optimization Case-2 Mean Energy Loss (kWh) 4190.3649 4190.3649
techniques provide the same solution for case-1 as well as for SD Energy loss (kWh) 17.0458 17. 0458
PFCS Location (Node) 2 16 31 2 16 31
case-2. But in case-3, HHO performs better than GWO. Figure SDG Location (Node) 24 14 30 24 14 30
4 depicts the convergence curves for all the cases with 100% Case-3 SDG size (MW) 1.396 1.054 1.371 1.396 1.054 1.371
EV penetration. It is observed in Figure 4 that the HHO Mean Energy Loss (kWh) 2894.4277 2894.4277
converged before the GWO. Moreover, HHO provides better SD Energy loss (kWh) 25.4753 25.4753
loss minimization in case-3 compared to GWO.
Figure 6 shows the energy loss comparisons between 100%
and 50% EV penetration and it is visualized that the energy
losses are less for all cases in 50% EV penetration scenario
compare to 100% penetration because of less number of EV
creates comparatively less load and less energy demand in the
system. Energy losses are also decreasing from case-1 towards
case-3 for both the EV penetration level due to case wise
development, i.e. apt assignment of EV at PFCSs and allocation
of SDGs at proper nodes with suitable sizes.

Fig. 4. Convergence curves of GWO and HHO for case-1 to 3 with 100% EV
penetration.

Fig. 6. Comparison of all the cases with 100% and 50% EV penetration.

Table 9 presents the sensitivity studies on different key items.

Fig. 5. Daily average voltage profiles of the distribution network for case-1 to It is noticed, when SDGs sizes are reduced (study-2), the energy
3 with 100% EV penetration.
loss is increased significantly. High EV penetration (study-3)
Figure 5 illustrates the daily average voltage profiles of the and low charging efficiency (study-4) both create extra load on
distribution network for all the cases with 100% EV the power system, which leads to increment of energy loss. The
penetration. It is seen in Figure 5, the voltage profile is worst energy loss is not affected by the increment of 20% traffic flow
among all for case-1, because of the allocation of PFCSs with (study-5), whereas 50% increment increases energy loss (study-
mismanagement of EVs appointment to the PFCSs. The voltage 6), because traffic flow creates congestions which burn more
profile of case-2 is comparatively better than case-1, which is energy of the EV batteries and increase the energy demand from
the system. Increased jam coefficient defines, the jam will be GWO 2 16 31
3 14 30
2961.0982 28.4861
occurred with more number of vehicle. Therefore, increased 2.235 1.107 1.312
24 14 30
jam coefficient reduces the congestions and consequently HHO 2 16 31
1.404 1.141 1.390
2918.5838 28.8568
energy loss is also reduced (study-7).
TABLE 9
SENSITIVITY ANALYSIS ON THE KEY FACTORS FOR CASE-3
In Table 10, the solution given by HBC is the lowest loss
Mean Energy Change from minimization among all, whereas PSO and HGSO perform
Study Test Item
Loss (kWh) study-1 (%) same as HHO and minimized the remarkable energy loss
1 Base Case-3 with 100% EV and 100% DG 2918.5838 - compared to other techniques. Therefore, it can be decided that
2 Decrease SDG sizes by 50% 3254.1446 ▲11.49
3 Increase EV penetration to 120% 2958.3145 ▲1.36 the location 2, 16 and 31 for PFCS and 24, 14 and 30 for SDGs
4 Decrease charging efficiency to 80% 2921.1453 ▲0.08 with sizes 1.4, 1.1 and 1.3 MW respectively, are the optimized
5 Increase traffic flow by 20% 2918.5838 ▲0 solution to achieve minimum energy loss for the proposed
6 Increase traffic flow by 50% 2922.4563 ▲0.13 problem.
7 Increase jam coefficient by 20% 2917.1569 ▼0.04
▲ – Increased, ▼ – Decreased
Figure 7 depicts the daily average voltage profiles of the
different sensitivity studies. It is seen in case of voltage also,
reduction of SDG’s sizes (study-2) drops the voltage profile
remarkably. Voltage profiles are not significantly deviated from
the base case (study-1) in case of study-3 & 4. Whereas, voltage
profile is not affected in study 5 – 7 and it is same as study-1.

Fig. 8. Allocation of PFCSs and SDGs.

Figure 8 shows the graphical allocation of PFCSs and SDGs,

where it is easily noticeable that the PFCSs are scattered in the
area which will offer well accessibility to all EV users with less
travel distance.
Fig. 7. Daily average voltage profiles of different sensitivity studies.
E. Statistical Hypothesis Test
D. Validation of the Solution
The statistical tests, wilcoxon signed rank test (Liu et al.
The actual targeted problem of this proposed work which is 2017) and quade test (Liu et al. 2017) are performed to check
case-3 with 100% EV penetration, is solved with eight other robustness of the solutions of the targeted problem and to
optimization techniques and the solutions are presented in confirm the hypothesis. Both the test are performed for 30
Table 10. The selected optimization techniques are Honey Bee sample with 95% confidence interval. The absolute values are
Colony (HBC), Backtracking Search Optimization Algorithm 137 (from alpha distribution table) for wilcoxon signed rank test
(BSA), Differential Evolution (DE), Biogeography Based and 2.73 (from 𝑓 distribution table) for quade test.
Optimization (BBO), Symbiotic Organisms Search (SOS), 1) Wilcoxon Signed Rank Test
Genetic Algorithm (GA), Particle Swarm Optimization (PSO) The wilcoxon signed rank test is executed with 95%
and Henry Gas Solubility Optimization (HGSO). confidence considering 30 samples and the results are shown in
TABLE 10
SOLUTIONS OF CASE-3 WITH DIFFERENT OPTIMIZATION TECHNIQUES FOR
Table 11. It is seen that the test values are higher than the
100% EV PENETRATION absolute value (137), for every optimization technique, which
Optimization PFCS SDG Location Mean Energy SD Energy accept the null hypothesis (H0) and confirm the constancy and
Technique Locations and Size (MW) Loss (kWh) Loss (kWh) robustness of the results.
6 15 31
HBC 2 16 31 2968.1383 30.2487 TABLE 11
1.561 0.941 0.905 WILCOXON TEST RESULT
3 14 30 HBC BSA DE BBO SOS GA PSO HGSO GWO HHO
BSA 2 16 31 2961.0982 29.1257
2.235 1.107 1.312 Test
3 14 30 191 182 198 185 171 208 196 183 202 210
DE 2 16 31 2961.0982 29.1257 Value
2.236 1.106 1.313
3 14 30
H0?          
BBO 2 16 31 2961.0982 29.1257
2.235 1.107 1.313
3 14 30 2) Quade Test
SOS 2 16 31 2961.0982 29.1257
2.235 1.107 1.312 Quade Test is performed with same confidence interval and
3 14 30 same sample size. Table 12 presents the test value considering
GA 2 16 31 2961.0981 29.1257
2.236 1.107 1.312
24 14 30 10 optimization techniques. H0 is rejected here for the higher
PSO 2 16 31 2918.5838 28.8568 statistic value than the absolute value (2.73), which confirms
1.404 1.141 1.390
HGSO 2 16 31
24 14 30
2918.5838 28.8568 significant better performance of HHO, PSO and HGSO.
1.404 1.141 1.390
 9

TABLE 12 hybrid optimization algorithm. Energy 133:70–78.

QUADE TEST RESULT https://doi.org/10.1016/j.energy.2017.05.094
No. Optimization Bodo N, Levi E, Subotic I, et al (2017) Efficiency Evaluation of Fully
No. of Scenario Alpha Test Value H0?
Technique Integrated On-Board EV Battery Chargers With Nine-Phase
30 10 0.05 55.83  Machines. IEEE Trans Energy Convers 32:257–266.
https://doi.org/10.1109/TEC.2016.2606657
Collin R, Miao Y, Yokochi A, et al (2019) Advanced Electric Vehicle Fast-
V. CONCLUSION Charging Technologies. Energies 12:1839.
https://doi.org/10.3390/en12101839
Lack of public charging station is the key obstacle for large Das S, Acharjee P, Bhattacharya A (2020) Charging Scheduling of Electric
scale acceptance of EVs. This article proposed a realistic Vehicle incorporating Grid-to-Vehicle (G2V) and Vehicle-to-Grid
solution for PFCS placement in a distribution network (V2G) technology considering in Smart-Grid. IEEE Transactions on
Industry Applications. https://doi.org/10.1109/TIA.2020.3041808
considering the traffic network. The locations are found out Davidov S, Pantoš M (2019) Optimization model for charging infrastructure
where the energy loss is minimum. The SDGs are also allocated planning with electric power system reliability check. Energy
for further reduction of energy loss and to improve the voltages. 166:886–894. https://doi.org/10.1016/j.energy.2018.10.150
The EVs are assigned at apt PFCSs by ILP with Dijkstra's Dong G, Ma J, Wei R, Haycox J (2019) Electric vehicle charging point
placement optimisation by exploiting spatial statistics and maximal
algorithm considering traffic congestions and shortest coverage location models. Transportation Research Part D:
distances. The apt assignment of EVs reduces the energy loss Transport and Environment 67:77–88.
indirectly. The locations and sizes of SDG are as important as https://doi.org/10.1016/j.trd.2018.11.005
Gong D, Tang M, Buchmeister B, Zhang H (2019) Solving Location Problem
PFCS locations in terms of energy loss minimization. Total 10 for Electric Vehicle Charging Stations—A Sharing Charging
optimization techniques are used to validate the final allocation Model. IEEE Access 7:138391–138402.
result. The statistical tests show the sturdiness of the result. For https://doi.org/10.1109/ACCESS.2019.2943079
Hadian E, Akbari H, Farzinfar M, Saeed S (2020) Optimal Allocation of
the future work, battery energy storages can be connected to Electric Vehicle Charging Stations With Adopted Smart
schedule the power generated from SDGs. The required Charging/Discharging Schedule. IEEE Access 8:196908–196919.
expansion of the network can also be planned. https://doi.org/10.1109/ACCESS.2020.3033662
Heidari AA, Mirjalili S, Faris H, et al (2019) Harris hawks optimization:
Algorithm and applications. Future Generation Computer Systems
DECLARATIONS 97:849–872. https://doi.org/10.1016/j.future.2019.02.028
Hong HP (1998) An efficient point estimate method for probabilistic analysis.
Funding: There is no funding for this work. Reliability Engineering & System Safety 59:261–267.
Conflicts of interest/Competing interests: The authors https://doi.org/10.1016/S0951-8320(97)00071-9
declare that they have no conflict of interest. Hosseini S, Sarder M (2019) Development of a Bayesian network model for
optimal site selection of electric vehicle charging station.
Availability of data and material: Data are available with the International Journal of Electrical Power & Energy Systems
authors. 105:110–122. https://doi.org/10.1016/j.ijepes.2018.08.011
Kong W, Luo Y, Feng G, et al (2019) Optimal location planning method of fast
Code availability: Codes are available with the authors.
charging station for electric vehicles considering operators, drivers,
Authors' contributions: Arnab Pal: Conceptualization, vehicles, traffic flow and power grid. Energy 186:115826.
Methodology, Software, Writing-original draft. Aniruddha https://doi.org/10.1016/j.energy.2019.07.156
Bhattacharya: Conceptualization, Methodology, Supervision, Liu L, Zhang Y, Da C, et al (2020) Optimal allocation of distributed generation
and electric vehicle charging stations based on intelligent algorithm
Writing-review & editing. Ajoy Kumar Chakraborty: and bi‐level programming. Int Trans Electr Energ Syst.
Conceptualization, Methodology, Supervision, Writing-review https://doi.org/10.1002/2050-7038.12366
& editing. Liu X, Bie Z (2019) Optimal Allocation Planning for Public EV Charging
Station Considering AC and DC Integrated Chargers. Energy
Procedia 159:382–387.
REFERENCES https://doi.org/10.1016/j.egypro.2018.12.072
Aktel A, Yagmahan B, Özcan T, et al (2017) The comparison of the Liu Z, Blasch E, John V (2017) Statistical comparison of image fusion
metaheuristic algorithms performances on airport gate assignment algorithms: Recommendations. Information Fusion 36:251–260.
problem. Transportation Research Procedia 22:469–478. https://doi.org/10.1016/j.inffus.2016.12.007
https://doi.org/10.1016/j.trpro.2017.03.061 Liu Z, Wen F, Ledwich G (2013) Optimal Planning of Electric-Vehicle
Amaliah B, Fatichah C, Riptianingdyah O (2016) FINDING THE SHORTEST Charging Stations in Distribution Systems. IEEE Trans Power
PATHS AMONG CITIES IN JAVA ISLAND USING NODE Delivery 28:102–110.
COMBINATION BASED ON DIJKSTRA ALGORITHM. https://doi.org/10.1109/TPWRD.2012.2223489
International Journal on Smart Sensing and Intelligent Systems Long Y, Xu D, Li X (2019) Channel coordination of battery supplier and battery
9:2219–2236. https://doi.org/10.21307/ijssis-2017-961 swap station of micro-grid with uncertain rental demand. Soft
Amiri SS, Jadid S, Saboori H (2018) Multi-objective optimum charging Comput 23:9733–9745. https://doi.org/10.1007/s00500-018-3542-x
management of electric vehicles through battery swapping stations. Luo L, Wu Z, Gu W, et al (2020) Coordinated allocation of distributed
Energy 165:549–562. https://doi.org/10.1016/j.energy.2018.09.167 generation resources and electric vehicle charging stations in
Andrade J, Ochoa LF, Freitas W (2020) Regional-scale allocation of fast distribution systems with vehicle-to-grid interaction. Energy
charging stations: travel times and distribution system 192:116631. https://doi.org/10.1016/j.energy.2019.116631
reinforcements. IET Generation, Transmission & Distribution Mehta R, Srinivasan D, Khambadkone AM, et al (2018) Smart Charging
14:4225–4233. https://doi.org/10.1049/iet-gtd.2019.1786 Strategies for Optimal Integration of Plug-In Electric Vehicles
Atat R, Ismail M, Serpedin E, Overbye T (2020) Dynamic Joint Allocation of Within Existing Distribution System Infrastructure. IEEE Trans
EV Charging Stations and DGs in Spatio-Temporal Expanding Smart Grid 9:299–312. https://doi.org/10.1109/TSG.2016.2550559
Grids. IEEE Access 8:7280–7294. Mkahl R (2015) Contribution to the modeling, dimensioning and management
https://doi.org/10.1109/ACCESS.2019.2963860 of the energy flows of an electric vehicle charging system: study of
Awasthi A, Venkitusamy K, Padmanaban S, et al (2017) Optimal planning of the interconnection with the electric network. Ph.D.dissertation,
electric vehicle charging station at the distribution system using Univ.Belfort-Montbéliard
Mkahl R, Nait-Sidi-Moh A, Gaber J, Wack M (2017) An optimal solution for
charging management of electric vehicles fleets. Electric Power
 Systems Research 146:177–188. Macroscopic Traffic Model. IEEE Photonics J 10:1–11.
https://doi.org/10.1016/j.epsr.2016.11.008 https://doi.org/10.1109/JPHOT.2017.2782801
Mohamed A, Salehi V, Ma T, Mohammed O (2014) Real-Time Energy Soroudi A, Aien M, Ehsan M (2012) A Probabilistic Modeling of Photo Voltaic
Management Algorithm for Plug-In Hybrid Electric Vehicle Modules and Wind Power Generation Impact on Distribution
Charging Parks Involving Sustainable Energy. IEEE Trans Sustain Networks. IEEE Systems Journal 6:254–259.
Energy 5:577–586. https://doi.org/10.1109/TSTE.2013.2278544 https://doi.org/10.1109/JSYST.2011.2162994
Motoaki Y (2019) Location-Allocation of Electric Vehicle Fast Chargers— Spieker H, Hagg A, Gaier A, et al (2017) Multi-stage evolution of single- and
Research and Practice. WEVJ 10:12. multi-objective MCLP: Successive placement of charging stations.
https://doi.org/10.3390/wevj10010012 Soft Comput 21:4859–4872. https://doi.org/10.1007/s00500-016-
Pal A, Bhattacharya A, Chakraborty AK (2019) Allocation of EV Fast Charging 2374-9
Station with V2G Facility in Distribution Network. In: 2019 8th Sultana S, Roy PK (2015) Oppositional krill herd algorithm for optimal location
International Conference on Power Systems (ICPS). IEEE, Jaipur, of distributed generator in radial distribution system. International
India, pp 1–6 Journal of Electrical Power & Energy Systems 73:182–191.
Pal A, Bhattacharya A, Chakraborty AK (2021) Allocation of electric vehicle https://doi.org/10.1016/j.ijepes.2015.04.021
charging station considering uncertainties. Sustainable Energy, Wang G, Xu Z, Wen F, Wong KP (2013) Traffic-Constrained Multiobjective
Grids and Networks 25:100422. Planning of Electric-Vehicle Charging Stations. IEEE Trans Power
https://doi.org/10.1016/j.segan.2020.100422 Delivery 28:2363–2372.
Pal A, Chakraborty AK, Bhowmik AR (2020) Optimal Placement and Sizing https://doi.org/10.1109/TPWRD.2013.2269142
of DG considering Power and Energy Loss Minimization in Wei W, Wu D, Wu Q, et al (2019) Interdependence between transportation
Distribution System. ijeei 12:624–653. system and power distribution system: a comprehensive review on
https://doi.org/10.15676/ijeei.2020.12.3.12 models and applications. J Mod Power Syst Clean Energy 7:433–
Ponnam VKB, Swarnasri K (2020) Multi-Objective Optimal Allocation of 448. https://doi.org/10.1007/s40565-019-0516-7
Electric Vehicle Charging Stations and Distributed Generators in Wu Y, Xu C, Huang Y, Li X (2020) Green supplier selection of electric vehicle
Radial Distribution Systems using Metaheuristic Optimization charging based on Choquet integral and type-2 fuzzy uncertainty.
Algorithms. Eng Technol Appl Sci Res 10:5837–5844. Soft Comput 24:3781–3795. https://doi.org/10.1007/s00500-019-
https://doi.org/10.48084/etasr.3517 04147-4
Rezaee S, Farjah E, Khorramdel B (2013) Probabilistic Analysis of Plug-In Xiong Y, Gan J, An B, et al (2018) Optimal Electric Vehicle Fast Charging
Electric Vehicles Impact on Electrical Grid Through Homes and Station Placement Based on Game Theoretical Framework. IEEE
Parking Lots. IEEE Trans Sustain Energy 4:1024–1033. Trans Intell Transport Syst 19:2493–2504.
https://doi.org/10.1109/TSTE.2013.2264498 https://doi.org/10.1109/TITS.2017.2754382
Sadeghi-Barzani P, Rajabi-Ghahnavieh A, Kazemi-Karegar H (2014) Optimal Yi T, Cheng X, Zheng H, Liu J (2019) Research on Location and Capacity
fast charging station placing and sizing. Applied Energy 125:289– Optimization Method for Electric Vehicle Charging Stations
299. https://doi.org/10.1016/j.apenergy.2014.03.077 Considering User’s Comprehensive Satisfaction. Energies 12:1915.
Schrieber J, Schuhmacher D, Gottschlich C (2017) DOTmark – A Benchmark https://doi.org/10.3390/en12101915
for Discrete Optimal Transport. IEEE Access 5:271–282. Zhang H, Tang L, Yang C, Lan S (2019a) Locating electric vehicle charging
https://doi.org/10.1109/ACCESS.2016.2639065 stations with service capacity using the improved whale
Shaaban MF, Mohamed S, Ismail M, et al (2019) Joint Planning of Smart EV optimization algorithm. Advanced Engineering Informatics
Charging Stations and DGs in Eco-Friendly Remote Hybrid 41:100901. https://doi.org/10.1016/j.aei.2019.02.006
Microgrids. IEEE Trans Smart Grid 10:5819–5830. Zhang Y, Zhang Q, Farnoosh A, et al (2019b) GIS-Based Multi-Objective
https://doi.org/10.1109/TSG.2019.2891900 Particle Swarm Optimization of charging stations for electric
Shafie-khah M, Heydarian-Forushani E, Osorio GJ, et al (2016) Optimal vehicles. Energy 169:844–853.
Behavior of Electric Vehicle Parking Lots as Demand Response https://doi.org/10.1016/j.energy.2018.12.062
Aggregation Agents. IEEE Trans Smart Grid 7:2654–2665. Zheng Y, Dong ZY, Xu Y, et al (2014) Electric Vehicle Battery Charging/Swap
https://doi.org/10.1109/TSG.2015.2496796 Stations in Distribution Systems: Comparison Study and Optimal
Shlayan N, Challapali K, Cavalcanti D, et al (2018) A Novel Illuminance Planning. IEEE Trans Power Syst 29:221–229.
Control Strategy for Roadway Lighting Based on Greenshields https://doi.org/10.1109/TPWRS.2013.2278852
Figures

Figure 1

Superimposed study area with power distribution and tra c network.

Figure 2

Annual average vehicle ow pro le of three types of road.

Figure 3

Annual average vehicle ow pro le of three types of road (Pal et al. 2020).
Figure 4

Convergence curves of GWO and HHO for case-1 to 3 with 100% EV penetration

Figure 5

Daily average voltage pro les of the distribution network for case-1 to 3 with 100% EV penetration.
Figure 6

Comparison of all the cases with 100% and 50% EV penetration.

Figure 7

Daily average voltage pro les of different sensitivity studies.

Figure 8

Allocation of PFCSs and SDGs.