Energy 245 (2022) 123180

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

An optimal energy management among the electric vehicle charging

stations and electricity distribution system using GPC-RERNN
approach
B. Rajani a, *, Bapayya Naidu Kommula b
a
Department of Electrical and Electronics Engineering, Aditya College of Engineering & Technology, Surampalem, Andhra Pradesh, 533437, India
b
Department of Electrical and Electronics Engineering, Aditya Engineering College, Surampalem, Andhra Pradesh, India

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

Article history: This paper proposes a hybrid strategy to manage the energy in electric vehicle charging station (EVCS)
Received 26 February 2021 and distribution system (DS). The proposed hybrid approach is the joint implementation of Giza Pyra-
Received in revised form mids Construction (GPC) and recalling-enhanced recurrent neural network (RERNN) hence it is named as
5 July 2021
GPC-RERNN. The main purpose is to give maximum amount of energy as restrictions are irregular and
Accepted 10 January 2022
Available online 11 January 2022
volatile nature of renewable energy sources, stochastic nature of EV, and local meteorological conditions.
Likewise, minimum system cost which incorporates land cost, station equipment, operating and pres-
ervation cost and also reduces the voltage deviation and power loss on distribution system. At first,
Keywords:
Energy management system
RERNN is used to originate the quality-of-service constrained decision form for EVCSs. The utility of DS is
Electric vehicle increase when scheduling the charging plans of EVCSs. To analyze energy interaction, GPC approach is
Electric vehicle charging stations (EVCS) used. Here, every EVCS is considered as leader and DS is considered as follower. It is used to establish, an
Cost optimization issue by equilibrium restrictions. The proposed approach analysis the bidirectional trading
Photovoltaic (PV) system of energy, effect of PV uncertainty under EMS operation, cost analysis based on selling energy. Finally the
Giza pyramids construction performance of the proposed approach is performed with the MATLAB/Simulink working platform and
RERNN likened with several existing approaches.
© 2022 Published by Elsevier Ltd.

1. Introduction [5,6].
EV is one of the new eco-friendly modes of transport, for
As global economic growth and natural resources dwindle, the reducing the emission; it acts as an important mode [7]. The EV has
energy crisis and environmental degradation have become serious become the hub of the trend for new power automobiles based on
issues for the stable improvement of the world today [1]. Therefore, continuous improvement of battery technology [8]. The enlarge-
to use energy effectively countries around the world use different ment of EVs is a great way to address concerns associated to fossil
approaches. Based on the Environmental Performance Index, many resource depletion as well as environmental pollution, which will
countries are producing highest CO2 emissions [2]. Both the growth promote sustainable urban growth [9]. The base of the industry
rate of energy use and CO2 emissions in transportation systems are growth for the electric vehicle is the Electric Vehicle Charging
higher than average, which has forced the world to seek more Station (EVCS) as power supplier of electric vehicle [10,11]. Profi-
efficient routes in the transportation sector [3]. Because of the cient, suitable and economical EVCS is increase the growth of in-
economic cost and environmental importance, the electric vehicle dustrial and also motivates the consumers to buy with greater
(EV) has attracted much attention as an alternative to internal desire [12].
combustion engine vehicles (ICEVs) [4]. Many government sup- The entire life cycle of EVCS, the important function is service
ports ongoing research on various features of EV application to quality and operational efficiency and also the site selection [13].
provide in-depth insights into the policy class on electric motion Therefore, the optimal EVCS site should be determined using the
correct decision-making method [14]. The major purpose of the
energy management of EV is meet present requirements without
* Corresponding author.
conciliation ability to meet the energy needs of future generations
E-mail address: [email protected] (B. Rajani). [15]. By creating renewable resources and creating cleaner and

https://doi.org/10.1016/j.energy.2022.123180
0360-5442/© 2022 Published by Elsevier Ltd.
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

more proficient technologies, this can ultimately result in more that energy on electric vehicle charging station (EVCS) and distri-
efficient and more secure use of it [16]. The joining of distributed bution system (DS). The proposed hybrid approach is the joined
generation like PV, wind etc. To the EVCS can make the EV with execution of both the Giza Pyramids Construction (GPC) and
high efficient manner [17]. However, the charging requirements of recalling-enhanced recurrent neural network (RERNN) thus it is
EVs influence the spatial and temporal distribution of electricity known as GPC-RERNN approach. The main objective of the pro-
charges [18]. Thus, it is essential to perform effectual energy posed approach is minimize the system cost, which incorporates
management in EVs, EVCS and Grid [19,20]. land cost, station equipment, operating and preservation cost and
With the improvement of PHEV technologies, various studies on also reduces the voltage deviation and power loss on distribution
EMS have been introduced. A useful strategy to energy manage- system. The remaining manuscript is mentioned as below: Section
ment is the heuristic system, the most representative are rule- 2 explains that literature review of current work, Section 3, explains
based approaches [6] and fuzzy logic controllers (FLC) [16e18]. system modelling and design analysis. In section 4 explains that
Rule-based strategies have few rules and algorithms depend on proposed optimization system and their process. In section 5, the
engineering intuition. For instance, the whole electric range-centric simulations and outcomes are obtained with objective functions of
approach and the combined approach are two rule-based ap- proposed GPC-RERNN system. At last, the conclusion is suggested
proaches are usually utilized for PHEV power management design on Section 6.
[6,19]. In the ARE-centric strategy (also named CD-CS strategy),
vehicles use electricity to initially work under CD phase and switch 2. Recent research work: a brief review
to CS mode while the electricity is approximately depleted. In the
combined strategy, the power demand is divided among engine Using various procedures and perspectives broad investigation
and battery based on the rules presented along the driving route. works have already existed on writing which depended on the
Referee [6] categorizes the combined approach into the dominant energy management among electric vehicle charging stations and
engine combined strategy and the dominant electricity combined electricity distribution system. Portion of the works are reviewed
strategy denotes that the fuel economy of these approaches is on here.
considerably affected by driving distance earlier than recharging. Fuzzy grasshopper optimization was introduced by Gampa et al.
An important benefit of rule-based strategies is that they may be [21] for the finest placing of DG and the EVCS. The load of EV and
performed under real time. Referee [18] proposes an FLC strategy the distribution system peak load were analyzed by the introduced
for series PHEVs combining the battery management system de- approach. The major contribution of the introduced approach was
notes that the presented FLC approach is effectual on enhancing substation power factor enhancement, voltage profile enhance-
fuel economy and the battery from being over discharged. Nor- ment, loss minimization. The multi objective of the introduced
mally, these approaches are effectual, simple to execute, highly approach was achieved by the introduced fuzzy grasshopper
robust and broadly utilized online based on the low computational optimization.
load requirement. Suganya et al. [22] have established modified particle swarm
Another methodology available online for EMS design is local optimization for the finest solution among the charging station and
optimization approaches, such as model predictive control (MPC) the grid station. Based on the owner place and the usage patterns
[20] and stochastic dynamic programming [20]. These approaches were used to select the location of charging station. The location of
depend on anticipated driving cycle period in the future and utilize EV was achieved by the two models of the introduced approach.
the optimal control policy in this short-term time horizon. In the- The residential and commercial areas were selected by the intro-
ory, local optimization approaches have the potential to accomplish duced approach for locating the EVCS. Considered each area have
near-optimal power management performance should be per- single or double charging stations. The introduced approach eval-
formed for real-time control if it is adequate and accurate forecasts uated the vehicle to grid process in the PHEV economy.
of future driving conditions. Nevertheless, this type of methodology Li et al. [23] have suggested the game model for energy man-
needs maximum computational capacity [10] that creates it hard to agement of the EVCS and distribution station interaction. The EVCS
use it in vehicles today. Furthermore, it is also difficult to obtain an quality of service was achieved by the network calculus theory. The
accurate prediction of driving conditions [20]. To get the most likely introduced method of Supply function equilibrium (SFE) was used
driving cycles on future, infrastructures outside the vehicle are to obtain the equilibrium of the EVCS and DS. The introduced
needed, like cloud and GPS signals. How to combine the exterior method analysis the cost of the system based on whole sale market
data with the signals of the integrated sensors is also a very hard price.
issue that has not yet been solved. Energy management strategy of electrical vehicle charging sta-
Global optimization approaches, which use advanced optimi- tions was suggested by Pflaum et al. [24] through the randomized
zation algorithms to solve the energy flow control issue, have been algorithm. Based on real time, day-ahead upper bound profiles on
extensively researched [20] in recent times. For instance, Genetic EVCS power consumption were determined by randomized
Algorithm Optimization (GA) is utilized for parallel HEV EMS in approach. With the preconfigured probability, the quality of service
Ref. [20], the optimal goal is to diminish fuel consumption and was guaranteed in introduced approach. improved GA PSO algo-
emissions. Three different drive cycles are utilized to assess the rithm have suggested by Mozafar et al. [25] for determine the
presented GA-based strategy, indicating that it is effectual on optimal location as well as capacity of RES with charging station.
diminishing fuel cost and emissions in the absence of sacrificing The introduced approach contribution was the loss minimization,
dynamic performance. Mainly all of these global optimization cost reduction as well as voltage enhancement. By the use of
strategies mentioned above are helpful in obtaining an optimal weighting coefficients approach, the introduced approach obtains
control policy for PHEVs. Though, the reliance on a priori driving the objective. The introduced approach reduced the loss through
cycles that may barely be get on real-time situation makes strate- the improved GA PSO approach. To allocate the EVCS in the proper
gies depends on global optimization algorithms difficult to imple- location Liu et al. [26] have suggested system dynamic approach.
ment online, which is the biggest demerits in this category. Until The energy storage system planning as well as determine the
now, the literature lacks discussion on how to implement strategies optimal EVCS location was the main contribution of the introduced
depends on global optimization in real-time vehicle control. approach. The introduced approach was analyzed the distributed
In this paper proposed a hybrid approach to optimally manage renewable integration and allocation. For locating the EVCS in a
2
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

certain place, Awasthi et al. [27] have suggested the hybrid GA and reduce costs, from network charges to general system charges,
PSO based approach. Power quality improvement and the cost involving taxes. The objective of demand-side management is to
reduction was the major contribution of the introduced approach. adapt the general consumption picture - consumption time profile,
By the use of introduced approach, the charging station criteria contractual supply parameters - in order to accomplish savings on
were met through imposing minimum pressure on the supply utility rates. To limit these impacts and make sure a balance among
system. energy consumption and the amount of energy that is fed into the
grid, administrators can now use generation and consumption
2.1. Background of research work systems that offer so-called “grid services” in exchange for a fee.
The initial step on Demand Side Management is to conduct an in-
The recent research review shows that the energy management depth analysis of consumption on site: this clarifies the peculiar-
amid the EVCS and distribution system is a significant contributing ities of every individual site and whether consumption habits may
factor. The industries have more focused on the EVCS with distri- be optimized in the absence of resorting to extra instruments.
bution system environment. An EVCS entity becomes highly Whenever a change in habits is not possible or simply not enough
attention at the DS environment. Therefore, to follow the utilization to accomplish the needed cost reductions, the on-site installation
of EVCS, it is essential to assume their interaction along with DS (behind the meter) of the subsequent solutions may be assessed:
which is inducing its more enhancements. By analysing the existing Renewable energy sources (photovoltaic). These solutions need an
studies that investigate gap majority has two features like the cost Energy Management System that is a enthusiastic computing
of the system is higher and the interactive process of energy platform that:
management amid EVCS and DS.
The existing approach some of them are Game based approach,
Monitor all the assets implicated (consumption sites, batteries,
modified particle swarm optimization; grasshopper optimization, production systems);
improved Genetic algorithm (GA) and Particle swarm optimization
Optimize the contributions of batteries and production systems
(PSO) algorithm, randomized algorithm. The main benefit of game on real time, reducing costs linked with intake from the grid and
based learning process is diminishing loss, enhance the efficiency diminishing battery aging;
and the limitation of this method is higher cost. The stagnation of
Utilize the assets implicated to supply services to grid.
local optimization is evicted using grasshopper optimization algo-
rithm. But the performance of the grasshopper optimization algo- Here, the charging station involves three functions under this
rithm is poor global search capacity as well as the convergence system: (i) system operator (ii) EV users, (iii) EV charging station
speed is slow. GA and PSO are better in the search, but one thing the operator. The system operator is defined as local grid system
process is sluggish and fitness coding is complex. The GA algorithm operator that is in charge for grid protection and presenting the
is simply to realize and assists multi-purpose optimization, how- fundamental power service for industries and residents [29].
ever, GA execution is still an art. GA needs lesser information about The system operator tends to supply a dynamic electricity tariff
the issue, but it is complex to design an objective function, get the for promoting end-users and contribute on demand-side energy
representation as well as operator's right. Owing to take more time, management for mitigating the load power requirement and
GA is computationally expensive. The PSO is simply to execute and diminish transmission maintenance costs. EVs have high flexible to
efficient, but the major disadvantages of the PSO algorithm are regulate based on time scale and charging power and developing
“simply to optimize locally in higher dimensional intervals and has optimal candidates for contributing at electricity market likened to
a lower convergence rate in functional performance. The advan- the household utilizations. However, it remains complex to inspire
tages and disadvantages of random algorithms are easy to imple- individual EV users because they are bound by daily routines and
ment and cannot attain the global optimal solution respectively. To charging facilities. It is more sensible for EVCS operator to regulate
address this issue, many methodologies are existed in the litera- the charging of various electric vehicles and recognize the demand
ture, but they express insufficient performance. Such drawbacks side power management through high flexibility. Therefore, the
and issues inspired to do this research work. EVCS operator acts as an intermediary for the charging service
between the users of electric vehicles and the dynamic price of the
3. System configuration electricity market activated by the system operator. In general, the
charging station operator considers the system operator, i.e., does
Fig. 1 shows that the system architecture of energy management not manipulate the electricity market compensation prices.
among EVCS and the distribution system. This architecture incor- Therefore, an adequate EMS is necessary to achieve the afore-
porated with PV with EVCS, energy storage system (ESS), con- mentioned objectives. In this paper EVCS is located at near the
verters, grid and EV. Here, the DC power bus is used to connect the building in which using certain SOC the EV reach EVCS randomly.
PV, EV which is used to lessen that investment cost of the system Before leaving the EV, the EVCS is should charge.
and also enlarge that efficiency of system. For keep the stable DC The process of charging is determined based on actual SOC
bus, the vehicle to grid process is achieved by the DC to DC con- (Eact ), initial SOC (Eini ), arrival time (at) and departure time (at),
verter [28]. The PV array and the boost converter are directly linked gv gv
charging power (Pi;t ) and discharging power (Pi;t ), it is described
with DC bus and assumed as PV is operating at maximal power as,
point of system. Consequently, the solar energy of PV array is given
to the grid by the AC/DC converter or to local EVs for direct vg !
X
dt Pi;t
charging. In this paper the major aim of the proposed approach is Eact ¼ Eini þ gv
Pi;t hi  Dt (1)
at
hi
the cost reduction, loss minimization and minimizes the voltage
deviation.

3.1. Demand-side energy management system

3.2. EV charging stations scheduling
The term Demand Side Management refers to a group of actions
intended to direct and optimize a site's energy consumption and The grid connected EV has many issue in the distribution system
3
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

Fig. 1. System architecture of energy management among EVCS and distribution system.

due to the charging process EV act as load, at idle state act as energy The distribution system provides the information about the
storage system. There are double loads like non-dispatchable charging plans like arrival and discharging time. The grid is offered
charging loads, dispatchable charging loads are considered to the three phase AC supply which is the source of EV. The capaci-
analyze the EV charging station scheduling. EV charging scheduling tance of the dc bus is sited in the dc side of the system for the
is the important parameter for cost minimization [30]. The time of stabilization of DC bus voltage due to high as well as immediate
arrival, discharging information is used for the energy management requirement of the load. The important portion of EVCS is DC bus in
system (EMS) of EVCS. Fig. 2 shows the information flow diagram of which total EVs are related to charging.
EVCS and distribution system. From this figure the demand of the
EV is given to EVCS. Therefore the charging power is provided by
the EVCS. Each EVCS is linked with distribution system. The 3.3. Electric vehicle charging station at quality of service
requirement of EVCS information is flow to the distribution system.
The quality of service of charging station is achieved by depends
4
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

Fig. 2. Information flow diagram of EVCS and DS.

on arrival time departure time and time of departure in minimum 3.4. Mathematical modelling
condition. When the EVCS is incorporated with the distribution
system then it gives the charging operation for EVs. Depend on the The modelling of the system is used to model the system. Here
arrival, departure times and the requirement of EV; the minimum PV, ESS, EV are modelled using the mathematical equation. The
condition departure is achieved. The charging requirement of EV is modelling of the PV, ESS, EV are described the following section.
satisfy at certain period of time is described as QoS is guaranteed.
Over all power requirement with maximum power is the arrival
time and the energy of EV is obtained by specified overall time is
described as departure time [31]. Ideally, it is arranged consequent 3.4.1. Photovoltaic (PV) model
plan to charge from EVCS. The minimum overall energy received by In the PV system, the solar energy is transformed into electrical
EV that meet QOS with certain time interval is described as mini- energy using photovoltaic cell. The PV cell is made up of semi-
mum departure condition. Arrival time (TA ) and departure time conductor material. Because of the semiconductor material PN
(TD ) is described as, junction is creating which is used to change the energy. In this
system to obtain the maximum power tracking strategy is utilized.
    The output power of PV system is described as,
TA ¼ min pmax t  tspe Dt; Cdema (2)
Irinci
ppv ¼ pstd ð1 þ kðtc  tr ÞÞ (6)
X
tspe
Irstd
TD ¼ pdema ðtÞDt (3)
t here pstd specifies the power at standard condition, Irinci specifies
The minimum departure time is described as, the incident irradiance, Irstd specifies irradiance at standard con-
dition, k specifies the temperature coefficient of power, tc specifies
TD;min ¼ minfCdema ; max½Cdema  pdema ðtde  tÞDt; 0g (4) cell temperature, tr and specifies the reference temperature. The PV
source is reversely fed the power to the grid which is used for
To obtain the QoS, EV must meet the following condition, household requirements. It is described by,

TA  TD  TD;min (5) pv pg

pt ¼ Pthouh þ Pt ; ct (7)
Here pdema ðtÞ is specified as required power of EV, Cdema is
specified as required charging energy for EV. Minimizes the total here ppv
t specifies the power injected to grid as PV (kW), pt
houh

load cost and the substation energy cost during the time horizon in denote as power utilized to meet household load as PV (kW). The
question minimizes the total load cost and the substation energy generated power is also depend on power directly sold to grid,
cost during the time horizon in question. building power, storing energy to the ESS which is described by,
5
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

pt
pv;gen
¼ Pt
pv;inj
þ Ptbuild þ Pt
pv;ess
; ct (8) Ptev;Load ¼ Ptev;Dis :hd:ev ; ct (16)

here Ptev;Dis is specified as EV discharging power (kW), hd:ev is

specified as discharging efficiency of EV. The charging power rule of
3.4.2. Modelling of energy storage system (ESS) EV is described as,
When the ESS is discharges means sold to grid or it is utilized for
building load demand. It is the actual power given through the ESS. Ptev;c  cr:ev :bev
t ; ct (17)
The ESS energy state is described as summation of energy obtained
in previously plus the energy provided by loading or difference of here cr:ev is specified as charging rate of EV (kW per time interval),
PV energy utilization and the discharging energy [32]. In the EMS bev
t is specified as EV is charging period t then the binary variable 1,
control the charge and discharge of the ESS. Additionally, it in- other wise 0. The discharging power of EV is defined as,
creases that performance of the system efficiently. The ESS fulfill
 
the load demand by, pev;
t
Dis
 dr:ev : 1  bev
t ct (18)
ess;inj
Pt þ Ptess;build ¼ Ptess;Dis :hd:ess ; ct (9) here dr:ev is specified as discharge rate of EV.

here Ptess;ini specified as initial load used the power of ESS (kW),
Ptess;build specified as building load used the power of ESS (kW), Ptess;Dis 3.5. Operational constraints of EVCS with DS
denoted as ESS discharging power (kW), hd:ESS specified as dis-
charging performance ESS. The limit of charging power of ESS is 3.5.1. Power balance
described by, The purchased energy from grid and sold energy to EV is in
equal manner. The load requirement is achieved by the power
Ptess;c  cr:ess :bess
t ; ct (10) balancing of system which is also used to diminish the system cost
and give that stable system [34]. The power balance constraint is
here cr:ess is specified as ESS charge rate (kW per time interval), bess described as,
t
is specified as ESS charges during period t then the binary variable
grid pv;build ess;build
1, other wise 0. The limit of discharging power of ESS is described Pt þ Pt;i þ Pt;i þ PiEV;build
by, ess;char ev;char
¼ Ptbuild þ Pt;i þ Pt;i ; ct2T (19)
 
Ptess;Dis  dr:ess : 1  bess
t ct (11) The above equation is used to balance the income and outcome
of electric power of the energy management system in the each
here dr:ess is specified as ESS discharge rate (kW per time interval). time period. The overall load of the system is incorporated with
The state-of-energy of ESS (kW h) is followed as, building load requirement, ESS charging necessitate, EV charging
necessitate is meet the request from grid and/or through the joint
Ptess;c Ptess;stor Ptess;d
t ¼ soet1 þ hc:ess :
soeess ess
þ  ; ct  1 (12) procurement of PV, ESS supplied power and exhaust power of EVs.
DT DT DT
If time is 1 h then the ESS energy state is equal to initial state of
energy. It is described, 3.5.2. Protection constraints
To ensure the safe operation of the DS, the below restrictions
soeess in ess related to the voltage level of every terminal and the power flow of
t ¼ soe ; If t ¼ 1 (13)
every branch must be observed.

soeess
t  soe
max ess
; ct (14) 2
Vmin 2
 Vi ðtÞ  Vmax ci2nodeset (20)

soeess
t  soe
min ess
; ct (15) 2
Iij  Imax (21)
here hc:ess is specified as charging efficiency of the ESS, t denote as
here Vmin and Vmax is specified as minimal and maximal values for
day index on time units (h or min) DT is specified number of time
voltage in bus. i
intervals under 1 h, Ptess;c , Ptess;stor , Ptess;d is specified as charging,
stored and discharging power of ESS respectively.
3.5.3. Limits of PV generation units operation
3.4.3. Modelling of electric vehicle By allowing for PV resource limits, it should operate under the
The EV model is incorporated with the driver design, power condition
train design along with vehicle dynamics design. At driver model
PV PV
the input is taken as real speed which is distinct with the reference Pi;t  Pi;max ; …i ¼ 1; 2; 3……; Nb (22)
speed. In the power train design accept the input from the driver
model. Based on the location of the brake pedal as well as the PV
here Pi;max is specified as maximal capacity of PV unit that may be
accelerator pedal the outcome of driver model is obtained. The installed.
driver pedal position is achieved by the power train which is used
to manage the energy of the system. Based on the force of vehicle,
dynamic model process the vehicle with real speed. The vehicle 3.5.4. Constraint of upper limit capacity of EVCS
forces such as traction force, aerodynamic drag force, rolling The capacity of every charging station on given area is described
resistance force, and force of gravity [33]. Mathematical expression as number of electric vehicles on charging station. The restriction of
of EV power is described as, the charging station is described by,
6
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

. . 
cs  csmax (23) C land ¼ C lease C rent C inv (31)

here CSmax is specified as maximal capacity of charging station. The land cost is considered as; it may be land lease cost C lease or
rental cost C rent or the buying cost of land C inv .
3.5.5. Constraints of SOC
At a given moment SOC is described as electric energy on EV 3.6.4. Power losses
battery. The essential portion of EV is rated power of battery's The power loss of the distribution system is described as line
inverter and allowable SOC. The charging power and SOC is loss which is described as,
described as,
pj ¼ i2 r (32)
min
Pi;t  Pit  Pimax (24)
The total power loss is described as,

socmin
i;t  socti  socmax
i (25) X
n
pt ¼ pj (33)
j¼1

3.6. Objective function

3.6.5. Voltage deviation

The major purpose of the proposed method is to diminish the
cost of the system that incorporates the cost of the land, the station The management of voltage is the important parameter for the
distribution network. The voltage can vary depend on time; it
equipment, the cost of operation and preservation also diminishes
the voltage deviation and the loss of energy on distribution system. should affect the performance of the system [36]. If minimize the
deviation of voltage the performance of the system is enhanced.
In this paper, the objective function is described by:
The voltage deviation is defined by,
X
OBJ ¼ MIN ðCmin þ Power loss þ Voltage deviationÞÞ N 
cos t;loss;vgedevi x X Vi  Vspec
Vd ¼ (34)
(26) i¼1
Vspec

 
cos tmin ¼ min C I þ C st:equip þ C O&M (27) here specified voltage is indicate as Vsp , voltage magnitude bus i is
indicate as Vi .

4. Proposed energy management system among EVCS and DS

using GPC-RERNN
3.6.1. Operation and maintenance cost
The cost of equipment is considered as one time investment. But
The proposed energy management system is managing the
the operation and maintenance cost is varies based on the opera- energy based on the hybrid strategy. The hybrid strategy is joined
tion [35]. It incorporated with fixed labor or other management
implementation of Giza Pyramids Construction (GPC) and
costs which is described by, recalling-enhanced recurrent neural network (RERNN) hence it is
 named as GPC-RERNN approach. At first, RERNN is used to originate
C O&M ¼ C electri þ C lab:pay þ C mainten (28) the quality-of-service decision model of EVCSs. The utility of DS is
increase when scheduling the charging plans of EVCSs. Then, to
here cost of electricity purchase is specified as C electri , cost of labor analyze the energy interaction, the GPC approach is used. In GPC
wage is specified as C lab:pay , cost of equipment maintenance is approach, every EVCS is considered from leader and DS is consid-
specified as C mainten . ered from follower. Fig. 3 shows the interactional development of
energy management amid the EVCS and DS. The reason for setting
3.6.2. Station equipment cost up EVCS as leaders is energy needs is provided maximum priority
The station equipment cost is cost of electric vehicle charging to maintain the wider consumption of EV [37]. The EVCS is sending
station which is described as, the different power needs with different rates to the operator of
distribution system. The charging plan of EVCS and the purchase
X
n X plans of market is optimized by the DS depend on cost and demand.
C equip ðxÞ ¼ jij pequip
ij
Nij ðxÞ (29)
This process increases the usability of the DS, when meeting the
i¼1 j2jequip
i
energy needs of EV users before departure. The proposed system
  .  determines the optimal solution based on interaction of EVCS and
jij ¼ RD 1 þ RD ð1 þ RD ÞLij ð1 þ RD ÞLij  1 (30) DS. Through which the utilization of EVCS is increasing.

Here capital recovery factor is specified as jij , discount rate is 4.1. Proposed recalling-enhanced recurrent neural network
(RERNN) for energy management system
specified as RD , lifetime of cost element is specified as Lij, total cost
paid per unit of production equipment is specified as pequip
ij
. Recalling-enhanced recurrent neural network (RERNN) is the
artificial neural network that utilized radial function implemented
3.6.3. Land cost in the field of mathematical modelling. Elman recurrent neural
The cost of land accounts for great portion of total cost of network have three layers but the recalling enhanced neural
recovering station investment, and subsidized form of grants may network consist of six layers and incorporated with selective
greatly reduce that cost of land with few extents. memory property. The RERNN is trained by an improved conjugate
7
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

Fig. 3. Interactional process of energy management amid the EVCS and DS.

algorithm with generalized Armijo search approach, which can 4.1.1. Step by step process of RERNN
improve the convergence speed [38]. The Elman neural network Step 1: Initialization.
has slow convergence to overcome that the RERNN is introduced. Initialize the parameters of EV like number of nodes, weight
The structure of RERNN is shown in Fig. 4. The RERNN layers are vectors W ¼ ½W1 ; W2 ; … Wn  and the iteration numbersi and
Input layer, State layer, Memory layer, Sum layer, Hidden layer, the hidden node number.
Delay layer Output layer. The input node accepts the input and Step 2: Random Generation.
hidden layer output with delay function. The state layer accepts the The input vectors are randomly generated after the process of
input that can be given to the memory layer. The memory layer is initialization. Here the input parameters are taken as EV system
accepting the previous outcome of sum layer and current outcome parameter like battery SOC, engine torque, EM torque and speed
of state layer. and battery power which are randomly generated.
The major function of the memory layer is determined the size Step 3: Check the iteration.
of the previous sum layer information transmits to next stage. Sum If the iteration is less than the maximum iteration then process
layer add the information of present input, final recurrent hidden the data otherwise stop the process.
outcome and the outcome of memory layer. The hidden layer Step 4: Find learning rate by generalized Armijo search
outcomes are providing the final probabilistic value at the output approach.
layer. The delay layer propagates back to present hidden layer The learning rate is found by the generalized Armijo search
output. Output vector calculate the output based on conjugate approach by the following condition such as,
gradient descent algorithm, generalized Armijo search approach
     t
[39]. In the proposed approach, the function of RERNN is origi-
E wk þ LR pk  E wk þ a1 LR Ew
k
pk a1  0 (35)
nating the quality-of-service decision model for EVCSs. Depending
on quality of service it learns the data of the EVCS and DS. Step 5: Calculate the new weight.
8
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

Fig. 4. Structure of RERNN.

The new weight is calculated by using the gradient descent al- EVCS functions in parallel. Here, the main problem is considered
gorithm, which is described as, cost reduction and the sub problem is considered energy loss and
voltage optimization. The step-by-step process of the GPC algo-
wkþ1 ¼ wk þ LR P k (36) rithm is explained below.

Step 6: Check the maximum iteration.

If the iteration is reached then stop the process. Otherwise in- 4.2.1. Step by step process of GPC
crease the iteration and go to step 6. Step 1: Initialization.
Step 7: Calculate the direction. The outcome of RERNN approach is given to the input of the
Using the conjugate gradient descent algorithm the direction of GPC. Using that approach, initialize the cost, population.
the learning process is calculated by, Step 2: Random Generation.
The input vectors are arbitrarily created after initialization
P k ¼  Ew
k
þ bP k1 (37) process. Here the input parameters are taken as EV system
parameter like battery SOC, engine torque, EM torque, and battery
 t power which are randomly generated.
aEw
k P k1
Step 3: Find the fitness function.
b¼  ; a2ð0; 1Þ (38) Check the best learning value that is considered as pharaoh's
P k1 P k1  Ew
k
agent.
Step 4: Calculate the stone displacement.
Then go to the step 4. Again repeat the process. Fig. 5 shows the
The displacement of the stone is specified by,
flow chart of RERNN approach.
v2
4.2. Proposed Giza Pyramids Construction GPC approach for solving D¼ (39)
2gðsin q þ m cos q
optimization issues
here D is specified as the value of displacement. g specified as
The first ancient-inspired metaheuristic algorithm is Giza Pyr-
gravity of Earth, v is specified as velocity, m is specified as kinetic
amid Construction (GPC) Optimization approach. The Giza Pyramid
friction coefficient.
Complex is a site of three great pyramids, the entire built of ancient
Step 5: Update the position of workers.
Egypt [40]. The inspiration of the algorithm is the activities of the
The position of workers is updated by,
workers and pushing that stone blocks on ramp. Here, the workers
carry that stone blocks. This process is carried out in direct v2
administration of Pharaoh's agent. In this document, GPC is utilized x¼ (40)
2g sin q
to skew that original optimization model into one parent problem
and various child problems. The main problems are improving the Using the value find the position and also check the fitness for
DS, while the secondary problems improve the EVCS function in finding the best workers.
parallel. The main problem has been solved, an optimal power Step 6: Calculate new position and new cost.
transfer program for ancillary problems is provided to enhance the Using the above change position set the new cost for each
9
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

Fig. 5. Flowchart of the recurrent neural network (RERNN).

worker. increased to reach the value of 28 kW. Then it decreased a small

Step 7: Compare the cost. value to reach 27 kW at the time period of 12 h. Then it again
Check the cost value is grater or less than the pharaoh's agent reduced to reach 26.5 kW 16 h. After that the power is increased to
cost. if the cost is less than pharaoh's agent cost then set new cost as reach the peak value of 29 kW at the time period of 15 h. Then it
pharaoh's agent cost otherwise go to next step. Based on the cost gradually decreased to reach the value of 17 kW at 22e24 h. The
comparison, find the best solution. cost of the system is varying based on the time interval is also
Step 8: Stopping criteria. illustrated in this Fig. 7. At 6 h, the price signal cost is 2.2 $. Then it
Check the stopping criteria. If the stooping criteria reach the increased to reach the value of 3.8 $ at the time of 8 h. Then the
maximum iteration stop the process, otherwise go to step 4. Fig. 6 price signal is gradually decreased to reach the value of 2.5 $ at the
shows the flowchart of the Giza pyramid construction. time of 14 h. After that the price signal is increased to reach the
value of 4.1 $ at the time period of 18 h. Then it gradually decreased
5. Results to reach the value of 2.2 $.
Fig. 8 shows the analysis of overall energy requested as grid
In this manuscript, a hybrid GPC-RERNN approach for EMS through EMS to every scenario. Fig. 8 (a) shows the total energy
among EV and distribution system is introduced. This paper in- demand as a network through EMS for scenario 1 and scenario 2
vestigates that cooperative evaluation of EMS operation using and scenario 3. For scenario 1, the maximum power requested as
several considerations, like bidirectional energy trading capabilities grid is 145 kW with a duration from 13 to 15 h. In scenario 2, the
of arrival time of electric vehicle fleet under driving plan of electric maximum power requested as grid is 144 kW with a time duration
vehicle, the force of photovoltaic indecision. At EMS operation, the of 7e11 h and 12e13 h. In the scenario 3, the maximum demand for
establishment of dissimilar Prioritization factors involve the sale of energy as a grid by the EMS is 144 KW in 10e12 h time interval.
energy to the resource network on total cost of the system [41]. The Fig. 8 (b) shows the total energy demand as grid through EMS for
implementation of the proposed model is performed at MATLAB/ scenario 4 and scenario 5 and scenario 6. For scenario 4, the
Simulink and the execution is assessed with the existing system. maximum power requested from the grid is 145 kW with duration
The proposed system analysis the result with six scenarios and of time from 8 to 9 h. In scenario 5, the maximum power requested
evaluate the result with existing approaches like GA, PSO, GL-BA. from the grid is 144 kW with time duration of 9e10 h and 15e16 h.
In scenario 6, the maximum demand for energy as a network by
5.1. Discussion EMS is 144 KW in the time period of 15e18 h.
Fig. 9 shows that analysis of the total energy injected into the
Fig. 7 shows the analysis of power demand of building in real network through EMS in each scenario. Fig. 9 (a) shows the total
time. Here, at time 0e6 h the power demand of the building is energy injected into the grid through EMS for scenario 1 and sce-
17 kW. From 6 h to 11 h the demand of the building is gradually nario 2 and scenario 3. For scenario 1, the maximum power injected

10
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

Fig. 6. Flowchart of the giza pyramids construction (GPC).

into the grid is 59 kW with duration from 8 to 9 h. In scenario 2, the

maximum power injected into the grid is greater than 100 kW with
time duration of 15e18 h. In scenario 3, the maximum energy
injected into the grid by the EMS is 90 KW in a period of time of
13e15 h. Fig. 9 (b) shows the total energy injected into the grid
through EMS for scenario 4 and scenario 5 and scenario 6 For
scenario 4, the maximum power injected into the grid is 140 kW
with a duration of time from 10 to 15 h. In scenario 5, the injected
maximum power to the grid is 144 kW at the time duration of
9e10 hrs and 15e16 h. In scenario 6, the maximum energy injected
into the grid by the EMS is 144 KW in the time period of 10e15 h.
Fig. 10 shows the total energy requested as a network and
injected into the network through EMS. The total energy requested
by a network through EMS is 40 kW in the time period of 0e7 h and
then the power is reduced to zero in the period of 9 h. After that, the
requested power is increased to reach the value of 140 kW in the
time period of 15e17 h. Then it decreased to zero from 18 to 21 h.
Then it increased slightly to reach the value of 5 kW in a time of
Fig. 7. Analysis of power demand of building and the price signal variation based on
time. 21e24 h. The total energy injected into the network through EMS is
0 in the initial time from 0 to 6 h. Then, the power is increased to
reach the value of 50 KW in a time interval of 7e8 h. In the time

11
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

Fig. 8. Performance of general energy requested as grid through the EMS to (a) Scenario 1e3, (b) Scenario 4-6.

Fig. 9. Performance of general energy injected back to the grid through the EMS to (a) Scenario 1e3, (b) Scenario 4-6.

Fig. 11 shows the decomposition of the energy fed to the grid as

PV, ESS and EVs for two scenarios and random method. Fig. 11(a)
shows the decomposition of energy fed to the grid as PV, ESS, and
EVs for scenario 1. The PV power of scenario 1 is zero for whole time
and the ESS power is varied from 0 to 18 kw at 5e10 h. Then it is
zero from 10 h to 16 h. At 16 h the ESS power is again increased to
reach 17 kW at 16e21 h, then again it reached to zero. In the EVs
power of scenario 1 is initially zero. At the time period among
6e10 h, it reaches the value of 40 kW. From 10 to 16 h the EVs
power is again zero. After, it increased to reach 25 kW at 16e19 h.
Fig. 11(b) shows that decomposition of energy fed to the grid from
PV, ESS, and EVs for scenario 6. The PV power of scenario 6 is zero at
the time period of 0e8 h. Then it increased to 25 kW at 10 h. Then it
slightly decreased and again increased to reach the value of 30 kW
at 12e14 h. Again it becomes zero and slight changes occur at 18 h
time interval. The ESS power is fed to the grid at the time period of
9e14 h and the maximum power 30 kW. In the EVs power of sce-
Fig. 10. Performance of general energy requested grid and injected into the grid via nario 6 is varied 9e14 h; at that time the power reaches the value of
EMS. 100 kW at two times. Fig. 11 (c) shows the random approach of
energy decomposition in which energy is fed to grid from PV, ESS
and EVs. Here PV, ESS, and EVs powers are varied only at 6e8 h and
period from 9 to 16 h, the power injected into the grid is zero. Then 16e19 h time interval. Other time the power becomes zero. The
it increased to 55 kW in the time period of 16e19 h. Again it is peak power value of PV is 5 kW, ESS peak power is 19 kW and the
reduced to zero in the time period of 19e24 h.

12
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

Fig. 11. Performance of decomposition of energy fed to the grid from the PV, ESS, and EVs for (a) scenario 1, (b) scenario 6 and (c) the random approach.

EVs maximum power is 35 kW. demonstrated at Fig. 12 (b). In scenario 6, the photovoltaic power is
Fig. 12 shows the decomposition of generated PV energy for PV varied up to 15 kW in the time interval of 9e12 h, and increased to
system (a) scenario 1 (b) scenario 6 and random strategy. In sce- 27 kW in the time interval of 14e15 h. The photovoltaic power at
nario 1, the analysis of the decomposition of the photovoltaic en- ESS is 23 kW and the power 0.6 kW in the time interval from 10 to
ergy generated is illustrated in Fig. 12 (a). In scenario 1, the 12 h and from 18 to 19 h. In a random approach, the analysis of the
photovoltaic power is very low, around 0.3 kW. The photovoltaic decomposition of the photovoltaic energy generated is illustrated
power for the building is 0.8 kW and 0.6 kW in the time interval in Fig. 12 (c). In the random approach, the photovoltaic power is
from 10 to 12 h and from 18 to 19 h. The power from PV to ESS is varied up to 15 kW in a time interval of 12 h. The power from PV to
2.3 kW in a time interval of 11e18 h. In scenario 6, the analysis of ESS is 14 kW.
the decomposition of the photovoltaic energy generated is Fig. 13 shows the decomposition of ESS presented energy for PV

Fig. 12. Performance of decomposition of generated PV energy in (a) scenario 1 (b) scenario 6 and the random approach.

13
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

Fig. 13. ESS decomposition of given energy for PV under (a) scenario 1, (b) scenario 6 and (c) Random approach.

under (a) scenario 1, (b) scenario 6. The decomposition of ESS maximum power is also around 17.5 kW.
presented energy for PV under scenario 1 is demonstrated at Fig. 14 shows the analysis of ESS state of energy of scenario 1, 2
Fig. 13(a). The maximal power of ESS to grid is 17.5 kW. ESS to build and the random approach. Fig. 14 (a) shows the analysis of ESS state
maximum power is also around 17.5 kW. The decomposition of ESS of energy of scenario 1, 2. In scenario 1 the state of energy is
presented energy for PV under scenario 6 is shown in Fig. 13(b). The increased from 50 kWh to 118 kWh. Then the energy is decreased to
maximum power of ESS to grid is 17.5 kW. ESS to build maximum reach the value of 25 kWh. Again the energy is increased to reach
power is also around 17.5 kW. The decomposition of ESS presented the value of above 120 kWh then the energy is decreased to reach
energy for PV under random strategy demonstrated at Fig. 13(c). the value of 30 kWh. At scenario 2, the state of charge is start at 1 h
The maximum power of ESS to grid is 17.5 kW. ESS to build and the energy value is increased to reach above 120 kWh at 8 h

Fig. 14. Performance of ESS state of energy for (a) scenario 1, 2, (b) random approach.

14
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

then it decreased to reach the value of 30 kWh at 15 h. Then the

energy value is increased to reach the value of 70 kWh. Then it
decreased and increased reach the value of 98 kW at 21e23 h.
Fig. 14 (b) shows the analysis of ESS state of energy in the random
strategy. Here the power is increased from 50 kW to 119 kW. Then
the power is decreased to reach the value of 70 kW at 11e13 h. The
power is increased to reach 110 kW at 18 h and then decreased to
reach the value of 30 at 21e24 h.
Fig. 15 shows the performance of decomposition of produced PV
energy for different scenarios. Fig. 15 (a) shows the performance of
decomposition of produced PV energy for scenario1. The EV to grid
power is varied among the 7e10 h and 17e19 h. The power value at
the period is 39 kW and 25 kW. EV to building power is varied at the
time intervals of the 5e12 h and 17e21 h. The power value at above
time duration is 20 kW and 35 kW. Fig. 15 (b) shows the perfor-
mance of decomposition of produced PV energy for scenario 6. In
this EV to grid and EV to create power is considered. The EV to grid
power is varied 11e14 h. The power value at the period is 95 kW. EV
to building power is varied at the time intervals of the 9e14 h and Fig. 16. Energy injected to the grid as PV for random strategy.
17e21 h. The power value at above time duration is 40 kW and
20 kW. Fig. 15 (c) shows the analysis of decomposition of created PV
energy for scenario 6. The EV to grid power is varied 6e10 h and grid in a time interval of 17e19 h. For electric vehicles, 48 kW and
16e20 h. The power value at the period is 27.5 kW and 32 kW. EV to 55 kW, the energy is injected into the grid in a period of time from 6
building power is varied at the time intervals of the 6e8 h and to 8 h and from 16 to 19 h. Fig. 17 shows the analysis of the
17e20 h. The power value at above time duration is 20 kW and decomposition of the energy fed to the grid in (a) PV, ESS and EV.
25 kW. Fig. 17 (a) shows the analysis of the decomposition of the energy fed
Fig. 16 shows the decomposition of the energy injected into the to the grid in PV in scenario 1, 6. In this Condition, the photovoltaic
grid from photovoltaic sources for the random approach. The power becomes zero in scenario 1. The power of scenario 6 is
decomposition of the energy injected into the network from ESS for modified only at the time from 11 to 12 h. Ranged up to 48 kW.
the deterministic approach remains constant. For electric vehicles, Fig. 17 (b) shows the breakdown analysis of power fed to the grid
56 kW of power is injected into the grid in a time interval of under ESS. In this condition, the power of scenario 1 is modified in 2
17e19 h. For photovoltaic energy, 5 KW of power is injected into the moments, such as from 6 to 10 h and from 16 to 21 h. The power of
scenario 6 is varied only at the moment from 10 to 14 h. Ranged up

Fig. 15. Performance of decomposition of produced PV energy for (a) Scenario 1, (b) Scenario 2 (c) random approach.

15
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

Fig. 17. Performance of decomposition of energy fed to grid as PV under (a) PV, ESS, and EVs.

to 150 kW. Fig. 17 (c) shows the analysis of the breakdown of energy around 9000. At scenario 3 the operating cost of the station 1, 2, 3
fed to the grid in electric vehicles. In this condition the scenario 1 are around 11,000. At scenario 4 the operating cost of the station 1,
power becomes changed at the time of 17e18 h (i.e a small change 2, 3 are also around 11,000. At scenario 5 the operating cost of the
occurs). The scenario 6 power is varied at the time of 9e14 h and station 1, 2, 3 are around 12,000. At scenario 6 the operating cost of
18e19 h. It varied up to 30 kW. the station 1, 2, 3 are around 13,000. Fig. 18 (b) shows the com-
In Fig. 18 shows the operation cost comparison of station 1, 2, 3 parison of operation cost of station 1, 2, 3 using PSO method. At
using GA, PSO, DL-GA and the proposed system. The cost of the scenario 1 the operating cost of the station 1 is around 9500,
system is compared to the scenarios. Fig. 18 (a) shows the com- operating cost of the station 2 is also around 9500 and operating
parison of operation cost of station 1, 2, 3 using GA method. At cost of the station 3 is 9000. At scenario 2 the operating cost of the
scenario 1 the operating cost of the station 1 is 9500, operating cost station 1, 2, 3 are around 11,000. At scenario 3 the operating cost of
of the station 2 is above 10,000 and operating cost of the station 3 is the station 1, 2, 3 are around 11,000. At scenario 4 the operating cost
9000. At scenario 2 the operating cost of the station 1, 2, 3 are of the station 1, 2, 3 are also around 11000and the station 1 cost is

Fig. 18. Comparison of operation cost of stations 1, 2, 3 using (a) GA, (b) PSO and (c) DL-GA (d) proposed method.

16
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

increased from 6 to 65 s at 1e7 days. The FFA-GBDT approach, CPU

time is increased from 7 to 68 s, at 1e7 days. The DL þ GA approach,
CPU time is increased from 10 to 72 s, at 1e7 days. The PSO
approach, CPU time is increased from 12 to 76 s, at 1e7 days. The GA
approach, CPU time is increased from 13 to 78 s, at 1e7 days.
Fig. 20 shows the sensitivity analysis of number of EVs is
charged on EVCS and the unit price of unconvinced energy de-
mands. Fig. 20 (a) shows the sensitivity analysis based on cost on
EVCS. In this analysis the proposed approach is compared to the
existing approaches such as GA, PSO, DL þ GA, FFA þ GBDT, GPC-
RERNN. The proposed approach cost is very low compared to
other methods. The proposed method cost is below 100 $, The FFA-
GBDT approach, the cost of system is 200$ for all the EVs. The
DL þ GA approach, the cost of system is below 350 $ for all the EVs.
The PSO approach, the cost of system is below 400 $ for all the EVs.
The GA approach, the cost of system is around the 400 $ for all the
EVs. Fig. 20 (b) shows sensitivity analysis of the unit price of
Fig. 19. Comparison of computation efficiency of proposed and existing approach. compensation for non satisfied energy demands based on cost. In
this analysis the proposed approach is compared to the existing
approaches such as GA, PSO, DL þ GA, FFA þ GBDT, GPC-RERNN. The
increased to 12,000. At scenario 5 the operating cost of the station 1, proposed approach cost is very low compared to other methods.
2, 3 are around 12,000. At scenario 6 the operating cost of the The proposed method cost is around 100 $, The FFA-GBDT
station 1, 2, 3 are around 12,500. Fig. 18 (c) shows the comparison of approach, the cost of system is 200$ for unsatisfactory energy de-
operation cost of station 1, 2, 3 using DL-GA method. At scenario 1 mands. The DL þ GA approach, the cost of system is increased from
the operating cost of the station is around 9500, operating cost of 0 to 850$ for unsatisfactory energy demands. The PSO approach,
the stations are also around 9500. At scenario 2 the operating cost the cost of system is below 400 $ for all the EVs. The GA approach,
of the station 1, 2, 3 are around 9800. At scenario 3 the operating the cost of system is 400 $ for all unsatisfactory energy demands.
cost of the station 1, 2, 3 are around 12,000. At scenario 4 the The GA approach, the cost of system is around the 500 $ for all
operating cost of the station 1, 2, 3 are also around 12000and the unsatisfactory energy demands.
station 1 cost is increased to 12,200. At scenario 5 the operating
cost of the station 1, 2, 3 are around 12,500. At scenario 6 the
6. Conclusion
operating cost of the station 1, 2, 3 are around 13,000. Fig. 18 (d)
shows the comparison of operation cost of station 1, 2, 3 using
In this paper proposed the EMS aid the EVCS and the distribu-
proposed method. At scenario 1 the operating cost of the station is
tion system based on GPC-RERNN approach. The proposed RERNN
around 9500, operating cost of the station 2 is above 10,000. At
approach is used to determine the quality of service of the EVCS.
scenario 2 the operating cost of the station 1, 2, 3 are around 10,000
The optimization problem is solved by the GPC approach. The
and below values. At scenario 3 the operating cost of the station 1,
proposed method reduces the cost effectively. The proposed
2, 3 are around 12,000. At scenario 4 the operating cost of the
approach is analyzed under six scenarios. Using that scenarios
station 2, 3 are also around 12000and the station 1 cost is increased
compared the performance of the proposed system with existing
to 12,500. At scenario 5 the operating cost of the station 1, 2, 3 are
approaches like GA, PSO, DL-GA, and FFA-GBDT. The proposed
around 13,000. At scenario 6 the operating cost of the station 1, 2, 3
method evaluates the cost of the system based on sensitivity
are around 13,500.
analysis. As a result, the result demonstrates that present system is
Fig. 19 shows the comparison of computation efficiency of the
effectual to get optimal solution using least computation and
proposed and the existing approach. The time is varied to 7 days
lessens complexity of needed algorithm. Future optimization of
and the CPU time is changed to 80 s. The proposed approach time is
electric vehicles focuses on technological optimization and market

Fig. 20. Sensitivity analyses of (a) EVCS based on cost and (b) Energy demands based on cost.

17
B. Rajani and B.N. Kommula Energy 245 (2022) 123180

improvement. We expect a significant change in the near future in system for improved integration of PHEVs and PV arrays. IEEE Trans Sustain
Energy 2015;6(3):659e74.
the development of the quality of the voltage, which will aim to
[17] Hafez O, Bhattacharya K. Integrating EV charging stations as smart loads for
diminish the long-term variations in the magnitude of the voltage demand response provisions in distribution systems. IEEE Trans Smart Grid
that arise in a context of decentralized RES generation. 2018;9(2):1096e106.
[18] Liao Y, Lu C. Dispatch of EV charging station energy resources for sustainable
mobility. IEEE Trans Transport Electrific 2015;1(1):86e93.
Credit author statement [19] Masoum A, Deilami S, Moses P, Masoum M, Abu-Siada A. Smart load man-
agement of plug-in electric vehicles in distribution and residential networks
B. Rajani: Conceptualization, Methodology, Writing e original with charging stations for peak shaving and loss minimisation considering
voltage regulation. IET Gener, Transm Distrib 2011;5(8):877.
draft. Bapayya Naidu Kommula: Supervision, Writing e review & [20] Suganya S, Charles Raja S, Venkatesh P. Smart management of distinct plug-in
editing. hybrid electric vehicle charging stations considering mobility pattern and site
characteristics. Int J Energy Res 2017;41(14):2268e81.
[21] Gampa S, Jasthi K, Goli P, Das D, Bansal R. Grasshopper optimization algorithm
Funding information based two stage fuzzy multiobjective approach for optimum sizing and
placement of distributed generations, shunt capacitors and electric vehicle
None. charging stations. J Energy Storage 2020;27:101117.
[22] Suganya S, Raja S, Venkatesh P. Simultaneous coordination of distinct plug-in
Hybrid Electric Vehicle charging stations: a modified Particle Swarm Opti-
Declaration of competing interest mization approach. Energy 2017;138:92e102.
[23] Li Y, et al. Supply function game based energy management between electric
vehicle charging stations and electricity distribution system considering
The authors declare that they have no known competing
quality of service. IEEE Transactions on Industry Applications; 2020. 1-1.
financial interests or personal relationships that could have [24] Pflaum P, Alamir M, Lamoudi M. Probabilistic energy management strategy for
appeared to influence the work reported in this paper. EV charging stations using randomized algorithms. IEEE Trans Control Syst
Technol 2018;26(3):1099e106.
[25] Mozafar M, Moradi M, Amini M. A simultaneous approach for optimal allo-
Acknowledgement cation of renewable energy sources and electric vehicle charging stations in
smart grids based on improved GA-PSO algorithm. Sustain Cities Soc 2017;32:
None. 627e37.
[26] Liu J, Zhang T, Zhu J, Ma T. Allocation optimization of electric vehicle charging
station (EVCS) considering with charging satisfaction and distributed re-
References newables integration. Energy 2018;164:560e74.
[27] Awasthi A, Venkitusamy K, Padmanaban S, Selvamuthukumaran R,
[1] Sun S, Yang Q, Yan W. Hierarchical optimal planning approach for plug-in Blaabjerg F, Singh A. Optimal planning of electric vehicle charging station at
electric vehicle fast charging stations based on temporal-SoC charging de- the distribution system using hybrid optimization algorithm. Energy
mand characterisation. IET Gener, Transm Distrib 2018;12(20):4388e95. 2017;133:70e8.
[2] Jiang C, Jing Z, Cui X, Ji T, Wu Q. Multiple agents and reinforcement learning [28] Shakerighadi B, Anvari-Moghaddam A, Ebrahimzadeh E, Blaabjerg F, Bak C.
for modelling charging loads of electric taxis. Appl Energy 2018;222:158e68. A hierarchical game theoretical approach for energy management of electric
[3] Wang S, Dong Z, Luo F, Meng K, Zhang Y. Stochastic collaborative planning of vehicles and charging stations in smart grids. IEEE Access 2018;6:67223e34.
electric vehicle charging stations and power distribution system. IEEE Trans [29] Zhang H, Hu Z, Xu Z, Song Y. Evaluation of achievable vehicle-to-grid capacity
Ind Inf 2018;14(1):321e31. using aggregate PEV model. IEEE Trans Power Syst 2017;32(1):784e94.
[4] Singh S, Jagota S, Singh M. Energy management and voltage stabilization in an [30] Tushar M, Zeineddine A, Assi C. Demand-side management by regulating
islanded microgrid through an electric vehicle charging station. Sustain Cities charging and discharging of the EV, ESS, and utilizing renewable energy. IEEE
Soc 2018;41:679e94. Trans Ind Inf 2018;14(1):117e26.
[5] Shojaabadi S, Abapour S, Abapour M, Nahavandi A. Simultaneous planning of [31] Chaudhari K, Ukil A, Kumar K, Manandhar U, Kollimalla S. Hybrid optimization
plug-in hybrid electric vehicle charging stations and wind power generation for economic deployment of ESS in PV-integrated EV charging stations. IEEE
in distribution networks considering uncertainties. 2020. Trans Ind Inf 2018;14(1):106e16.
[6] Shin M, Choi D, Kim J. Cooperative management for PV/ESS-Enabled electric [32] Yan Q, Zhang B, Kezunovic M. Optimized operational cost reduction for an EV
vehicle charging stations: a multiagent deep reinforcement learning charging station integrated with battery energy storage and PV generation.
approach. IEEE Trans Ind Inf 2020;16(5):3493e503. IEEE Trans Smart Grid 2019;10(2):2096e106.
[7] Shakerighadi B, Anvari-Moghaddam A, Ebrahimzadeh E, Blaabjerg F, Bak C. [33] Shojaabadi S, Abapour S, Abapour M, Nahavandi A. Simultaneous planning of
A hierarchical game theoretical approach for energy management of electric plug-in hybrid electric vehicle charging stations and wind power generation
vehicles and charging stations in smart grids. IEEE Access 2018;6:67223e34. in distribution networks considering uncertainties. 2020.
[8] Luo L, Wu Z, Gu W, Huang H, Gao S, Han J. Coordinated allocation of [34] Li D, Zouma A, Liao J, Yang H. An energy management strategy with renewable
distributed generation resources and electric vehicle charging stations in energy and energy storage system for a large electric vehicle charging station.
distribution systems with vehicle-to-grid interaction. Energy 2020;192: eTransportation 2020;6:100076.
116631. [35] Zhang H, Hu Z, Xu Z, Song Y. Evaluation of achievable vehicle-to-grid capacity
[9] Moghaddam Z, Ahmad I, Habibi D, Phung Q. Smart charging strategy for using aggregate PEV model. IEEE Trans Power Syst 2017;32(1):784e94.
electric vehicle charging stations. IEEE Trans Transport Electrific 2018;4(1): [36] Li L, Dai S, Cao Z, Hong J, Jiang S, Yang K. Using improved gradient-boosted
76e88. decision tree algorithm based on Kalman filter (GBDT-KF) in time series
[10] Zhao T, Li Y, Pan X, Wang P, Zhang J. Real-time optimal energy and reserve prediction. J Supercomput 2020;76(9):6887e900.
management of electric vehicle fast charging station: hierarchical game [37] Shoaran M, Haghi B, Taghavi M, Farivar M, Emami-Neyestanak A. Energy-
approach. IEEE Trans Smart Grid 2018;9(5):5357e70. efficient classification for resource-constrained biomedical applications. IEEE J
[11] Petoussis S, Zhang X, Godfrey K. Electricity market equilibrium analysis based Emerging Selected Topics Circuits Sys 2018;8(4):693e707.
on nonlinear interior point algorithm with complementarity constraints. IET [38] Sun R, Wang G, Zhang W, Hsu L, Ochieng W. A gradient boosting decision tree
Gener, Transm Distrib 2007;1(4):603. based GPS signal reception classification algorithm. Appl Soft Comput
[12] Su J, Lie T, Zamora R. Modelling of large-scale electric vehicles charging de- 2020;86:105942.
mand: a New Zealand case study. Elec Power Syst Res 2019;167:171e82. [39] Harifi S, Mohammadzadeh J, Khalilian M, Ebrahimnejad S. Giza Pyramids
[13] Shekari T, Golshannavaz S, Aminifar F. Techno-economic collaboration of PEV Construction: an ancient-inspired metaheuristic algorithm for optimization.
fleets in energy management of microgrids. IEEE Trans Power Syst Evolutionary Intelligence; 2020.
2017;32(5):3833e41. [40] Gao T, Gong X, Zhang K, Lin F, Wang J, Huang T, Zurada JM. A recalling-
[14] García-Villalobos J, Zamora I, San Martín J, Asensio F, Aperribay V. Plug-in enhanced recurrent neural network: conjugate gradient learning algorithm
electric vehicles in electric distribution networks: a review of smart charging and its convergence analysis. Inf Sci 2020;519:273e88.
approaches. Renew Sustain Energy Rev 2014;38:717e31. [41] Thomas D, Deblecker O, Ioakimidis C. Optimal operation of an energy man-
[15] Franco J, Rider M, Romero R. A mixed-integer linear programming model for agement system for a grid-connected smart building considering photovol-
the electric vehicle charging coordination problem in unbalanced electrical taics' uncertainty and stochastic electric vehicles' driving schedule. Appl
distribution systems. IEEE Trans Smart Grid 2015;6(5):2200e10. Energy 2018;210:1188e206.
[16] ElNozahy M, Salama M. Uncertainty-based design of a bilayer distribution

18