Received 20 December 2022, accepted 27 December 2022, date of publication 23 January 2023, date of current version 30 January 2023.

Digital Object Identifier 10.1109/ACCESS.2023.3238667

Utilization of EV Charging Station in Demand Side

Management Using Deep Learning Method
ABDUL HAFEEZ , RASHID ALAMMARI , (Senior Member, IEEE),
AND ATIF IQBAL , (Senior Member, IEEE)
Department of Electrical Engineering, Qatar University, Doha, Qatar
Corresponding author: Rashid Alammari (ralammari@qu.edu.qa)
This publication was made possible by Qatar University Research grant# [QUCP-CENG-2020-2] from Qatar University, Qatar. The APC
for the article is funded by the Qatar National Library, Doha, Qatar.

ABSTRACT Conventional energy sources are a major source of pollution. Major efforts are being made by
global organizations to reduce CO2 emissions. Research shows that by 2030, EVs can reduce CO2 emissions
by 28%. However, two major obstacles affect the widespread adoption of electric vehicles: the high cost
of EVs and the lack of charging stations. This paper presents a comprehensive data-driven approach based
demand-side management for a solar-powered electric vehicle charging station connected to a microgrid. The
proposed approach utilizes a solar-powered electric vehicle charging station to compensate for the energy
required during peak demand, which reduces the utilization of conventional energy sources and shortens
the problem of fewer EVCS in the current scenario. PV power stations, commercial loads, residential loads,
and electric vehicle charging stations were simulated using the collected real-time data. Furthermore, a deep
learning approach was developed to control the energy supply to the microgrid and to charge the electric
vehicle from the grid during off-peak hours. Furthermore, two different machine learning approaches were
compared to estimate the state of charge estimation of an energy storage system. Finally, the proposed
framework of the demand management system was executed for a case study of 24 hours. The results reflect
that peak demand has been compensated with the help of an electric vehicle charging station during peak
hours.

INDEX TERMS CO2 emission, data-driven approach, deep learning, demand-side management, electric
vehicle charging station, peak clipping.

I. INTRODUCTION generation is conventional. However, to satisfy the rapidly

As environmental problems continue to worsen to the point increasing demand for power, renewable energy systems have
where they threaten the entire globe, it is imperative that been integrated into traditional power grids. Consequently,
humans take immediate action to cut emissions of green- the previously centralized power system is evolving toward
house gases (GHG). Many environmental organizations have more decentralized structures, with an increasing number of
developed plans and policies to reduce CO2 emissions [1]. small units connected to distribution grids. Distributed energy
The use of renewable energy sources and the electrification resources (DER) can supplement central power generation,
of transportation systems are two of the most promising by adding capacity to the utility grid or directly to end-
approaches, being considered as a means to address rising users. However, these renewable power generation systems
environmental concerns and energy supply [2]. Therefore, are climate-dependent and highly stochastic [3]. Therefore,
power generation and transportation networks are shifting during peak hours, the load demand may exceed maximum
towards the utilization of renewable energy and electri- capacity, which could cause instability or even a blackout
cal vehicles (EV). Currently, a major part of the electrical if the balance in the grid is not effectively maintained [4].
The peak demand lasts for a short time, and most of the
The associate editor coordinating the review of this manuscript and time, the generation capacity is not fully utilized. Typically,
approving it for publication was Qiang Li . pump storage or diesel engine plants are utilized to manage

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
VOLUME 11, 2023 8747
 A. Hafeez et al.: Utilization of EV Charging Station in Demand Side Management Using Deep Learning Method

short-term peak loads. Advancements in battery technologies and cost reduction strategies. Raghavan et al. demonstrate
and EV have made it possible to use them as temporary the day-ahead participation of EVs employing large-scale
energy storage devices for peak load management. This is integration in demand response, to optimize the net profit and
because idle time of an EV is significantly longer than its excess power output of independent operators. When making
charging time [5]. Therefore, EVs with other energy storage the model, power generation capacity, the grid, and the power
systems (ESS) play an important role in maintaining the balance are taken into account. Preventing the avalanche
power balance of the power system. The use of a clean energy effect, real-time regulated charging and the leverage acquired
storage system for peak demand management (PDM) leads by such charging methodologies are major benefits to inde-
to a low consumption of fossil fuels and less environmental pendent operators [13]. The resulting grouping of EVs leads
pollution. to a financially beneficial DSM adoption [14]. Rezaee et al.
Grid ancillary services are implemented on a hundred kW discussed the application of plug-in electric vehicles (PEVs)
basis for peak load management of the grid, whereas single in DSM, utilizing parking lots as aggregators. A methodology
EVs can only provide a limited amount of power (10-20 kW). is presented to estimate the impact of EV aggregated in the
The concept of an aggregator was implemented to achieve parking lot on the grid. In terms of bus voltage and grid
large-scale power rating. A third party between EVs and power loss, the results indicated acceptable levels of pene-
power grids, such as EV charging station (EVCS), can be used tration. However, constraints on the availability of vehicles,
as an EV aggregator. When aggregated in suitable numbers, load, and renewable energy sources were not included in this
EVs become a reliable integral part of the DER as a large- research [15]. Tong et al. present a single-family home DSM
scale energy source with the help of a base ESS [6]. setup combining PV and ESS as an energy buffer to relay
The use of electric vehicles significantly contributes to a surplus photovoltaic output during off-peak hours. In this
healthier environment, with the lowest CO2 emissions and study, day-ahead market pricing signals were obtained to
much less noise. Research shows that by 2030, EVs could launch bids, expanding the viability of retired vehicle trac-
reduce CO2 emissions by 28% [7]. However, two major tion batteries in their second life form [16]. Arellano et al.
obstacles affect the widespread adoption of electric vehicles: used a simulation tool to investigate the impact of changing
the high cost of EVs and the lack of charging stations [8]. EV penetration levels into the distribution grid at the DSM
The current problem is that EV owners are awaiting proper with different load patterns. To improve performance and
charging infrastructure; however, EVCS investors are waiting efficiency, an automatic demand response (ADR) approach
for sufficient EV adoption for a profitable business. This was introduced in a test case scenario in the control aspect
situation is like the ‘‘chicken or egg’’ theory [9]. There is a of the smart grid infrastructure. ADR employs a demand
need to increase the profit of the EVCS infrastructure. This response (DR) approach to relay control signals from the
motivates investors to establish EVCSs. A renewable-energy- end user to the utility’s main control unit without the need
integrated EVCS already has an ESS for energy storage. From for human interaction [17]. The ADR technique was also
an economic perspective, EVSC can be utilized for grid peak- implemented by Xiang et al. with the integration of EVs
load management. Moreover, with the help of an EVSC, the and ESSs via blockchain profiling [18]. To achieve optimal
EV system can also be used for peak load management, which management operation of grid-connected buildings, Christos
is a mutual profit for both systems [10]. This approach will et al. suggested a distributed feedback-based optimization
shorten the problem of a lower number of EVCS in the current method based on the principles of approximate dynamic
scenario because EVCS will have a source of earnings other programming. In this study, a multi-criteria approach was
than the charging of EV. An EV charging station aggregated used to minimize energy expenditures without compromising
with EVs and an ESS behaves like a dynamic battery that end-users values. Thermostatically controllable loads, such
is utilized in demand-side management (DSM) during peak as water heaters and clothes dryers, were featured with EVs
hours [11]. EV aggregators act as energy sources and have and ESS [19]. An optimization technique for DSM was devel-
the advantages of no start-up and shut-down costs and fast oped by Zihao Dong et al., which implemented Time-of-Use
response speed in solving the intermittency issue of renew- (ToU) based load shifting at the residential level. Multiple
able energy sources. Fig. 1 shows a potential framework for adaptable home appliances, EVCS, and rooftop PV systems
the integration of EVCS into DSM operations. are taken into account. In this case, optimization led to a
There has been an uptick in the number of utility operator- 19% reduction in daily electricity costs. In addition to low-
implemented programs that use DSM strategies for cost ering the carbon intensity of the grid and providing energy,
and energy management. Participating in DSM programs rooftop PV reduces domestic carbon emissions by 12%. With
can help electric power markets to run more efficiently and the growing use of electric vehicles and renewable energy,
profitably [12]. DSM schemes that incorporate EVs, aim to the smart scheduling of household loads has a substantial
maximize customer and utility profitability while minimizing impact on grid resilience and energy efficiency [20]. Ran
overall system losses. Existing research incorporates a variety et al. came up with a two-stage planning optimization model
of financial models, with a focus on financial objectives. Gen- for shared EVs to investigate the efficient coordination of
erally, they employ energy efficiency, energy management, shared EV operations under a demand response. The location

8748 VOLUME 11, 2023

A. Hafeez et al.: Utilization of EV Charging Station in Demand Side Management Using Deep Learning Method

FIGURE 1. A potential framework for the integration of EVCS into DSM operation.

of the charging facility was the first step, followed by the household and commercial loads are not considered in this
relocation of vehicles in the second step. In addition, both study. Modeling the charging load of an EV is challenging
supply-side and demand-side uncertainties were taken into owing to its complexity. However, it provides a framework
account and approximated into a form that was manageable for future research on the impact of EVCS on demand-side
by using a sample average approximation. This policy made management. Yang et al. proposed a set of equations to build
easier to make decisions under, an uncertain charging ability. a probabilistic load model. The method of parameter identi-
However, more DR elements, like power generation (solar fication was based on ant colony algorithms. A real battery-
and wind energy), are not considered [21]. Shariff et al. swapping charging station was used to identify and simulate
presented the design aspects and practical implementation the suggested concept. The findings demonstrated that the
of a modern solar-powered EVCS, controlled by a Type-1 model’s applicability and accuracy were satisfactory in the
vehicle connector. The designed model was built in MAT- case of a battery-swapping EVCS [25].
LAB/Simulink. Its circuit functioning and methodical model Moghaddam et al. presented a dynamic pricing model for
were analyzed to establish the parametric design character- peak load adjustment. To minimize the overlap between the
istics. A complete hardware configuration was created to PEV and residential peak load periods, a constraint opti-
demonstrate the power factor adjustment with varying steady- mization problem was formulated and optimized using a
state loads. However, research has mainly focused on the heuristic solution [26]. Numerous factors must be considered
design aspect of solar-connected EVCS. The effect of EVCS for the proper implementation of DSM programs, including
on DSM was not analyzed in this study [22]. To improve load and power forecasts, State of Charge (SoC) estimation,
the quality of service at public stations and increase their identification of appropriate consumers to engage in schemes,
utilization, [23] presented a data-driven performance analysis and developing automated systems that manage demand-side
method for the charging behavior of EV at charging stations. resources [27]. Nowadays, model predictive control (MPC)
The purpose of the data-driven model was to be adaptive to and data-driven methods are used for dynamic nonlinear
any type of issue that the EVCS can face to find a better system design of power systems. In the literature, data-driven
solution. However, the study did not consider the constraints approaches are used in power systems for different practices,
of other systems connected to charging stations. The author such as optimal charging [28], economic dispatch [29], uncer-
of [24] proposed an EV charging network as a cyber-physical tainties of renewable energy resources [30], and EV and SoC
system (CPS). The EVCS is regarded as a component of the prediction of ESS [31]. For the proper operation of a grid-
micro grid (MG), whereas EVs are part of the transportation connected, PV-integrated EVCS, the aggregator needs knowl-
network. The findings showed that the suggested algorithms edge of the SoC level of the ESS and EVs participating in the
optimized the charging costs and balanced the regional load plan. In addition, the reserved energy level of each system
profiles and EV charging behaviors successfully. However, participates in load management. Commonly used methods
the limitations of batteries, renewable energy resources, and of SoC estimation are mathematical modeling-based [32].

VOLUME 11, 2023 8749

 A. Hafeez et al.: Utilization of EV Charging Station in Demand Side Management Using Deep Learning Method

These models have less accuracy due to the nonlinear and exponential smoothing [50] have been used for a long
complex nature of the battery. The majority of these mod- time to predict load and price in DR. This kind of model
els can only function under certain conditions, such as a is usually linear, and has been found to be less accurate at
specific battery type and a fixed temperature. When other forecasting load [51]. The lower prediction accuracy of clas-
factors are considered, new models should be developed. sical approaches can be related to their linearity assumptions.
Machine-learning (ML)-based methods have been used to Therefore, ANNs, which can approximate highly nonlinear
overcome these limitations. ML-based methods can model relationships, have been used for load and price forecasting in
the nonlinearity of a system using collected data [32]. Com- DSM. In addition, as demand becomes increasingly nonlinear
mon machine learning techniques include, fuzzy logic [33], and variable, DL methods will likely produce even more
support vector machine [34], and artificial neural networks accurate load and price forecasts [27].
(ANNs) [35], [36]. The existing research survey indicates that numerous tech-
Machine learning techniques have been utilized for high- niques and optimization algorithms have been developed and
complexity global optimization issues [36] and have shown implemented in EV DSM programmers to address optimiza-
positive results in peak demand reduction [37]. A genetic tion in the SG environment. These strategies and algorithms
algorithm (GA) based power mitigation method was investi- may have single or multiple objectives, as well as a sin-
gated in paper [38]. GA is used to automate the optimization gle or hybrid strategy. Linear programming (LP), dynamic
of demand-side management. As a case study for minimizing programming (DP), fuzzy computation, particle swarm opti-
harmonic distortion for all 24-hour time steps, a basic indus- mization (PSO), genetic algorithm (GA), differential evolu-
trial grid with five machines was utilized. The drawback of tion (DE), ant colony optimization (ACO), stochastic opti-
this study is that only the total harmonic distortion (THD) mization artificial, game theory algorithm and neural network
of the voltage and total demand distortion (TDD) of the (ANN) are the most widely used optimization algorithms
load current are utilized to optimize operational schedules to current employed in the domain of EV DSM optimization
reduce harmonic distortion in the industrial grid. problems. As for the current research directions, the algo-
The Neural network-based deep-learning (DL) methods rithms used to schedule and control devices must be more
have received considerable attention from the research com- flexible and able to work in a more diverse environment.
munity in recent years. Deep learning is the process of learn- In this research, renewable energy and EV transportation
ing several levels of representation and abstraction and is systems integrated power systems require a DSM framework
capable of analyzing data in its raw format as well as dis- that is automated, can adapt to a changing environment and
covering the representations that are required for detection or can learn what users need. For the large-scale uncertain-
classification [39]. Deep neural networks can be built in many ties present in this type of integrated model, a data-driven
different ways. Most often, feed-forward NNs [40], convolu- approach integrating machine learning with optimization
tional NNs [41], and recurrent neural networks (RNN) [42] needs to be adapted [21]. Although DL approaches are used
are used for supervised learning, whereas autoencoders [43] in public charging station occupancy prediction to reduce
and Restricted Boltzmann Machines [44] are used in unsu- electric vehicle operator and user inconvenience [52], load
pervised learning. Deep learning has the potential to learn forecasting [53], [54], [55], [56], their direct implementation
highly nonlinear, complicated relationships, and correlations for peak load management in an MG integrated with RE and
between input and output data in comparison with traditional EV transportation system cannot be found.
techniques. Because of this, the DSM literature shows that The research contributions of this paper mainly include
deep learning methods are usually better at making predic- deep learning for optimal scheduling of loads and electrical
tions than traditional methods, such as SVR [45], [46], [47], transportation networks for peak load management in renew-
shallow ANNs [45], and Random Forest [45]. Furthermore, able energy-oriented power systems. It focused on creating
it is unclear why they do so well in particular sorts of a DL-based peak load management system for an MG that
problems [48], and it should be emphasized that arbitrarily is integrated with renewable resources and an EVCS in the
increasing the depth of an ANN may not necessarily pro- second stage of control. However, the first stage of con-
duce the best results [49]. Long short-term memory (LSTM) trolling EVs and PV systems integrated into the electrical
networks, which are a form of RNN, are able to manage transportation system was performed. Moreover, the research
better long-term dependencies at the expense of higher com- considered a data-driven approach for modeling the differ-
putational costs. By contrast, convolutional neural network ent components of the proposed transportation network to
(CNN) networks are ideally suited for processing data with a overcome the nonlinearities of the components. However,
grid-like architecture. in the case of mathematical modeling of components, some
Demand response in smart grids is moving toward a future variables are very difficult to implement in real-time.
in which end-user loads can be controlled in a detailed In the comparison of the discussed literature the key con-
manner. This means that load and price forecasts must be tribution of this manuscript are as follows:
more accurate. Time-series models such as autoregressive • In this work, a comprehensive data-driven approach-
(AR), auto-regressive integrated moving average (ARIMA), based DSM for an EVCS-connected MG is proposed.

8750 VOLUME 11, 2023

A. Hafeez et al.: Utilization of EV Charging Station in Demand Side Management Using Deep Learning Method

FIGURE 2. Techniques of demand side managements in power system.

• A data-driven approach is used to incorporate the PV out additional energy demand [53]. The proposed methodology
in the proposed EVCS-connected MG. utilizes an electrical transportation system for peak-demand
• Based on the real-time commercial and residential col- management. In the subsequent section, the modelling and
lected load data, the load connected to the EVCS-based control of the constituents of the proposed methodology are
MG is forecasted. discussed
• The SoC estimation of the ESS system present in EVCS
is performed by the LSTM method. Comparative result A. PROPOSED SYSTEM MODELLING
of LSTM with Vector Autoregressive moving average This work considers a PV-connected EVCS for grid peak
(VARIMA) shows its suitability. demand management as well as EV charging station profit
• To execute the proposed DSM, a deep learning-based improvement. The proposed system contains a conventional
efficient controller is designed and validated through a power generator, residential loads, critical commercial loads,
case study of 24-hour power consumption. renewable energy sources, and an EVCS. The EVCS unit
The rest of the paper is organized as follows: Section II contains the ESS and EVs, which are utilized as charg-
describes the framework of the proposed demand-side ing/discharging loads. A two-stage framework is presented
management strategy, modeling, and their constraints for power coordination between the EVCS and grid, as shown
are explained in the related subsections. In Section III, in Fig. 3. The first stage represents a dynamic energy source
a case study of demand-side management using data-driven model of the ESS and EVs integration in the direct mode.
approaches is presented. In addition, the results and analysis In the second stage, the power allocations for each unit are
of deep-learning approaches for SoC estimation are pre- presented, including the ESS and PV systems. PV and load
sented. Finally, the conclusions of the results and analysis are significant sources of uncertainty in the MG. In the pro-
are presented in Section IV, followed by references. cess of power distribution among the units, accurate modeling
of these unknown factors plays a critical role. To overcome
II. METHODOLOGY the effect of these uncertainties, a deep neural networks
Demand side management is an area of energy management (DNN) based controller is used, for the required energy flow
that focuses on the monitoring and management of peak between the two stages, for the optimal cost of energy con-
demands, as well as the smoothing of the load profile through- sumed by the ESS.
out the day [57]. DSM is required to reduce power plant
capital expenditure and enhance the financial performance of 1) MICRO GRID MODELING
electrical utilities [58]. As shown in Fig. 2, DSM methods The modelling objective is to maximize the total profit of the
can be categorized into six major categories: peak clipping, EVCS during the dispatch time to minimize the overall cost,
valley filling, load shifting, strategic conservation, strategic which can be expressed as in Eq. (1).
load growth, and flexible load shape [59].Utilizing an energy
storage system for demand-side management is one of the kn
X
most effective and promising strategies for peak clipping. min Cp = (CGRIDs (k) + CGRIDc (k) + CEVcc (k)
The ESS stores energy during off-peak load conditions and K=k0
supplies it to the grid during peak hours to compensate for + CEVdg (k) + CESSdg (k)) (1)

VOLUME 11, 2023 8751

 A. Hafeez et al.: Utilization of EV Charging Station in Demand Side Management Using Deep Learning Method

FIGURE 3. Energy communication diagram between first stage and second stage.

CGRIDs (k) is the grid tariff during off-peak load, which is 2) ENERGY STATE OF THE STORAGE ELEMENT (SOC)
defined in Eq. (2). The state of charge is a representation of energy available in
an energy storage system. The model of an ESS in terms of
CGRIDs (k) = βgs (k) Pgs (k) 1k (2) SoC can be represented as in Eq. (7).
CGRIDc (k) is energy cost supply by EVCS during the off- SoCEVCS (k + 1)
SoCEVCS (k) + PEVCS (k) 1kη
 
peak load, which is defined in Eq. (3).

CSCH


 EEVCS
if PEVCS (k) ≥ 0

CGRIDc (k) = βCSs (k) |PCS (k)| 1k

(3) 
=   (7)
1k
where Pgs (k) and PCS (k) are the power consumed and sup- SoC (k) + P (k) otherwise

EVCS EVCS EEVCS ηCSDC



plied by the EVCS from the main grid respectively. βgs (k)


and βCSs (k) are the energy sale and purchase prices from the
grid respectively and 1k represents the time period. where SoCEVCS (k) is the SoC of EVCS at time k, PEVCS (k)
CEVcc (k) is the energy cost during EV charging, and isnde- represents the dispatched power from EVCS while ηCSCH
fined in Eq. (4). and ηCSDC are charging and discharging efficiencies. EEVCS
represents the EVCS energy demand.
CEVcc (k) = βEVcc (k) |PEVC (k)| 1k (4) The model represented in Eq. (7) has constrains that are
represented in Eq. (8), Eq. (9), and Eq. (10).
where PEVC is the power consumed by the EV during charg-
EVCS ≤ SoCEVCS (k) ≤SoCEVCS
SoCmin max
(8)
ing. βEVcc (k) is the energy sale price of the grid during EV
charging. Pmin
EVCS ≤ PEVCS (k) ≤Pmax
EVCS (9)
CESSdg (k) andCEVdg (k) are ESS and EV battery degrada- maxCS
PGRID ≤
maxgd
PGRID (k) ≤PGRID (10)
tion cost, which are defined in Eq. (5) and Eq. (6).
where; SoCmin max
EVCS and SoCEVCS are denoting the minimum
CESSdg (k) = βESdg (k) |PES (k)| 1k (5) and maximum SoC of EVCS. Pmin max
EVCS and PEVCS are minimum
CEVdg (k) = βEVdg (k) |PEVd (k)| 1 k (6) and maximum charging power required, respectively. PGRID
repersents the grid capacity limit. Further, the maximum
where PES (k) and PEVd (k) are overall power consumed by the possible power flow between the grid and EVCS in terms of
ESS and EV. βESdg and βEVdg are the average charging cost the EVCS to the grid and grid to the EVCS are PmaxCS
GRID and
maxgd
for ESS and EV respectively. PGRID .

8752 VOLUME 11, 2023

A. Hafeez et al.: Utilization of EV Charging Station in Demand Side Management Using Deep Learning Method

3) EVCS MODELING where; ŷ is the forecasted output, w0 is the bias term, and the
The size of the power coordination problem will be reduced weight matrix for x is w and power m is any real number.
by treating EVCS as a single unit rather than the combination Commonly used regression techniques for nonlinear sys-
of PV unit, ESS and multiples EVs. An EVCS demand model tems are ARMA) and ARIMA [19]. A moving average model
is constructed to take all EV’s charging demand and tech- (MA) with order q and an autoregressive model (AR) with
nological limitations into account. In this section, an EVCS order p constitute the two basic components of ARMA, which
demand model is developed, considering all energy charging are represented in Eq. (16),
and discharging systems and their technological constraints. p q
X X
It was assumed that EV owners will share the required infor- ŷt = c + εt + ϕi yt−i + θi εt−i (16)
mation such as SoC, departure, and arrival time. At each i=1 i=1
time slot, the EVCS also gathers the most recent SoC data of
where ϕ1, . . . . . . , ϕp represents the weights for the autore-
energy sources. Eq. (11) and Eq. (12) represent the constraints
gressive model, θ1, . . . . . . , θ q denotes the weights for the
for EVCS operation with a time slot.
X moving average model, ε is white noise and c denotes a
EVCS (k) =
Pmax Pmax max constant. ARMA models were used for stationary time-series
EV,n + PESS,n (11)
nϵNk data prediction. Whereas, the performance of the arima model
X is better for nonstationary time series data in comparison to
EVCS (k)
Pmin Pmin min
= EV,n + PESS,n (12)
the arma model. Currently, advanced arima is used for the
nϵNk
prediction of more complex systems that can extract features
where Pmax max
EV,n and PESS,n are the maximum power supply from many variables [61]. Although, models based on statis-
limits of nth EV and ESS, respectively, and Pmin min
EV,n and PESS,n tical methods perform well; however, in the event of a sudden
represents minimum limit of power that can be supplied by change in the model’s attributes or the presence of a statistical
nth EV and ESS, respectively. error in the data, deficiencies emerge and prediction accuracy
The total energy that can be supplied by the ESSs, con- decreases. This has a significant impact on the load patterns
nected to the EVCS can be presented in Eq. (13) and the and profiles [60]. Using ann-based techniques, this type of
energy supply by the EVs for the entire time slot can be problem can be handled and accurate predictions can be made
defined in Eq. (14). without any loss of precision.
N=n
X
EESS (k) = EESS,n (k) (13) 2) DEEP LEARNING METHODS
N=1 A mathematical representation of a biological neuron is
N=n
X known as an artificial neuron. It depicts nonlinear responses
EEV (k) = (SoCEV,n (k + 1) − SoCEV,n (k))EEV,n (k) to input signals, similar to those of biological neurons [56].
N=1 The artificial neural network structure is based on a net-
(14) work of artificial neurons, known as a multilayer perceptron
(MLP). The multilayer perceptron is a three-layered structure
where; EESS,n (k) and EEV,n (k) are total energy supply nth
that includes input, hidden, and output layers. Each layer
ESS and EV respectively at time k. SoCEV,n (k) is the SoC
arranges neurons such that there are no intra layer connec-
level of the nth EV at time k.
tions between neurons. However, neurons have full-weighted
connections between layers. A typical ANN structure depicts
B. DEEP-LEARNING BASED SoC ESTIMATION
the artificial neuron’s structure and describes the mechan-
1) DATA-DRIVEN APPROACHES FOR VARIABLE ESTIMATION
ics of the neuron’s nonlinear reactions in response to input
Smart grids use many smart meters that send information signals.
about user behavior to a central server. Using data-driven An ANN structure is trained with the help of a dataset to
techniques, these data can be processed to obtain insights learn the weights and bias with the appropriate number of
useful for MG analysis and control. Forecasting the future neurons, hidden layers, and activation functions as shown in
load (electricity consumption) and SoCs of energy-storing Eq. (17). ANN with two hidden layers is capable of training
elements are crucial tasks for achieving grid intelligence. with arbitrary accuracy. If the number of hidden layers is
If these forecasts are accurate, a utility provider will be able increased it is called deep learning.
to plan resources and take measures to balance the supply
and demand of electricity. These forecasts can be made using ŷ = ∅ (wout h + bout ) = ∅ [wout σ (wx + b) + bout ] (17)
data-driven predictive models. Fig. 4 shows the various types
where; ∅ is the activation function of the output layer; repre-
of data-driven approaches. A popular statistical technique for
sents the output of the hidden layer, h =(wx+b); σ is the acti-
investigating the relationship between input variables and a
vation function for the hidden layer; wout and w are the weight
response variable is linear regression. Eq. (15) depicts its
matrix; bout and b are the bias terms. Deep neural networks,
general form:
convolutional neural networks, and recurrent neural networks
ŷ = w0 + wxm (15) are three categories of ANN-based deep learning methods.

VOLUME 11, 2023 8753

 A. Hafeez et al.: Utilization of EV Charging Station in Demand Side Management Using Deep Learning Method

FIGURE 4. Types of data driven approaches, machine learning and statistical models.

The DNN is a fully connected feedforward network with- time-series model that uses multiple features to estimate the
out lopping back as shown in Fig. 5(a). DNN models have output variable. The purpose of this model is to filter out
two common issues: overfitting and computational complex- meaningful patterns in the series to predict the next value.
ity. Convolutional Neural Networks, or CNNs, are a special LSTM stores information about previous inputs in hidden-
class of DNNs that use the convolutional layer architecture cell memories, making it more suitable for handling time-
(depicted in Fig. 5(b)). CNN reduces the likelihood of over- series data. Data were obtained from a public database for
fitting by simplifying the network structure and decreasing its modeling a battery and preprocessed for unfitting and missing
connectedness scale. RNNs differ from other deep learning information. The final preprocessing step divided the given
algorithms in that their structure includes loops (illustrated dataset into training and test data. After data preparation,
as the cycle in Fig. 5(c)), which allows information to flow in appropriate data-driven algorithms were chosen and trained.
both directions. An RNN can remember information from the Mean square error (MSE) is used to evaluate the overall loss
previous state, making it the most effective technique for fore- function at the end of each forward pass during the training
casting tasks. On the other hand, traditional RNN gradients process. MSE can be represented in mathematical form as in
tend to explode or vanish when the loop is run many times Eq.(18).
because the weight of the loop is constant across all time K
steps. This is called ’’long dependency’’. One widely used 1X
2
MSE = yk − ŷk (18)
RNN model, LSTM, can be applied to remember information K
k=1
for a long period of time.
where; yk is the true SoC value while ŷk is the output of
the proposed network at time k. To minimize the total loss,
3) SoC ESTIMATION
the Adam optimizer is used, which changes the biases and
The Proper operation of the MG requires energy-storing com-
network weights on the loss function’s gradient. In the LSTM
ponents for an automatic DSM scheme. As the EV charging
layer, a dropout rate is employed to account for possible
behavior is typically complex and the battery degrades with
overtraining during the training phase. The root mean square
repeated charge and discharge cycles, a battery management
error (RMSE) is used in the testing phase to evaluate the
system (BMS) is required to monitor the battery health sta-
performance which can be represented as in Eq.(19).
tus and protect it from overcharging and over discharging. v
One of the key states in the BMS is the state-of-charge, u
u1 X K

2
which reflects the remaining battery charge after one charge- RMSE = t yk − ŷk (19)
discharge cycle. The SoC can only be accurately estimated K
k=1
using current, voltage, temperature, and additional measur-
able variables. A new updated version of the ARIMA network 4) DATA-DRIVEN LOAD FORMULATION
is chosen to model the highly nonlinear dynamics of batteries Load is a crucial aspect of dsm. for utilities, it is typically
and estimate the battery SoC from measurable voltage, cur- the most practical to view dsm in terms of general load-
rent, and temperature variables. VARIMA is a multivariable shaping goals. The load shape represents the time of day,

8754 VOLUME 11, 2023

A. Hafeez et al.: Utilization of EV Charging Station in Demand Side Management Using Deep Learning Method

FIGURE 5. The basic structure of (a) DNN, (b) CNN, and (c) RNN.

weekday, and seasonal distribution of electricity consump-

tion. Consequently, the dsm on the grid requires accurate
load modeling and forecasting. Mathematical modeling of
residential and commercial demands has become a challeng-
ing task. Load profiles exhibit variability and uncertainty
in behavior [61]. A data-driven approach has better accu-
racy than mathematical modeling. Multiple linear regres-
sion, stochastic time series, general exponential smoothing,
state space method, and ann-based approach are frequently
employed load forecasting techniques. First, in a data-driven
process, raw information is gathered from some source, like
a computer simulation, a survey, or a publicly available
database. Following data preparation, appropriate data-driven
algorithms are chosen and trained. The accuracy of forecasted
load will depend upon the quality of collected data and the
validation efficiency selected methods.

C. DL-BASED CONTROLLER
An Intelligent controller is needed for controlling vari-
able loads and charging/discharging of the different energy
sources participating in demand-side management. A deep
learning based controller is chosen for the energy flow
between the MG and EVCS system. The flowchart of control
logic for peak load management is shown in Fig. 6. The logic
checks if the overall Preload on the grid (Pload ) is greater
than or equal to specify peak load limit (Pmin ) and if the
time is within the specified limit by the grid (Tpmin ≤ T ≤
Tpmax ). Moreover, if the energy level of the EVCS system is
above the threshold level then only it participates in the peak
load management. Once, T ≥ Tpmax then check, SoCEV ≥
FIGURE 6. Flowchart of control logic for grid peak load management.
SoCEVmin or SoCESS ≥ SoCEVmin and charge the energy
storage system only if, Pload is less than the Pmin .

III. CASE STUDY RESULTS AND DISCUSSION A. SoC ESTIMATION USING DL METHODS
To formulate the micro grid, industrial and commercial load The data of an energy storage system are analyzed in Fig. 7,
data were collected from a local real power substation. Data which demonstrates how the various parameters of the system
for energy source modeling has been collected from an are correlated with one another. A scatter plot illustrates the
open online data source https://data.nasa.gov/dataset. The relationship between the state of charge of a battery and
proposed method was simulated using MATLAB 2020a. the input state variables such as the current, voltage, and
Load models for residential and commercial loads were sim- temperature.
ulated by using collected data and forecasted for a time We found that the SoC was strongly dependent on the
duration with the help of the deep learning time series parameters utilized during the training of the DL model.
method. A deep learning model was trained using these parameters,

VOLUME 11, 2023 8755

 A. Hafeez et al.: Utilization of EV Charging Station in Demand Side Management Using Deep Learning Method

FIGURE 7. Correlation between the different variables of the energy source.

FIGURE 8. Change in parameter of the ESS with respect to time.

and the desired energy source model parameters with respect the modeled ESS were almost constant with time, and the
to time are shown in Fig. 8. The voltage and temperature of SoC level of the ESS changed without changes in the supply

8756 VOLUME 11, 2023

A. Hafeez et al.: Utilization of EV Charging Station in Demand Side Management Using Deep Learning Method

FIGURE 9. Performance accuracy of the ML algorithms.

FIGURE 12. Forecasted load profile of commercial load 3 for 24 hours.

FIGURE 10. Forecasted load profile of commercial load1 for 24 hours.

FIGURE 13. Forecasted load profile of residential load1 for 24 hours.

FIGURE 11. Forecasted load profile of commercial load 2 for 24 hours. FIGURE 14. Forecasted load profile of residential load 2 for 24 hours.

voltage and current. The results presented in Fig. 8 were

obtained from a deep learning model using the LSTM algo-
rithms. Initially, two different machine-learning approaches,
VARIMA and LSTM, were used to model the EVCS, and
their performances were compared, as shown in Fig. 9. The
LSTM network provided a better estimation with an RMSE of
0.49 % compared to the VARIMA network with 0.87 %. The FIGURE 15. Forecasted load profile of total load connected to microgrid.
for 24 hours.
LSTM validation accuracy was better than that of VARIMA
Therefore, LSTM was chosen for the SoC estimation of the
energy source. After that, the estimated SoC of ESS is given
to the controller for peak load management of the MG

B. LOAD PROFILE ANALYSIS

Commercial and residential loads connected to the MG were
simulated using the collected load data. Three commercial
FIGURE 16. Power profile of EV charging station for 24 hours.
loads were simulated, and their behaviors were similar to
those of the real loads connected to a grid. Power consump-
tion patterns of commercial loads are shown in Fig.10 to
Fig. 12.
The overall load connected to the grid is shown in Fig.15.
Moreover, EVCS power consumption is shown in Fig 16.
It shows that EVCS consumes power during off-peak-load
and delivers power during the peak load.
FIGURE 17. Profile of solar power station for 24 hours.

C. GRID PROFILE
An MG was simulated using a 120 kW generator, 10 kW shown in Fig. 10–12. The residential loads are shown in
solar power source, three commercial loads, two residential Fig. 13 and 14. The charging and discharging behaviors of
loads, and an EVCS. The profiles of commercial loads are the EVCS are displayed in Fig. 16, which shows that the

VOLUME 11, 2023 8757

 A. Hafeez et al.: Utilization of EV Charging Station in Demand Side Management Using Deep Learning Method

efficient deep learning controller. Initially, the power sup-

ply by the conventional source was above 100 kW between
10 am to 5 pm. Moreover, supply at peak load was up to
182 kW. Analysis of results shows that power supply by the
conventional power source is confined below 100 kW after
the DSM using a designed controller. Also, it was analyzed
that the proposed approach is less complex, more efficient
FIGURE 18. Profile of conventional generator without peak load
compensation for 24 hours. and modeling of the system is more realistic as compared to
conventional mathematical modeling methods. The proposed
method reduces the GHG emission, increases the reliability of
the DER microgrid, and increases the profit of EVCS which
will motivate the investor to establish the new EV charging
station.
Future research could combine this more realistic model
with other intelligent operational models to increase EVCS
FIGURE 19. Profile of conventional generator with peak load utilization by bringing it closer to real-world scenarios. More-
compensation using EVCS for 24 hours. over, this model can be used in conjunction with algorithms
to assist operators with the placement and sizing of EVCS in
EVCS consumes power during off-peak hours and supplies order to improve operations.
power to the grid during peak hours. Fig. 17 characterizes the ACKNOWLEDGMENT
power profile of the solar system. Fig. 18 characterizes the This publication was made possible by Qatar University
power supply by the conventional power generator without Research grant# [QUCP-CENG-2020-2] from the Qatar Uni-
peak load. To reduce the power supply from the conventional versity. The statements made herein are solely the responsi-
source, the EVCS supplies power to the grid, and the result bility of the authors. The APC for article is funded by the
is shown in Fig. 19. It was found that the power shared Qatar National Library, Doha, Qatar.
by the conventional source was high, reaching 182 kW in REFERENCES
the case of peak demand. The profile of the conventional [1] F. Chien, C.-C. Hsu, I. Ozturk, A. Sharif, and M. Sadiq, ‘‘The role of renew-
power source was modified by the proposed demand-side able energy and urbanization towards greenhouse gas emission in top Asian
countries: Evidence from advance panel estimations,’’ Renew. Energy,
management framework and the power supply by the con- vol. 186, pp. 207–216, Mar. 2022, doi: 10.1016/J.RENENE.2021.12.118.
ventional generator is confined below 100 kW. It successfully [2] X. Eric Yu, Y. Xue, S. Sirouspour, and A. Emadi, ‘‘Microgrid and trans-
compensated the peak load demand between 10 to 17 hours. portation electrification: A review,’’ in Proc. IEEE Transp. Electrific. Conf.
Expo. (ITEC), Jun. 2012, pp. 1–6, doi: 10.1109/ITEC.2012.6243464.
The central DSM controller receives load forecast data, SoC [3] S. Talari, M. Shafie-khah, G. J. Osório, J. Aghaei, and J. P. S. Catalão,
information, PV generation, and day-ahead demand, and then ‘‘Stochastic modelling of renewable energy sources from operators’ point-
performs the required actions for the energy flow between of-view: A survey,’’ Renew. Sustain. Energy Rev., vol. 81, pp. 1953–1965,
Jan. 2018, doi: 10.1016/J.RSER.2017.06.006.
the MG and EVCS. The EVCS participates in the DSM [4] N. Etherden and M. H. J. Bollen, ‘‘Overload and overvoltage in low-voltage
program and receives signals via two-way communication and medium-voltage networks due to renewable energy—Some illustrative
during hourly dispatch. Consequently, a peak-shaping DSM case studies,’’ Electric Power Syst. Res., vol. 114, pp. 39–48, Sep. 2014,
doi: 10.1016/J.EPSR.2014.03.028.
strategy is implemented to reduce system peak and total [5] P. Kou, D. Liang, L. Gao, and F. Gao, ‘‘Stochastic coordination of plug-
operating costs. in electric vehicles and wind turbines in microgrid: A model predictive
control approach,’’ IEEE Trans. Smart Grid, vol. 7, no. 3, pp. 1537–1551,
IV. CONCLUSION May 2016, doi: 10.1109/TSG.2015.2475316.
[6] C. Pang, M. Kezunovic, and M. Ehsani, ‘‘Demand side manage-
Considering the alarming situation due to on-road IC engine- ment by using electric vehicles as distributed energy resources,’’ in
based vehicles and other GHG emissions. In this manuscript, Proc. IEEE Int. Electric Vehicle Conf., Mar. 2012, pp. 1–7, doi:
a comprehensive framework of a data-driven-based demand- 10.1109/IEVC.2012.6183273.
[7] N. Adnan, S. M. Nordin, M. A. B. Bahruddin, and M. Ali, ‘‘How trust
side management scheme for electric vehicle-connected can drive forward the user acceptance to the technology? In-vehicle tech-
micro gird is proposed. Loads connected to the MG were nology for autonomous vehicle,’’ Transp. Res. A, Policy Pract., vol. 118,
forecasted using the deep learning time-series method. Data pp. 819–836, Dec. 2018, doi: 10.1016/J.TRA.2018.10.019.
[8] F. Ahmad, A. Iqbal, I. Ashraf, M. Marzband, and I. Khan, ‘‘Optimal
required to train the DL method were collected from a power location of electric vehicle charging station and its impact on distribution
substation. Moreover, modeling of PV connected to the MG network: A review,’’ Energy Rep., vol. 8, pp. 2314–2333, Nov. 2022, doi:
was performed using a data-driven approach. The EVCS 10.1016/J.EGYR.2022.01.180.
[9] R. Wolbertus, S. Jansen, and M. Kroesen, ‘‘Stakeholders’ perspectives
modeling was done by considering it as a single energy source on future electric vehicle charging infrastructure developments,’’ Futures,
and the SoC of the associated EV battery was estimated using vol. 123, Oct. 2020, Art. no. 102610, doi: 10.1016/J.FUTURES.2020.
the deep learning method. The LSTM model was used for 102610.
[10] T. Jiang, ‘‘Multi-objective optimal scheduling method for regional
SoC estimation because its validation efficiency was found photovoltaic-storage-charging integrated system participating in demand
to be 99.51 %, which is better than VARIMA model with response,’’ in Proc. 8th Renew. Power Gener. Conf. (RPG), 2019, pp. 1–8,
99.13%. Finally, DSM was achieved using the proposed doi: 10.1049/CP.2019.0383.

8758 VOLUME 11, 2023

A. Hafeez et al.: Utilization of EV Charging Station in Demand Side Management Using Deep Learning Method

[11] A. Ul-Haq, C. Cecati, and E. Al-Ammar, ‘‘Modeling of a photovoltaic- [29] Y. Guo, J. Xiong, S. Xu, and W. Su, ‘‘Two-stage economic operation of
powered electric vehicle charging station with vehicle-to-grid implementa- microgrid-like electric vehicle parking deck,’’ IEEE Trans. Smart Grid,
tion,’’ Energies, vol. 10, no. 1, p. 4, Dec. 2016, doi: 10.3390/EN10010004. vol. 7, no. 3, pp. 1703–1712, May 2016, doi: 10.1109/TSG.2015.2424912.
[12] L. Tronchin, M. Manfren, and B. Nastasi, ‘‘Energy efficiency, demand side [30] Y. Zhang, L. Fu, W. Zhu, X. Bao, and C. Liu, ‘‘Robust model pre-
management and energy storage technologies—A critical analysis of pos- dictive control for optimal energy management of island microgrids
sible paths of integration in the built environment,’’ Renew. Sustain. Energy with uncertainties,’’ Energy, vol. 164, pp. 1229–1241, Dec. 2018, doi:
Rev., vol. 95, pp. 341–353, Nov. 2018, doi: 10.1016/J.RSER.2018.06.060. 10.1016/J.ENERGY.2018.08.200.
[13] S. S. Raghavan, ‘‘Impact of demand response on electric vehicle charging [31] M.-F. Ng, J. Zhao, Q. Yan, G. J. Conduit, and Z. W. Seh, ‘‘Predicting
and day ahead market operations,’’ in Proc. IEEE Power Energy Conf. at the state of charge and health of batteries using data-driven machine
Illinois (PECI), Feb. 2016, pp. 1–7, doi: 10.1109/PECI.2016.7459218. learning,’’ Nature Mach. Intell., vol. 2, no. 3, pp. 161–170, Mar. 2020, doi:
[14] M. Kühnbach, J. Stute, and A.-L. Klingler, ‘‘Impacts of avalanche effects 10.1038/s42256-020-0156-7.
of price-optimized electric vehicle charging–does demand response make [32] F. Yang, Y. Xing, D. Wang, and K.-L. Tsui, ‘‘A comparative study
it worse?’’ Energy Strategy Rev., vol. 34, Mar. 2021, Art. no. 100608, doi: of three model-based algorithms for estimating state-of-charge
10.1016/J.ESR.2020.100608. of lithium-ion batteries under a new combined dynamic loading
[15] S. Rezaee, E. Farjah, and B. Khorramdel, ‘‘Probabilistic analysis of plug- profile,’’ Appl. Energy, vol. 164, pp. 387–399, Feb. 2016, doi:
in electric vehicles impact on electrical grid through Homes and parking 10.1016/J.APENERGY.2015.11.072.
lots,’’ IEEE Trans. Sustain. Energy, vol. 4, no. 4, pp. 1024–1033, Oct. 2013, [33] A. J. Salkind, C. Fennie, P. Singh, T. Atwater, and D. E. Reisner, ‘‘Deter-
doi: 10.1109/TSTE.2013.2264498. mination of state-of-charge and state-of-health of batteries by fuzzy logic
[16] S. Tong, T. Fung, and J. W. Park, ‘‘Reusing electric vehicle battery for methodology,’’ J. Power Sources, vol. 80, nos. 1–2, pp. 293–300, Jul. 1999,
demand side management integrating dynamic pricing,’’ in Proc. IEEE Int. doi: 10.1016/S0378-7753(99)00079-8.
Conf. Smart Grid Commun. (SmartGridComm), Nov. 2015, pp. 325–330, [34] J. c. A. Antón, P. J. G. Nieto, C. B. Viejo, and J. A. V. Vilán, ‘‘Sup-
doi: 10.1109/SMARTGRIDCOMM.2015.7436321. port vector machines used to estimate the battery state of charge,’’ IEEE
[17] B. Arellano, S. Sena, S. Abdollahy, O. Lavrova, S. Stratton, and Trans. Power Electron., vol. 28, no. 12, pp. 5919–5926, Dec. 2013, doi:
J. Hawkins, ‘‘Analysis of electric vehicle impacts in new Mexico urban 10.1109/TPEL.2013.2243918.
utility distribution infrastructure,’’ in Proc. IEEE Transp. Electrific. Conf. [35] L. Kang, X. Zhao, and J. Ma, ‘‘A new neural network model for the state-
Expo. (ITEC), Jun. 2013, pp. 1–6, doi: 10.1109/ITEC.2013.6573495. of-charge estimation in the battery degradation process,’’ Appl. Energy,
[18] K. Xiang, B. Chen, H. Lin, Y. Shen, Y. Du, and T. Yan, ‘‘Automatic vol. 121, pp. 20–27, May 2014, doi: 10.1016/J.APENERGY.2014.01.066.
demand response strategy of local pure electric vehicle with battery energy [36] P. A. Vikhar, ‘‘Evolutionary algorithms: A critical review and its future
storage system based on blockchain technology,’’ in Proc. 2nd IEEE prospects,’’ in Proc. Int. Conf. Global Trends Signal Process., Inf. Comput.
Conf. Energy Internet Energy Syst. Integr. (EI), Oct. 2018, pp. 1–6, doi: Commun. (ICGTSPICC), Dec. 2016, pp. 261–265, doi: 10.1109/ICGT-
10.1109/EI2.2018.8582289. SPICC.2016.7955308.
[19] C. D. Korkas, M. Terzopoulos, C. Tsaknakis, and E. B. Kosmatopou- [37] S. Wenninger, C. Kaymakci, C. Wiethe, J. Römmelt, L. Baur,
los, ‘‘Nearly optimal demand side management for energy, thermal, EV B. Häckel, and A. Sauer. (Jan. 2022). How Sustainable is Machine
and storage loads: An approximate dynamic programming approach for Learning in Energy Applications?—The Sustainable Machine
smarter buildings,’’ Energy Buildings, vol. 255, Jan. 2022, Art. no. 111676, Learning Balance Sheet. Accessed: Sep. 26, 2022. [Online]. Available:
doi: 10.1016/J.ENBUILD.2021.111676. https://aisel.aisnet.org/wi2022/sustainable_it/sustainable_it/1
[20] Z. Dong, J. Jiang, H. Qian, and H. Sun, ‘‘Demand side management [38] A. Eisenmann, T. Streubel, and K. Rudion, ‘‘Power quality mitigation via
considering household appliances and EV,’’ in Proc. 6th Int. Conf. Smart smart demand-side management based on a genetic algorithm,’’ Energies,
Grid Smart Cities (ICSGSC), Chengdu, China, Oct. 2022, pp. 22–24. vol. 15, no. 4, p. 1492, Feb. 2022, doi: 10.3390/EN15041492.
[Online]. Available: https://ieeexplore.ieee.org/xpl/conhome/1821804/all- [39] Y. LeCun, Y. Bengio, and G. Hinton, ‘‘Deep learning,’’ Nature, vol. 521,
proceedings no. 7553, pp. 436–444, May 2015, doi: 10.1038/nature14539.
[21] C. Ran, Y. Zhang, and Y. Yin, ‘‘Demand response to improve
[40] D. E. Rumelhart, G. E. Hinton, and R. J. Williams, ‘‘Learning rep-
the shared electric vehicle planning: Managerial insights, sustainable
resentations by back-propagating errors,’’ Nature, vol. 323, no. 6088,
benefits,’’ Appl. Energy, vol. 292, Jun. 2021, Art. no. 116823, doi:
pp. 533–536, Oct. 1986, doi: 10.1038/323533a0.
10.1016/J.APENERGY.2021.116823.
[41] Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, ‘‘Gradient-based learn-
[22] S. M. Shariff, M. S. Alam, F. Ahmad, Y. Rafat, M. S. J. Asghar, and
ing applied to document recognition,’’ Proc. IEEE, vol. 86, no. 11,
S. Khan, ‘‘System design and realization of a solar-powered electric vehicle
pp. 2278–2323, Nov. 1998, doi: 10.1109/5.726791.
charging station,’’ IEEE Syst. J., vol. 14, no. 2, pp. 2748–2758, Jun. 2020,
[42] Z. C. Lipton, J. Berkowitz, and C. Elkan, ‘‘A critical review of recurrent
doi: 10.1109/JSYST.2019.2931880.
[23] J. Antoun, M. E. Kabir, R. F. Atallah, and C. Assi, ‘‘A data driven perfor- neural networks for sequence learning,’’ 2015, arXiv:1506.00019.
mance analysis approach for enhancing the QoS of public charging sta- [43] J. Masci, U. Meier, D. Cireşan, and J. Schmidhuber, ‘‘Stacked convolu-
tions,’’ IEEE Trans. Intell. Transp. Syst., vol. 23, no. 8, pp. 11116–11125, tional auto-encoders for hierarchical feature extraction,’’ Lecture Notes in
Aug. 2022, doi: 10.1109/TITS.2021.3100875. Computer Science (Including Subseries Lecture Notes in Artificial Intel-
[24] X. Chen, H. Wang, F. Wu, Y. Wu, M. C. Gonzalez, and J. Zhang, ligence and Lecture Notes in Bioinformatics), vol. 6791, no. 1. Berlin,
‘‘Multimicrogrid load balancing through EV charging networks,’’ IEEE Germany: Springer, 2011, pp. 52–59, doi: 10.1007/978-3-642-21735-7_7.
Internet Things J., vol. 9, no. 7, pp. 5019–5026, Apr. 2022, doi: [44] G. E. Hinton and R. R. Salakhutdinov, ‘‘Reducing the dimensionality of
10.1109/JIOT.2021.3108698. data with neural networks,’’ Science, vol. 313, no. 5786, pp. 504–507,
[25] S. Yang, M. Wu, X. Yao, and J. Jiang, ‘‘Load modeling and iden- 2006, doi: 10.1126/SCIENCE.1127647.
tification based on ant colony algorithms for EV charging stations,’’ [45] D. Liu, Y. Sun, Y. Qu, B. Li, and Y. Xu, ‘‘Analysis and accurate
IEEE Trans. Power Syst., vol. 30, no. 4, pp. 1997–2003, Jul. 2015, doi: prediction of user’s response behavior in incentive-based demand
10.1109/TPWRS.2014.2352263. response,’’ IEEE Access, vol. 7, pp. 3170–3180, 2019, doi:
[26] Z. Moghaddam, I. Ahmad, D. Habibi, and M. A. S. Masoum, ‘‘A coordi- 10.1109/ACCESS.2018.2889500.
nated dynamic pricing model for electric vehicle charging stations,’’ IEEE [46] C. Giovanelli, S. Sierla, R. Ichise, and V. Vyatkin, ‘‘Exploiting artificial
Trans. Transport. Electrific., vol. 5, no. 1, pp. 226–238, Mar. 2019, doi: neural networks for the prediction of ancillary energy market prices,’’
10.1109/TTE.2019.2897087. Energies, vol. 11, no. 7, p. 1906, Jul. 2018, doi: 10.3390/EN11071906.
[27] I. Antonopoulos, V. Robu, B. Couraud, D. Kirli, S. Norbu, A. Kiprakis, [47] G. Mohi Ud Din, A. U. Mauthe, and A. K. Marnerides, ‘‘Appliance-
D. Flynn, S. Elizondo-Gonzalez, and S. Wattam, ‘‘Artificial intelligence level short-term load forecasting using deep neural networks,’’ in Proc.
and machine learning approaches to energy demand-side response: A Int. Conf. Comput., Netw. Commun. (ICNC), Mar. 2018, pp. 53–57, doi:
systematic review,’’ Renew. Sustain. Energy Rev., vol. 130, Sep. 2020, 10.1109/ICCNC.2018.8390366.
Art. no. 109899, doi: 10.1016/j.rser.2020.109899. [48] H. W. Lin, M. Tegmark, and D. Rolnick, ‘‘Why does deep and cheap
[28] J. Sachs and O. Sawodny, ‘‘A two-stage model predictive control strategy learning work so well?’’ J. Stat. Phys., vol. 168, pp. 1223–1247, Sep. 2017,
for economic diesel-PV-battery island microgrid operation in rural areas,’’ doi: 10.1007/S10955-017-1836-5.
IEEE Trans. Sustain. Energy, vol. 7, no. 3, pp. 903–913, Jul. 2016, doi: [49] S. Sun, W. Chen, L. Wang, X. Liu, and T.-Y. Liu, ‘‘On the depth of deep
10.1109/TSTE.2015.2509031. neural networks: A theoretical view,’’ 2015, arXiv:1506.05232.

VOLUME 11, 2023 8759

 A. Hafeez et al.: Utilization of EV Charging Station in Demand Side Management Using Deep Learning Method

[50] F. Javed, N. Arshad, F. Wallin, I. Vassileva, and E. Dahlquist, ‘‘Fore- RASHID ALAMMARI (Senior Member, IEEE)
casting for demand response in smart grids: An analysis on use received the B.S. degree from Qatar University,
of anthropologic and structural data and short term multiple loads in 1985, the M.Sc. degree from Washington State
forecasting,’’ Appl. Energy, vol. 96, pp. 150–160, Aug. 2012, doi: University, Pullman, WA, USA, in 1989, and the
10.1016/J.APENERGY.2012.02.027. Ph.D. degree from the University of Strathclyde,
[51] A. R. Khan, A. Mahmood, A. Safdar, Z. A. Khan, and N. A. Khan, Glasgow, U.K., in 1996, all in electrical engineer-
‘‘Load forecasting, dynamic pricing and DSM in smart grid: A review,’’ ing, majoring in power systems. He started as a
Renew. Sustain. Energy Rev., vol. 54, pp. 1311–1322, Feb. 2016, doi:
Teaching Assistant with an industry experience
10.1016/J.RSER.2015.10.117.
partnership with the Ministry of Electricity and
[52] T.-Y. Ma and S. Faye, ‘‘Multistep electric vehicle charging station occu-
pancy prediction using hybrid LSTM neural networks,’’ Energy, vol. 244, Water. He became an Assistant Professor, in 1996,
Apr. 2022, Art. no. 123217, doi: 10.1016/j.energy.2022.123217. and was promoted to an Associate Professor, in 2003. He was appointed as
[53] J. Zhu, Z. Yang, Y. Chang, Y. Guo, K. Zhu, and J. Zhang, ‘‘A novel the Head of the QU Foundation Program, from 1998 to 2000, then as the
LSTM based deep learning approach for multi-time scale electric vehi- Chairperson of the Department of Electrical Engineering, from 2000 to 2004,
cles charging load prediction,’’ in Proc. IEEE PES Innov. Smart Grid leading the department to its first ABET accreditation. He was appointed
Technol. Asia (ISGT), May 2019, pp. 3531–3536, doi: 10.1109/ISGT- as the Dean of the College of Engineering, Qatar University, from October
ASIA.2019.8881655. 2012 to March 2016. He is a published author of many academic studies on
[54] H. H. H. Aly, ‘‘A proposed intelligent short-term load forecasting hybrid power systems and power quality. He received the University Distinguished
models of ANN, WNN and KF based on clustering techniques for smart Faculty Research Award, in 2004. In 2012, he achieved the State of Qatar
grid,’’ Electric Power Syst. Res., vol. 182, May 2020, Art. no. 106191, doi: Incentive Award in electrical engineering. He received the scholarship from
10.1016/J.EPSR.2019.106191. Qatar University.
[55] G. Zhang, X. Bai, and Y. Wang, ‘‘Short-time multi-energy load fore-
casting method based on CNN-Seq2Seq model with attention mech-
anism,’’ Mach. Learn. Appl., vol. 5, Sep. 2021, Art. no. 100064, doi:
10.1016/J.MLWA.2021.100064.
[56] M. Dabbaghjamanesh, A. Moeini, and A. Kavousi-Fard, ‘‘Reinforce-
ment learning-based load forecasting of electric vehicle charging station
using Q-Learning technique,’’ IEEE Trans. Ind. Informat., vol. 17, no. 6,
pp. 4229–4237, Jun. 2021, doi: 10.1109/TII.2020.2990397. ATIF IQBAL (Senior Member, IEEE) received the
[57] C. W. Gellings, ‘‘Power/energy: Demand-side load management: The B.Sc. (Hons.) and M.Sc. degrees in engineering
rising cost of peak-demand power means that utilities must encourage (power system and drives) from Aligarh Muslim
customers to manage power usage,’’ IEEE Spectr., vol. S-18, no. 12, University (AMU), Aligarh, India, in 1991 and
pp. 49–52, Dec. 1981, doi: 10.1109/MSPEC.1981.6369703. 1996, respectively, the Ph.D. degree from Liv-
[58] C. W. Gellings, ‘‘Evolving practice of demand-side management,’’ J. Mod- erpool John Moores University, Liverpool, U.K.,
ern Power Syst. Clean Energy, vol. 5, no. 1, pp. 1–9, Jan. 2017, doi:
in 2006, and the D.Sc. degree (Habilitation) in con-
10.1007/S40565-016-0252-1.
trol, informatics, and electrical engineering from
[59] D. Groppi, A. Pfeifer, D. A. Garcia, G. Krajačić, and N. Duić, ‘‘A review
the Gdansk University of Technology, in 2019.
on energy storage and demand side management solutions in smart
energy islands,’’ Renew. Sustain. Energy Rev., vol. 135, Jan. 2021, He has been a Lecturer with the Department of
Art. no. 110183, doi: 10.1016/J.RSER.2020.110183. Electrical Engineering, AMU, since 1991, where he worked as a Full Profes-
[60] Y. Rui and A. A. El-Keib, ‘‘A review of ANN-based short-term load sor, until August 2016. He is currently a Full Professor with the Department
forecasting models,’’ in Proc. 27th Southeastern Symp. Syst. Theory, 1995, of Electrical Engineering, Qatar University, Doha, Qatar, and a former Full
pp. 78–82, doi: 10.1109/SSST.1995.390613. Professor with the Department of Electrical Engineering, AMU. He is the
[61] J. A. Jardini, C. M. V. Tahan, M. R. Gouvea, S. U. Ahn, and Head of the Design Team of power electronics and drives equipment at
F. M. Figueiredo, ‘‘Daily load profiles for residential, commercial and Power Lab Instruments, Chennai, India. He has been listed in top 2% highly
industrial low voltage consumers,’’ IEEE Trans. Power Delivery, vol. 15, cited scientists of the world (data released by Stanford University, USA).
no. 1, pp. 375–380, Jan. 2000, doi: 10.1109/61.847276. The world ranking in 2019 was #649 and the current ranking is #622.
He has supervised several large research and development projects worth
more than multi million USD. He has published widely in international
journals and conferences on his research findings related to power electron-
ics, variable speed drives, and renewable energy sources. He has authored
or coauthored more than 520 research articles and four books and several
chapters in edited books. His research interests include smart grid, complex
energy transition, active distribution networks, electric vehicles drivetrain,
ABDUL HAFEEZ received the B.Tech. degree sustainable development and energy security, distributed energy generation,
in electrical engineering and the M.Tech. degree and multiphase motor drive systems. He is a fellow of IET and IE. He was
in engineering (instrumentation and control) from a recipient of the Maulana Tufail Ahmad Gold Medal for standing first
Aligarh Muslim University (AMU), Aligarh, at the B.Sc.Engg. (Electrical) Exams from AMU, in 1991. He was also a
India, in 2011 and 2013, respectively. He was recipient of the Outstanding Faculty Merit Award from 2014 to 2015, and the
an Assistant Professor at AKTU University, Uttar Research Excellence Awards at Qatar University, in 2015 and 2019. He has
Pradesh, India, from 2013 to 2019. He joined a received several Best Research Paper Awards, such as IEEE ICIT-2013, IET-
research project as a Research Assistant at REC SEISCON-2013, SIGMA 2018, IEEE CENCON 2019, IEEE ICIOT 2020,
Ambedkar Nagar, Uttar Pradesh, in 2019. He is ICSTEESD-20, Springer ICRP 2020, and IEEE GUCON 2021. He is the
currently working as a Research Assistant with the Vice-Chair of the IEEE Qatar Section. He is also an Associate Editor of
Department of Electrical Engineering, Qatar University, Doha, Qatar. He has the IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS and IEEE ACCESS and
authored or coauthored more than seven publications in international journals the Editor-in-Chief of the Journal on Electrical Engineering (I’manager).
and conference proceedings. His research interests include machine learning, He was a Former Associate Editor of the IEEE TRANSACTIONS ON INDUSTRY
smart grid, electric vehicles, data-driven control, optimal control, sustainable APPLICATION and a Former Guest Associate Editor of the IEEE TRANSACTIONS
development, and energy security. He is also an active review member of ON POWER ELECTRONICS.
various reputed international journals and conferences, including IEEE.

8760 VOLUME 11, 2023