ISA Transactions 97 (2020) 431–447

Contents lists available at ScienceDirect

ISA Transactions
journal homepage: www.elsevier.com/locate/isatrans

Practice article

Internet of Things based real-time electric vehicle load forecasting and

charging station recommendation
George F. Savari a , Vijayakumar Krishnasamy a , Jagabar Sathik a , Ziad M. Ali b,c ,
∗
Shady H.E. Abdel Aleem d ,
a
SRM Institute of Science and Technology, Chennai 603-203, India
b
Electrical Engineering Department, College of Engineering at Wadi Addawaser, 11991, Prince Sattam bin Abdulaziz University, Saudi Arabia
c
Electrical Engineering Department, Aswan faculty of Engineering, 81542, Aswan University, Egypt
d
15th of May Higher Institute of Engineering, Mathematical and Physical Sciences, Helwan, Cairo, Egypt

highlights graphical abstract

• Inadequate charging infrastructure

is a major barrier to the wide ac-
ceptance of EVs.
• We propose a real-time server-based
forecasting application.
• We present an EV charging system
by utilizing the advantages of the
IoT technology.
• The application avoids third-party
intervention and protects EV user
privacy.
• The effectiveness of this application
is validated through a low-cost test
system.

article info a b s t r a c t

Article history: Electric vehicles (EVs) are emerging as a favorable strategy to meet the increasing environmental
Received 19 January 2019 concerns and energy insufficiency, and this trend is expected to grow in the near future. However,
Received in revised form 4 August 2019 the inadequate charging infrastructure is becoming a major barrier to the wide acceptance of EVs.
Accepted 4 August 2019
Deployment of this infrastructure is expected to maximize the adoption of EVs to facilitate users’
Available online 6 August 2019
range anxiety. Therefore, connectivity between the charging stations (CS) is mandatory. Understanding
Keywords: the real-time status of CSs can provide valuable information to users such as availability of charging
Charging stations provisions, reserves and the time to reach the CS. The intent of this paper is to provide a better
Forecasting application EV charging system by utilizing the advantages of the Internet of Things (IoT) technology. The IoT
Cloud storage paradigm offers the present facilities a real-time interactional view of the physical world by a variety
EV charging of sensors and broadcasting tools. This research article proposes a real-time server-based forecasting
Internet of Things application: i) to provide scheduling management to avoid waiting time; and ii) to provide a real-
Renewable energy resources
time CS recommendation for EVs with an economic cost and reduced charging time. In addition,
the proposed scheme avoids third-party intervention and protects EV user privacy and complex
information exchange between the user and CS. The end users can easily use the CS based on their
requirements. This synergetic application is built up through the PHP programming language in the
Linux UBUNTU 16.04 LTS operating system, and all relevant information is processed and managed
through Cloud Structured Query Language (CSQL) from a Google cloud platform. The effectiveness

∗ Corresponding author.
E-mail address: [email protected] (S.H.E. Abdel Aleem).

https://doi.org/10.1016/j.isatra.2019.08.011
0019-0578/© 2019 ISA. Published by Elsevier Ltd. All rights reserved.
432 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447

of this application is also validated through a low-cost test system using LTC 4150, ESP 8266 Wi-Fi
module and Arduino.
© 2019 ISA. Published by Elsevier Ltd. All rights reserved.

Abbreviations Os Optimization score

Pr Price rating
AES Advanced encryption standard Dr Distance rating
AI Availability index Sar Slot availability rating
API Application program interface Ap Actual price
BDR Battery drain rate Mp Maximum price
CS Charging station Ad Actual distance
CSP Charging service provider Md Maximum distance
CSQL Cloud structured query language Asa Actual slot available
DNS Domain name system Msa Maximum slot available
DSO Distribution system operator
EMC Electromagnetic compatibility
EVs Electric vehicles 1. Introduction
EVSE EV supply equipment
GNSS Global navigation satellite system In recent years, countries around the world have implemented
GPS Global positioning system regulatory requirements to solve issues in the transport sector,
HEVs Hybrid EVs such as the escalating requirement for non-renewable fuels and
HTTP Hypertext transfer protocol pollution in metropolitan areas. So, great attention has been
HTTPS Hypertext transfer protocol secure given to eco-friendly EVs due to concerns about the availability
ICE Internal combustion engine of fossil fuel, the environment and the regulation on emissions.
Many countries have stipulated significant guidelines to ease
ICT Information and communications technology
the expansion and adoption of EVs [1,2]. For example, several
IEX Indian Energy Exchange
cities of North America are in the changeover period to rapidly
IoT Internet of things replace conventional vehicles with hybrid plug-in vehicles and
IP Internet protocol then to EVs [3]. EVs have great benefits in terms of cost and
ISO International standards organization convenience. Unlike internal combustion engine (ICE) vehicles,
IX Internet exchange points the maintenance costs (costs of the cooling liquid, starter battery,
Li-ion Lithium ion (a type of battery) lubricant, and exhaust system) do not exist in an EV [4]. The
MAE Metropolitan area exchanges driving range of EVs is also growing faster than that in the past.
NAP Network access points The compact Mahindra Reva car with a rated capacity of 16 kWh
NSP Network service providers can provide a driving range of 80 miles. Tesla model X (SUV) cars
PHEV Plug-in hybrid EV with a capacity of 94 kWh provide a mileage of 400–450 km.
A typical EV charging system is represented in Fig. 1. Though
RET Retailer
the driving ranges of EVs are determined by the capacity of the
SAE Society of automobile engineers
battery, the powertrain’s complexity plays a major role. Recent
SOC State of charge research is concentrating on the enhancement of the driving
SSL Secure socket layer range of EVs [5], so battery monitoring and management systems
V2D Vehicle-to-driver are necessary. In [6], a broad study on the Li-ion batteries used
V2G Vehicle-to-grid for EVs is discussed. The basic operation concepts, assembly, and
V2I Vehicle-to-infrastructure performance of many types of Li batteries are discussed and
V2V Vehicle-to-vehicle suggested for high-capacity batteries, for which a quick charging
system is required. In [7], fast CSs are capable of charging the
Nomenclature battery from its state of charge (SOC) of 20% to about 80% within
Bc Battery capacity 30 min. Conversely, for level II and level I, as given in Table 1,
Er Estimated range the charging system takes a long time to charge. For those EV
DTE Distance to empty in kilometers users who do not own chargers, a public CS is a better charging
choice. Naoui et al. in [8] have analyzed the best charger mode
Ib Battery current in ampere
that guarantees efficiency and the minimum charging time for
Iloss The consumed current in ampere
EVs and hybrid electric vehicles (HEVs).
Crated Rated capacity of the battery SOC is one of the important key enablers to provide the
Ec (t) Energy cost at a particular hour present condition of the battery to charge and discharge safely,
SOC (ko ) Primary SOC of the battery which in turn enhances the battery life. It can be stated as the
Cc Charging cost ratio between the balance capacities of the battery and the rated
Rp Required power capacity of the battery. Thus, SOC assists in managing batteries [9,
10]. Various SOC estimation methods are presented in [11]. In the
automotive industry, smart communication systems are catego-
rized as (1) vehicle-to-infrastructure (V2I), (2) vehicle-to-driver
(V2D), (3) and vehicle-to-vehicle (V2V).
 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447 433

Fig. 1. Typical EV charging system. Fig. 2. Example of selecting wrong CS.

Table 1
Common charging power range [15].
sharing the information of the EV to three different operators,
Level Grid supply Voltage (V) Current (A) Nature of namely the charging service provider (CSP), distribution system
charging
operator (DSO), and retailer (RET). However, the communica-
Level I Single-phase (1Φ ) 230 16 Slow
tion infrastructure between EVs, the retailer, and the DSO is not
Three-phase (3Φ ) 400 16 Slow
Level II 1Φ 230 32 Slow described in detail. In [20], optimal online scheduling was imple-
3Φ 400 32 Slow mented by using a price-responsive charging scheme. Therefore,
Level III 1Φ 230 32 Slow the aggregator may investigate the plug-in pattern of the EV.
3Φ 690 250 Medium In [21], a problem consisting of an EV aggregator that manages
Level IV (DC) Nil 600 400 Fast
the charging controls of an EV fleet was addressed under the
mobility uncertainty. Bidirectional communication was used for
Table 2 monitoring and controlling data exchange. The decision making
Charging point infrastructure. for on-the-move EV charging management depends on the global
Organization Region Public charge aggregator. In [22], details of the EV fleet were collected with in-
point ratio
formation and communications technology (ICT) and shared with
European Council European Union 10 the clients and aggregators for optimal charging in the energy
National Development and China 8 (pilot cities),
Reform Commission 15 (other cities)
market. In [23], using modern communication technology, the
Electric Power Research United States 7–14 EV information is shared with the energy supplier to coordinate
Institute charging and discharging. From the above-mentioned works, it
National Renewable Energy United States 24 was noted that the aggregator controls the entire operation and
Laboratory
the EV user depends on the aggregator. If the aggregator fails to
California Energy California 27
Commission operate, the charging management system will not perform its
function. Moreover, the system may suffer more from security
aspects.
While EVs are charging, one of the major problems is not to
Because of the huge investments of governments, the com- reveal personal information, i.e., the vehicle’s registration num-
munication systems like V2V and V2I have been well studied. ber [24,25]. In earlier works [26,27], EVs usually send a request to
The main goal of using these communication systems is to de- the global aggregator to make the decision. Throughout this pro-
crease traffic congestion, to make road transport safer [12], and to cess, the information on the EVs will be released as anticipated.
avoid vehicle collisions [13,14]. Still, there are only few works to The privacy of the EV charging state is necessary, as a malevolent
confirm the effectiveness of vehicle-to-device systems for using trader may attack an individual EV with spontaneously manufac-
an EV, and more works are needed in this research area. The tured goods or services to the EV location. The EV user may not
data communication deliberately helps the EV user and CS with know if the nearest CS is available or not, which leads to selecting
efficient scheduling to manage the power. The estimated SOC the wrong CS, as illustrated in Fig. 2.
information is communicated to the nearest CS to allot the slot A mobile android application was developed by Bedogni et al.
based on the power demand of the grid. [28] with the facility to provide a list of positions of CSs in
In [16], the estimated SOC information is communicated to geographical manner nearest to the EV’s position. Even though
the aggregator based on the power demand of the grid. Ag- the application provides an estimation of driving range and a
gregators could utilize cellular communication to provide the list of available CSs, it is not smart enough to find the real-
charging/discharging facility to the EV users. Details of EV users time status (i.e. slot availability, price) of the CSs to reserve a
are communicated to a charge controller using an automated slot. In [29], Tannahill et al. have investigated the SOC estimation
communication infrastructure such as internet or wireless net- along with the CS recommendation. The recommended algorithm
works [17]. After gathering all information, the charge controller can alert the EV driver when the safe SOC limit is reached, and
has an additional degree of freedom to charge the EV. In [18], the it can provide an accurate estimation of the EV charging time.
aggregator has a precondition that the aggregator is to be notified However, the model presented in [29] is not able to make the
with the available SOC, in addition to the expected departure geographical location and route view of CSs available with mul-
time whenever a vehicle is plugged in or out. The information is tiple requirements. The drawbacks presented in the existing EV
needed for optimal charging control of EV. In addition, the driver charging methods, such as the location of EV, user phone number,
would be required to sign a contract to retain the EV connected vehicle details and required charging power, may cause different
to the grid for a certain time period even though charging is not privacy issues to the users. Hence, it is crucial to find a program
required. In [19], the flexible charging of EVs was performed by that allows customers to query the real-time status of the CSs
434 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447

Table 3
Comparison of the reviewed research works to the proposed application.
Ref. Year Objective Tool/Technique used Pricing method EVa Comments
1 2 3 4 5 6
16 2018 To minimize charging Matlab Time of use ✔ ✘ ✘ ✔ ✘ ✘ Installing smart chargers at user’s
cost by a coordinated homes, offices, and public places
charging method creates challenges.
Personal data from the EVs has to
be fast, secure, and reliable.
17 2011 To minimize charging Matlab Contract ✔ ✘ ✘ ✘ ✔ ✘ The predicted driving pattern and
cost and time by a its energy demand will not be
two-stage heuristic accurate.
iterative method The aggregator should control the
sequence, duration and the rate of
charging for each EV.
18 2010 To compute optimal Matlab Contract ✔ ✘ ✘ ✘ ✘ ✘ The aggregator should control the
charging of EVs by charging/discharging under the
controlling charging rate customer’s agreement.
and charging sequence
19 2012 To implement a flexible N/A Time of use ✔ ✘ ✔ ✔ ✘ ✘ Information of the EV is shared to
charging for the EV user three different operators. In
addition, a complex information
exchange is required.
20 2019 To minimize charging CPLEX 12.6 Time of use ✔ ✘ ✔ ✘ ✔ ✘ An aggregator uses the
cost by a upper-level controller for a
price-responsive early large-scale coordination of EV
charging control charging.
The method does not encourage
charging during high-priced time
periods, and this may lead to
forecasting errors.
21 2018 To manage EV charging N/A Time of use ✔ ✘ ✔ ✘ ✘ ✘ The decision making for
to minimize user trip on-the-move EV charging
duration using management depends on the
intermediate charging at global aggregator. If the
CSs aggregator fails to operate, the
charging management system will
not work. Moreover, the system
may suffer more from security
aspects.
22 2017 To optimize EV charging Matlab Time of use ✔ ✘ ✔ ✔ ✘ ✘ The aggregator will have to
costs through smart optimize power delivered to the
charging considering EV battery.
user preference Three different tariffs are provided
to EV user preference.
23 2012 To minimize charging GAMS/SCENRED, Time of use ✘ ✘ ✔ ✘ ✘ ✘ A predetermined contract should
cost by a risk-based Monte be signed between the aggregator
approach Carlo simulation and the user.
Regulation capacity and price are
deterministic. Monte Carlo
simulations were used to consider
uncertainties corresponding to the
EV driving patterns and energy
market prices.
26 2013 To schedule EV charging Optimal control Time of use ✔ ✘ ✘ ✘ ✘ ✘ User’s privacy such as personal
according to the control theory with convex and social information can be
signal broadcast by the function easily obtained in the digitalized
utility company energy system. Therefore, privacy
protection has been a great
challenge in EV charging.
27 2014 To make optimal V-Charge Time of use ✔ ✘ ✔ ✘ ✘ ✘ The privacy for an EV charging
combination of public scenario is important as
and individual businesses may bombard an
transportation for individual EV with unwanted
coordinated recharging products or services in relation to
strategies EV’s location.
28 2014 To assist EV drivers Semantic Time of use ✘ ✔ ✔ ✔ ✘ ✘ It is not smart enough to inform
through battery information the driver on how long the
monitoring, dynamic Broker/GPS charging time will be. Also, it
range prediction requires the driver to enter
information such as the current
position of the EV, the amount of
energy to be charged and the
desired distance to reserve a CS.
(continued on next page)
 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447 435

Table 3 (continued).
Ref. Year Objective Tool/Technique used Pricing method EVa Comments
1 2 3 4 5 6
29 2015 To reduce the range Microsoft excel N/A ✘ ✔ ✔ ✔ ✘ ✘ Estimated the EV range for a
anxiety for EV user specific route. Model based on
offline data. It must be able to
collect and update all the data in
real time.
36 2017 A real time charging Traction power N/A ✘ ✔ ✔ ✘ ✘ ✘ A real-time range/SOC estimation
recommendation for EV estimation model and an intelligent recharge
driver based on an alerting system are proposed to
accurate real-time range reduce the driver range anxiety.
indicator system It was only assumed that the
system can detect if the
recommended charging station is
available or not.
If it is available, reserve a CS. If it
is not available, then the vehicle
can select another nearby CS that
is available for use.
Proposed application Web (cloud) Time of use ✘ ✔ ✔ ✔ ✔ ✔ Information sharing between CSs
and users will be of benefit to
both the user and the CSs.
Customer oriented, user friendly.
a
1: Third-party intervention, 2: EV user privacy, 3: CS location information, 4: User preference, 5: Load forecasting recommendation, and 6: Hassle-free information
exchange network.

within a certain area. The international trends for charging point Accordingly, this work proposes a real-time server-based fore-
infrastructure are given in Table 2. casting application: (i) to provide scheduling management to
EV load forecasting is an important issue for the efficient oper- avoid waiting time; and (ii) to provide a real-time CS recom-
ation and management of utility companies. In [30], the charging mendation for EVs with an economic cost and reduced charging
profiles are assumed such that vehicle charging will start at 17:00, time. The advantages of the proposed method compared to other
18:00 or 19:00 h daily. However, actual EV charging loads will methods presented in the literature are given in Table 3. This
highly depend on the travel patterns, which vary tremendously synergetic application is built up through the PHP programming
from one driver to another and from one day to another. In [31], language in the Linux UBUNTU 16.04 LTS operating system, and
GPS data from 76 vehicles are used to predict the EV charging all relevant information is processed and managed through Cloud
profiles in different regions. However, the study assumed that Structured Query Language (CSQL) from a Google cloud platform.
all the EVs are sedan type having similar characteristics. In [32], The effectiveness of this application is also validated through a
it is assumed that the plug-in portion of the fleet consists of low-cost test system using LTC 4150, ESP 8266 Wi-Fi module and
20% of fully electric vehicles, 30% of full hybrid, and 50% of mild Arduino.
hybrid for EV charging loading. EV charging demand was simu- The main contributions of the proposed scheme are summa-
lated in [33] by considering the spatial and temporal distribution, rized as follows:
based on the parking duration. In [34], EV load forecasting was
investigated and performed by analyzing the historical traffic and • The developed application can alert the EV driver when
meteorological data, and the charging performance of EVs was charging is required; recognize the nearest CSs with the
classified by using the decision tree algorithm. However, this names, addresses, and geographical route coordinates; and
study assumes only Li-ion battery cars with a capacity of 27 kWh, reserve the slot to charge the EV to reach the destination
and also it was assumed that a car could be charged only once a without the necessity of creating a complex information en-
day. In [35], the authors have used many methods such as Monte vironment. The data required for the algorithm do not need
Carlo forecasting technique and support vector machines for es- any complex information exchange since all information can
timating the charging demand of EVs as a decision maker has be obtained through GPS and the mobile internet system
the need for accurate forecasts in the power and energy sectors automatically using publicly available websites with a little
since most of the decisions are necessarily based on the fore- input from the driver.
casts of future demands. Several methods including similar day • With a minimal requirement of data input from the driver,
approach, regression models, time series, neural networks, ex- the developed application will be smart enough to provide
pert systems, fuzzy logic, and statistical learning algorithms were adequate information about the CSs to the driver, while
used for short-term forecasting. The development, improvements reducing the risk of collision with other vehicles.
and investigation of the appropriate mathematical tools can help • With the available SOC, the available number of CSs, their
develop more accurate load forecasting techniques. In [36], the present locations, the cost of CS and privacy information
authors have used traction power estimation to present a real- protection, a flexible real-time based online charging
time charging recommendation for an EV driver based on an scheduling scheme is proposed in this work, in which each
accurate real-time range indicator system. However, in the static moving EV can make its own choice rather than depending
system presented, the driver cannot get any information about on the choice from the aggregator. In this view, our pro-
the CS, i.e. the CS is functioning or not, how many vehicles are posed scheme is advantageous, since the privacy of EVs will
waiting for charging and what are the available charging points. not be shared through any medium because the choice is
From the above works, it was noted that a real-time application is made only on the user end. Additionally, the scheme also
required to collect transportation information, EV battery status provides data related to distance and cost, depending upon
and the CS status for forecasting possible loads on CSs and to the battery characteristics, and the accessibility of slots at
avoid depending on aggregators. the nearest CSs. The status of the CS, shared through web
436 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447

pages, can allow management of the energy system to react

depending upon the vehicle’s SOC and the vendor’s point
of view. To view the required information for the user, a
hypertext pre-processor is used to design the webpage. So,
the users can decide the CS based on the cost per unit
and distance. The database is designed in the cloud SQL
format. The database will have the necessary information
such as the cost, SOC and average rate etc. This database is
connected with the webpage and conveys the information
to the user with the help of the internet.
• Regarding accurate forecasting of possible loads on CSs, a
real-time application that collects information of transporta-
tion from application program interface (API), EV battery
status and the CS status (cost/kwh, range, total slots, and
availability of fast charging option) is developed.

Also, the advantages of the proposed EV charging scheme can be

summarized as follows:

1. The proposed scheme does not require any third-party Fig. 3. Electricity cost for 24 h.
(aggregator) intervention to allocate charging provision at
CSs.
2. The CSs are chosen by the customer. Hence, it is a
customer-oriented user-friendly application.
3. The amount of energy consumed to charge the EV and the
cost for the energy will not be revealed to any third party
except the CS and customer.
4. The customer can find the real-time status of CSs which
will avoid the congestion at a particular CS.
5. Based on the proposed scheme, the EV load forecasting can
be done efficiently.

Finally, introducing an intelligent system between vehicles

and drivers can be of great interest to the technology users
and developers and modern generation stakeholders who would
expect this facility in their vehicles. It is predicted that such sys-
tems could help support the adoption of EVs into the automobile
industry, particularly in developing countries.
The remainder of this paper is organized as follows. Sec-
tion 2 presents the test system and the measurement of SOC and
EV load forecasting. In Section 3 the EV charging infrastructure
Fig. 4. Number of EVs in different charging stations [37].
with proposed IoT application is explained. Also, it gives the EV
charging station network data along with the integrated database
of the developed application. Section 4 presents the web and
android applications, in addition to discussion of the results. noted that the unit cost is expressed by rupees per kilowatt-hour
Finally, Section 5 is dedicated to conclusions drawn from this (Rs/kWh), where the rupee is the official currency of India.
work, summary of the work done and future works. The proposed communication network will work to reduce the
variations. It will update the user concerning the accessibility of
2. EV load forecasting and SOC measurement slots and the cost at each CS. Hence, a user may prefer a sta-
tion where there is availability nearby. Obviously, intimating the
Load forecasting is an important process that can increase accessibility of slots during a precise time will avoid congestion
the efficiency of the distribution network. The accuracy of load significantly throughout peak hours at the CSs. So, insufficient
forecasting has a huge impact on the operation of a utility cor- numbers of CSs will lead to complication in load forecasting as
poration. In [37,38], the outcome of a statistical study on the well as in energy distribution for the EV users.
EV charging behavior of 220 domestic users across the UK is From the discussion above, the energy distribution crisis may
provided. The study was carried out for one year with more than affect the EV user. To overcome this problem, a better battery
68,000 samples. The authors also generated the probability dis- management system is needed. Hence, the SOC of the battery
tribution functions for various charging methods. These sample is to be monitored. Assessment of the SOC is a primary issue
results will help to forecast the maximum power demand of the for battery use. Describing the available SOC of a battery is a
EVs in the particular hour of the day. The commercial electricity key parameter for control strategies [39]. Battery performance
cost for 24 h from the Indian Energy Exchange (IEX) is given is determined by the SOC, so exact estimation of the SOC can
in Fig. 3. In general, this cost will vary according to the peak enhance the lifetime and also save energy by allowing the system
demand. The EV charging data of each station is given in Fig. 4. A to create an intelligent control strategy [40]. However, the main
particular CS may have additional power demand during a precise problem with the battery is that it is powered by chemical energy,
time duration while other CSs do not have this. The difference in and chemical energy cannot be measured directly. This leads to
the unit cost for the CS varies as shown in Fig. 5. It should be difficulty in the estimation of SOC.
 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447 437

The current readings are mathematically integrated over the

usage time to measure the SOC values, which can be given as
follows:
∫ k
1
SOC = SOC (ko ) + (Ib − Iloss ) dt (1)
Crated ko

where SOC (ko ) is the primary SOC of the battery, Crated is the
battery rated capacity, Ib is the battery current, and Iloss is the
consumed current.
It finds the available SOC by calculating the charge moved
either into or out of the battery. If the battery cell specification
states that it has 60 Ah, we can estimate 216,000 (ampere-
seconds) for coulomb counting. Therefore, 0% SOC will mean the
battery has 0 coulomb counts and 100% SOC will mean a count of
216,000, and this is used to estimate the rate of charge. The SOC
of the Li-ion battery [41] is given in Fig. 6.
In this work, LTC 4150 is used, which actually does the
Fig. 5. Charging cost at various charging stations [38]. coulomb counting method because of its simplicity and low
complexity. Basically, the coulomb counting method is based on
current integration and quantifying the charge delivered by the
There are a number of methods to assess the SOC of a bat- battery through sensing its input and output current. It acts by
tery [10,11,42,43]. These techniques can be further classified integrating the active flowing current over time to derive the total
sum of energy entering or leaving the battery. This results in a ca-
into three methods: (i) electrochemical; (ii) adaptive; and (iii)
pacity that is measured in ampere-hours typically. Certainly, the
electrical. The electrochemical techniques are highly accurate method is reliable as long as the current measurement is accurate.
but require access to the battery’s chemical composition. So, It can be applied to all the batteries used for EV application.
it is difficult to implement in software. Adaptive techniques The device monitors the current through an external resistor
between the battery’s positive terminal and the load. A voltage-
require an equivalent circuit model and a solution procedure to-frequency converter transforms the current sense voltage into
for a battery [44]. The accuracy of the adaptive techniques is a series of output pulses at the interrupt pin. These pulses cor-
determined by the efficiency of the equivalent model. However, respond to a fixed quantity of charge flowing into or out of the
battery. The interface is connected with an Arduino. Outputs of
electrical techniques require only measurable parameters such
the Arduino codes are obtained in digital format. Monitoring of
as charge/discharge current, terminal voltage, and internal resis- the SOC is important, as it is used to calculate the remaining
tance. Due to its easy implementation and low complexity, the energy and the maximum possible distance that can be driven,
coulomb counting (electric) technique is one of the commonly as well as to get the maximum output (speed or distance).
used techniques to estimate SOC [45]. This method is also known 3. EV charging infrastructure with proposed IOT application
as the ampere-hour (Ah) current integration technique. It is a less
complex and easy to implement algorithm. The available capacity The GSM Association has predicted that by the year 2030,
each and every EV will be linked together, and this will earn
in a cell is calculated by knowing the current that is charging or
profits of over 19B euros. In recent years, several standards and
discharging the cells and accumulating this over time. If the initial technologies for V2V and V2I communications have emerged. Dif-
SOC is known, then it is possible to obtain the precise SOC [46]. ferent types of communications are already available to exchange

Fig. 6. SOC of a Li-ion battery [41].

438 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447

Fig. 7. Flowchart of the EV charging reservation using IoT application.

the information (i.e. cables and radio exchanges). So, in order charging of EVs, and it will also enhance the grid by avoiding peak
to achieve an efficient EV charging scheme, V2G communication demand at a particular time.
standards are important. The Society of Automobile Engineers
(SAE), the International Standards Organization (ISO), and the 3.1. Test system for SOC measurement
International Electrotechnical Commission (IEC) have proposed
The experiment is done with a 9 V, 300 mAh Li-ion battery
several standardizations which describe various communication
and a coulomb counter, Arduino ESP 8266 and a 9 V DC motor.
fundamentals [47]. Various standards, such as charging com- The current rating of the Li-ion battery is 300 mAh, which means
munication protocols, charging connectors, safety and charging it can deliver 300 mA for 60 min, and this is said to be the
topology, are given in Table 4. initial charge of the battery cell. The motor load is 70 mAh, and
The developed application (web, android) has three divisions, it can run for 4.28 h with the available charge. Now, the cost for
namely (i) SOC estimation, (ii) selection of CS, and (iii) database charging the battery is given by:
and data processing. The overall flowchart for the proposed
Cc = Rp × Ec (t ) (2)
scheme is represented in Fig. 7.
The LTC coulomb counter is connected with the Arduino, and where Ec(t) is the energy cost at a particular hour.
it will estimate the SOC of the battery. The data will be read by The algorithm for experimental validation of SOC estimation
the laptop or PC and will be stored in the database of the server is given in Table 5.
using the ESP8266 Wi-Fi module. The database and the webpage If the motor runs for one hour, then it will consume 70 mAh.
will operate at the same time through an internet protocol. Hence, the remaining value of SOC is 76.66%, the observed power
is 0.63 kWh, and the result is depicted in Appendix A.1.
Depending on the user’s request, the webpage will show the
locations of the CSs with the help of GPS. Based on the availability 3.2. CS selection algorithm
of slots, charging cost and distance, the user will have the option
to select a particular CS. Once the CS receives the request from the The CS selection algorithm will provide relevant information
EV, it will allocate the charging slot based on the standards for the to the EV end user. When the charging points in the CSs are en-
EV charging infrastructure. This system will ensure the optimal gaged with the EV, the points transmit information to the server.
 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447 439

Table 4
Standard charging power levels [48,49].
No Standards Scopes Contents
1 SAE-J1772 Connector EV conductive charge coupler
2 IEC-62196 Connector Physical elements for energy transfer and communication (i.e. charging plugs, sockets,
charging couplers and inlets)
3 ISO-15118 Communication (charging) V2G communication interface (communication protocol)
4 SAE-J2293 Communication (charging) Energy transfer system
5 SAE-J2836 Communication (charging) Communication between PHEV and the grid, EV supply equipment (EVSE)
6 IEC-61850 Communication (power systems) Communication networks and the substation systems
7 IEC-61439 Topology (charging) LV switchgear and control gear assembly
8 IEC 61851 Topology (charging) EV conductive charging system, on-board charger EMC requirements and general requirements

Table 5
Algorithm for Experimental Validation of SOC Estimation.
Step Description
Step 1 Initialize the coulomb counter and Arduino
Step 2 Assign the voltage as zero.
Step 3 Make a communication between the system and Arduino.
Step 4 Connect the load with the battery and run it.
Step 5 Coulomb counter will feed the results to Arduino.
Step 6 Convert and store the results from analog to digital format.
Step 7 Display the results
Step 8 When the battery reaches the safe SOC, notify to the user
through a message.
Step 9 If the battery is plugged in for charging, the coulomb counter
goes to the idle state until completion of charging, or if not
plugged in it will continuously send alert messages up to fully
discharging of battery and terminate the connection.
Step 10 After charging of battery, the coulomb counter changes the
state from idle to running state.
Fig. 8. Optimal point for CS selection.

Large networks known as network service providers (NSPs) sup-

port the information sharing between CSs and the server. The Estimated range (Er ) = Y (km/kWh)
NSP is connected with the metropolitan area exchanges (MAEs) Distance to empty
through the network access points (NAPs). NAPs were the original
(DTE) = (X × Y × SOC ) km (3)
internet interconnects points. Both NAPs and MAEs are referred
to as internet exchange points (IXs). After that, the information Step 4: The calculated DTE will be the maximum distance that
will be passed through the domain name system (DNS). The the EV can drive with the available SOC and any EV CS within the
DNS keeps track of computer’s names and their corresponding IP limit will be fetched.
addresses on the internet. Finally, to ensure the data transferred Step 5: Then, the availability index (AI) in the CS (0 for available;
remains private, secure socket layer (SSL) is used. 1 for waiting no. 1; 2 for waiting no. 2) will be fetched for each
Meanwhile, the information about the EV location is moni- station.
tored by GPS. The battery drain rate is an important parameter Step 6: Both the DTE and AI data are taken into account and
to measure the available SOC. The coulomb counting test method optimized as illustrated in the sample Fig. 8.
is used to calculate the battery drain rate and available SOC at The EV can travel up to a maximum distance of 27 km with the
the regular intervals. With the available SOC, it is possible to remaining SOC. With an average speed, the battery will be empty
calculate the remaining distance that can be covered by the EV. before 40 min, and the EV will not be able to move. Meanwhile, at
The results will be stored in the local host, and the same will be the particular CS, the availability of charging slots is also varying.
communicated to the developed application. The current location Step 7: The booking option and all the relevant information of the
of the EVs and the CSs are fetched to the application by using CS will be displayed. This procedure is repeated for each CS.
the API and maps. An API is a code that allows two different Step 8: The station with the lowest optimal point will be given as
software programs to communicate with each other. The web the best option, and the results will be sorted in ascending order.
API is accessed using a hypertext transfer protocol (HTTP). To Now, the EV user can select the CS as desired.
ensure the safety of the information, the website uses the secure (Note: Total time calculation is given in the appendix for two
hypertext transfer protocol (HTTPS). Hence, the application has all cases if a user selects a CS based on the shortest distance due).
the inputs from the CSs and from the EVs. An availability index It should be mentioned that if an EV user arrives at the booked
is given to the slot availability, and this will be compared with CS earlier, he will use reserve points allowed by the CS for EV
the remaining battery capacity to find the optimized CSs. The users with no booking or early-arrived. These reserve points
step-by-step procedure is given below. help support the CS by fast chargers in case of emergency and
Step 1: The input from the CSs and EVs are collected dynamically abnormalities. However, non-flexible booking drivers will pay
for an interval of one minute. more than the drivers use the application. Arrival time-based
Step 2: The SOC% is the primary input from which the battery priority (ATP) algorithm will be used in such cases. ATP allows the
drain rate (BDR) will be calculated. vehicles into service once they arrive and hence reduces the delay
Step 3: Based on the BDR, the estimated distance to empty (DTE) time and avoids the charger to be idle. ATP algorithm lineups
will be calculated, as given in Eq. (3). all vehicles to charging points until all the charging points are
Battery capacity (Bc ) = X (kWh) engaged based on arriving time of the vehicles. Once a vehicle is
440 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447

Fig. 9. Flowchart of ATP algorithm.

 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447 441

Fig. 10. Integrated database structure of the developed application.

allotted to a charging point, the algorithm will update the waiting

time for the other vehicles. The same procedure will be repeated
until all the vehicles are charged. Fig. 9 illustrates flowchart of
the EV charging schedule by the ATP algorithm.

3.3. Integrated database for the proposed application

The input data received from the local host and CS will be sent
to the database. The database is created and managed by cloud
SQL. The advantage of cloud SQL is that it offers high performance,
scalability, and convenience. The files stored in this platform can
be accessed from anywhere in the world through an internet
connection, and the cloud environment has 100% durability so old
data can be retrieved at any time.
File security is a must-have feature during transfer and stor-
age. The cloud storage uses SSL or 256-bit advanced encryption
standard (AES-256 bit), and this reduces the risk of hacking. The
desktop view of the database is given in Fig. 10. The database
contains the CS and EV data to be processed. The contents of the
EV and CS data received are presented in Appendix A.2.
In India, the EV market is at a nascent stage and consists
of less than 1% of the total vehicles. So, it is assumed that the
Fig. 11. CSs from Google Maps [50].
CS is available at petrol pumps. We considered an EV network
system [50] for the Chennai to Chengalpattu highway road as
shown in Fig. 11. The EV network consists of 14 CSs. Some of
4. Web and android application
these are operated with fast charging DC and level II medium
charging. EVs including the Mitsubishi MiEV (16 kWh), Honda
Fit (20 kWh), and Nissan Leaf (30 kWh) are considered to have
a fixed minimum SOC of 25%. The number of sockets at a CS is The developed application will receive the updated available
distributed uniformly in the 5 to 15 range. It is assumed that each slots from the CSs and the location of EVs with an interval of one
socket of the DC fast charging option supplies 50 kW power and minute. Once the SOC of the battery reaches the specified limit
for the AC level II option supplies power of 19.2 kW [51]. (safe SOC limit), the EV user will take the decision to charge.
442 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447

Fig. 12. SOC status of an EV with respect to the vehicle range.

Fig. 13. Homepage of the developed application — desktop view.

The SOC status of an EV is given in Fig. 12. The proposed likelihood of selecting the wrong CS, unnecessary waiting time,
application has the homepage shown in Fig. 13, while the Android and congestion at the CS will be avoided significantly.
app view of the application developed is given in Fig. 14. Considering an EV such as a Mitsubishi MiEV (16 kWh) travel-
The user has options for the total battery capacity, location, ing to Pondicherry from Chennai, it will have reached the safe SOC
and available SOC. The user has to fill the battery capacity of (25%) when crossing Potheri. Therefore, the user has to recharge
the vehicle. The location and SOC of the EV can be identified ei- the vehicle’s battery. The user can travel up to a maximum dis-
ther automatically with Arduino and GPS or else manually. Since tance of 22.5 km at an average speed. Fig. 16 shows the DTE range
the GPS system employs the Global Navigation Satellite System
of the EV, and Fig. 17 shows the estimated distance calculation for
(GNSS), it can integrate multiple satellites that use microwave
the journey using API and maps by the application.
signals to provide data on speed, location time and direction.
Fig. 18 shows the status of the nearer CSs. It shows the rel-
Thus, real-time and historical navigation data on any kind of
evant information of the CSs such as the name of the service
journey can potentially be given by a GPS tracking system. Once
the required data are entered successfully, the application will provider, distance from the present location, reachability, total
process the input data. When the processing of all the inputs number of charging slots, engaged and available slots for charg-
given by the user and cloud SQL system is completed, the pro- ing, electricity price per kWh, availability of fast charging option
posed IoT application will give the results as given in Fig. 15. The and an optimization score. Based on the remaining battery capac-
optimized results can be viewed by clicking the respective option ity, the application will indicate whether the CSs are within the
available in the application. Now the user can decide at which CS range or not. The optimization score (Os ) can be calculated by var-
the EV can be charged without any difficulty. By doing so, the ious factors such as price (Pr ), distance (Dr ) and slot availability
 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447 443

Fig. 16. DTE Range — Desktop view.

Fig. 14. Android app view of the application.

Fig. 17. Validation of input data by the user and calculated data by the
developed application.

Asa
Sar = (7)
Msa
where the actual price (Ap ) is the charging price at a particular
CS, the maximum price (Mp ) is the highest charging price among
Fig. 15. Android app view of the application with results.
the CSs, the actual distance (Ad ) is the distance between the EV
and a particular CS, the maximum distance (Md ) is the longest
distance among CSs from the EV’s actual location, actual slot
availability (Asa ) is number of free charging slots at a particular CS,
(Sar ). Os is calculated as follows: an maximum slot availability (Msa ) is the highest number of free
charging slots among all CSs. The user can reserve a slot at any CS.
Os = α (Pr ) + β (Dr ) + γ (Sar ) (4) In addition, α , β , and γ are coefficients to balance Eq. (4) in which
the application enables the user to change the optimization score
so that,
weighting factors value as desired.
Ap
Pr = 1 − (5)
Mp 5. Conclusion

Ad Power systems engineers are always at the forefront to exploit

Dr = 1 − (6) new emerging technologies with the purpose of improving the
Md
444 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447

Fig. 18. Real-time status of CSs — desktop view.

power systems. In this regard, the paradigm of IoT is increas- Declaration of competing interest
ingly integrated with real-world applications, particularly in the
energy sector. Accordingly, in this paper, a real-time solution is The authors declare that they have no known competing finan-
proposed to improve the charging scheduling of EVs based on cial interests or personal relationships that could have appeared
modern communication between vehicles and CSs through state- to influence the work reported in this paper.
of-the-art technologies. The proposed scheme provides more se-
curity in terms of not sharing end-user information to aggregator,
Acknowledgments
unwanted rush in charging stations can be avoided, it is user
friendly, the user will receive the data of all the CSs nearby,
The authors appreciate and thank the editor and the anony-
and the end user can decide the cost or time based on their re-
mous reviewers for their constructive comments and suggestions.
quirements. The application is already developed (domain name:
www.optimalev.app, IP address: 35.200.233.16) by considering
various factors like the user convenience, variation of electricity Appendix
prices, and the willingness of the user to pay This robust applica-
tion is beneficial not only to EV users but also to the CSs to predict
the future load, to avoid congestion, and for energy consumption
A.1. Coulomb counting method
reduction and/or load balancing.
In future work, to ensure the reliability of the grid, V2V and
Fig. A.1 shows the Coulomb counting method — test results.
V2G can be incorporated with the developed application. One
thing that was slowing down the implementation of smarter grids
is the mindset that we have to have some kind of communi- A.2. Data structure in the database
cations from each device back to the grid operator. However,
effectively gathering information from all EVs and CSs is time- Fig. A.2 shows the vehicle data structure in the database, while
consuming to attain an optimized solution. Thus, the appliances Fig. A.3 shows the CS data structure.
can be made smart by connecting to a web page somewhere
Total time calculation if a user selects a charging station based
for that customer, and then a device is programmed to respond
on shortest time
appropriately to the grid operator to best suit the load and grid Case 1: If Ns > 0
operator’s needs. In opposition, the complexity and computation
burden of this centralized solution will increase exponentially Total time = T (EV, CS) + Ts
with the increasing number of EVs. Since the mobile network
Case 2: If Ns < 0
is cheap, it may cause over-congestion which will degrade the
quality of the communication. Total time = T (EV, CS) + T w s + Ts
G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447 445

Fig. A.1. Coulomb counting method — test results.

Fig. A.2. Vehicle data structure in the database.

446 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447

Fig. A.3. CS data structure in the database.

where: [16] Clairand J, Rodríguez-García J, Álvarez Bel C. Smart charging for elec-
tric vehicle aggregators considering users’ preferences. IEEE Access
2018;6:54624–35.
Ns Number of slots [17] Rotering N, Ilic M. Optimal charge control of plug-in hybrid elec-
T (EV, CS) Time taken by EV to travel to CS tric vehicles in deregulated electricity markets. IEEE Trans Power Syst
2011;26(3):1021–9.
Tws Waiting time at CS
[18] Han S, Han S, Sezaki K. Development of an optimal vehicle-to-
Ts Charging time grid aggregator for frequency regulation. IEEE Trans Smart Grid
2010;1(1):65–72.
References [19] Sundstrom O, Binding C. Flexible charging optimization for electric ve-
hicles considering distribution grid constraints. IEEE Trans Smart Grid
2012;3(1):26–37.
[1] Thomas CES. Transportation options in a carbon-constrained world: Hy-
[20] Yang Shaobing. Shaobing yang price-responsive early charging control
brids, plug-in hybrids, biofuels, fuel cell electric vehicles, and battery
based on data mining for electric vehicle online scheduling. Electr Power
electric vehicles. Int J Hydrog Energy 2009;34(23):9279–96.
Syst Res 2019;167:113–21.
[2] Dickerman L, Harrison J. A new car, a new grid. IEEE Power Energy Mag
[21] Cao Y, Wang T, Kaiwartya O, Min G, Ahmad N, Abdullah AH. An EV charg-
2010;8(2):55–61.
ing management system concerning drivers’ trip duration and mobility
[3] More Than 35 Technologies. [Online]. Available: http://www.apta.com/ uncertainty. IEEE Trans Syst Man Cybern Syst 2018;48(4):596–607.
mediacenter/pressreleases/2013/Pages/130422_Earth-Day.aspx.
[22] A. Perez-Diaz E. Gerding F. McGroarty coordination and payment mecha-
[4] Vrazic M, Vuljaj D, Pavasovic A, Paukovic H. Study of a vehicle conversion nisms for electric vehicle aggregators. Appl Energy 2018;212(Feb.):185–95.
from internal combustion engine to electric drive. In: IEEE international [23] Pantos M. Exploitation of electric-drive vehicles in electricity markets. IEEE
energy Conference. 2014, p. 1544–8. Trans Power Syst 2012;27(2):682–94.
[5] Dost PK-H, Spichartz P, Sourkounis C. Charging behavior of users utilizing [24] Lisovich MA, Mulligan D, Wicker SB. Inferring personal information from
battery electric vehicles and extended range electric vehicles within the demand-response systems. IEEE Secur Priv 2010;8(1):11–20.
scope of a field test. IEEE Trans Ind Appl 2018;54(1). [25] Döbelt S, Jung M, Busch M, M. Tscheligi consumers’ privacy concerns and
[6] Hannan MA, MdM Hoque, Hussain A, Yusof Y, Ker PJ. State-of-the-art and implications for a privacy preserving smart grid architecture—Results of
energy management system of lithium-ion batteries in electric vehicle an austrian study. Energy Res Soc Sci 2015;9:137–45.
applications: Issues and recommendations. IEEE Access 2018;6:19362–78. [26] Gan L, Topcu U, Low S. Optimal decentralized protocol for electric vehicle
[7] Navid R, Kashanizadeh B, Vakilian M, Barband SA. Optimal placement of charging. IEEE Trans Power Syst 2013;28(2):940–51.
fast charging station in a typical microgrid in Iran. In: Proc. 2013 10th [27] Timpner J, Wolf L. Design and evaluation of charging station schedul-
IEEE int. conf. eur. energy market (EEM). p. 1–7. ing strategies for electric vehicles. IEEE Trans Intell Transp Syst
[8] Mohamed N, Aymen F, Mouna BH, Alassaad S. Review on autonomous 2014;15(2):579–88.
charger for EV and HEV. In: 2017 International Conference on Green Energy [28] Bedogni L, et al. A mobile application to assist electric vehicles’ drivers
Conversion Systems, Mar. 2017, p. 1–6. with charging services. In: Proc. Next Gener. Mobile Apps, Serv. Technol.
[9] Ramadan HS, Becherif M, Claude F. Energy management improvement of Conf. 2014. p. 78–83.
hybrid electric vehicles via combined GPS/rule-based methodology. IEEE [29] Tannahill VR, Muttaqi KM, Sutanto D. Driver alerting system using range
Trans Autom Sci Eng 2017;14(2):586–97. estimation of electric vehicles in real time under dynamically varying
[10] Gholizadeh M, Salmasi FR. Estimation of state of charge unknown environmental conditions. IET Elect Syst Transp 2015;6:107–16.
nonlinearities and state of health of a lithium-ion battery based [30] Acha S, Green TC, Shah N. Optimal charging strategies of electric vehicles
on a comprehensive unobservable model. IEEE Trans Ind Electron in the UK power market. In: Proc IEEE ISGT, 2011. http://dx.doi.org/10.
2014;61(3):1335–44. 1109/ISGT.2011.5759128.
[11] Tang X, Liu B, Gao F. State of charge estimation of LiFePO4 battery [31] Ashtari A, Bibeau E, Shahidinejad S, Molinski T. PEV charging profile
based on a gain-classifier observer. Energy Proc 2017;105:2071–6. http: prediction and analysis based on vehicle usage data. IEEE Trans Smart
//dx.doi.org/10.1016/j.egypro.2017.03.585. Grid 2012;3(1):341–50.
[12] Di Lecce V, Amato A. Route planning and user interface for an advanced [32] Shahidinejad S, Filizadeh S, Bibeau E. Profile of charging load on the grid
intelligent transport system. IET Intel Transp Syst 2011;5(3):149–58. due to plug-in vehicles. IEEE Trans Smart Grid 2012;3(1):135–41.
[13] Taleb T, Benslimane A, Letaief KB. Toward an effective risk conscious and [33] Hongcai Z, et al. A prediction method for electric vehicle charging load
collaborative vehicular collision avoidance system. IEEE Trans Veh Technol considering spatial and temporal distribution. Automat Electr Power Syst
2010;59(3):1474–86. 2012;38(1):13–20.
[14] Barrios C, Motai Y, Huston D. Trajectory estimations using smartphones. [34] Arias MB, Bae S. Electric vehicle charging demand forecasting model based
IEEE Trans Ind Electron 2015;62(12):7901–10. on big data technologies. Appl Energy 2016;183(Dec.):327–39.
[15] The International Electrotechnical Commission (IEC) Electric Vehicle Con- [35] Majidpour M, Qiu C, Chu P, Pota HR, R. Gadh forecasting the EV charging
ductive Charging System—Part 1: General Requirement. In: IEC 61851-1 load based on customer profile or station measurement? Appl Energy
the international electrotechnical commission. 2001. 2016;163:134–41.
 G.F. Savari, V. Krishnasamy, J. Sathik et al. / ISA Transactions 97 (2020) 431–447 447

[36] Sarrafan K, Muttaqi KM, Sutanto D, Town GE. An intelligent driver alerting [44] Wei J, Dong G, Chen Z. Lyapunov-based state of charge diagno-
system for real-time range indicator embedded in electric vehicles. IEEE sis and health prognosis for lithium-ion batteries. J Power Sources
Trans. Ind. Appl. 2017;53(3):1751–60. 2018;397:352–60.
[37] Xydas ES, Marmaras CE, Cipcigan LM, Hassan AS, Jenkins N. Forecasting [45] Lu L, et al. A review on the key issues for lithium-ion battery management
electric vehicle charging demand using support vector machines. In: in electric vehicles. J Power Sources 2013;226:272–88.
Proceedings of the 2013 48th international Universities’ Power Engineering [46] Tang X, Wang Y, Chen Z. A method for state-of-charge estimation of
Conference, 2013. lifepo4 batteries based on a dual-circuit state observer. J Power Sources
[38] Quiros-Tortos J, Ochoa L, Lees B. A statistical analysis of EV charging 2015;296:23–9.
behavior in the UK. In: Innovative Smart Grid Technologies Latin America [47] Pereirinha PG, Trovo JP. Standardization in electric vehicles. In: XIICLEEE
(ISGT LATAM) 2015 IEEE PES. 2015, p. 445–9. 12th Portuguese-Spanish conference on electrical engineering, 2001.
[39] Cai ZH, Liu GF, Luo J. Research state of charge estimation tactics of nickel- [Online]. Available: http://www.uc.pt/en/efs/research/EESEVS/f/XIICLEEE_
hydrogen battery. In: Proc. int. symp. intelligence information processing 1844_PereirinhaTrovao.pdf.
and trusted computing. 2010. p. 184–7. [48] Schmutzler J, Wietfeld C, Andersen CA. Distributed energy resource man-
[40] He HW, Xiong R, Guo HQ. Online estimation of model parameters and agement for electric vehicles using IEC 61850 and ISO/IEC 15118. 2012, p.
state-of-charge of LiFePO4 batteries in electric vehicles. Appl Energy 1457–62.
2012;89(1):413–20. [49] Falk R, S. Fries securely connecting electric vehicles to the smart grid. Int
[41] Battery and Energy Technologies. https://www.mpoweruk.com/lithium_ J Adv Internet Technol 2013;6(1):57–67.
failures.htm. [Accessed 12 October 2018]. [50] Google Maps. https://www.google.com/maps/search/petrol+bunk+near+
[42] Hu X, Zou C, Zhang C, Li Y. Technological developments in batteries: A me/@12.8172622, 79.9754415, 12z. [Accessed 10 October 2018].
survey of principal roles types and management needs. IEEE Power Energy [51] Bowermaster D, Alexander M, Duvall M. The need for charging: Evaluat-
Mag 2017;15(5):20–31. ing utility infrastructures for electric vehicles while providing customer
[43] Zhang C, Wang LY, Li X, Chen W, Yin GG, Jiang J. Robust and adaptive support. IEEE Electr Mag 2017;5(Jan.):59–67.
estimation of state of charge for lithium-ion batteries. IEEE Trans Ind Electr
2015;62(8):4948–57.