You are on page 1of 4

Proc.

of the 5th International Conference on Power Generation Systems and Renewable Energy Technologies (PGSRET)
26-27 August, Turkey

The Optimal Placement Model for Electric Vehicle


Charging Stations
Kerem Can Arayici Gokturk Poyrazoglu
Electrical & Electronics Engineering Electrical & Electronics Engineering
Ozyegin University Ozyegin University
Istanbul, Turkey Istanbul, Turkey
kerem.arayici@ozu.edu.tr gokturk.poyrazoglu@ozyegin.edu.tr

Abstract— Nowadays, due to the increase in prices and the Furthermore, it is aimed to keep the cost of installation of the
environmental pollution of fossil fuels, individuals and charging station at the lowest level.
governments are tending towards the concept of electric In Section II-A, indices and sets are described for the
vehicles. Therefore, the popularity of electric vehicles is optimization problem in the standard form of MILP. All
increasing. With the development of technologies, the number
parameters and variables such as time periods, locations and
of electric vehicles raised the required quantity of charging
stations. This study introduces an optimization model that deals vehicles, etc. are defined in a linear algebraic way. In Section
with optimal placement of charging stations in terms of II-B, locations which are crowded in terms of business and
investment and ease of use. First, the potential locations for vital areas in the city of Istanbul are discussed. In Section II-
charging stations are determined. With the help of the historical C, the model’s objective function and constraints are
route of electric vehicles, a MILP problem is solved to find the explained in detail. A mixed integer linear programming
optimal placements. In this study, the city of Istanbul is (MILP) model is developed taking the number of vehicles
examined as a case study in a simplistic matter. The model can and their routes as an input to find the optimal charging
achieve the optimal solution regardless of the size of data. stations locations with the considering the installation cost. In
Keywords— Electric Vehicle, Charging Station Placement,
Section II-D, the contributions of the proposed model are
Optimization discussed in comparison to the other studies. Finally, in
Section III and IV, the demonstration of the model’s results
I. INTRODUCTION with the numerical values and the brief perspective for future
The increasing use of fossil fuel shows that in the works are covered.
following years the total fossil fuel supply will gradually II. THE PROPOSED OPTIMAL PLACEMENT MODEL
decrease, and eventually, fossil fuel sources will be
completely exhausted by the end of 2050 [1]. Also, the world The optimal placement of charging stations is a trending
is facing with the global warming issue so technology tends topic in investment management due to an increased number
to grow in the field of renewable energy, smart grid systems, of electric vehicles. Better investment decisions can be made
and electric vehicles (EV). All these technological by developing optimization models to detect the best
developments made for the purpose of living in a sustainable locations for charging stations. Andrews et. al. [3] developed
world. Therefore, the concerns about fossil fuel supplies and an optimization model for detecting the best locations for
global warming make the electric vehicle concept in the charging stations. In this section of the paper, Andrews’s
center of the attraction. In this way, the use of electric model is given first to clarify the contributions of the present
vehicles began to increase day by day. In the environmental study. Then the proposed model by including the missing
side, the penetration of electric cars will reduce CO2 parts of Andrews’s work is discussed.
emissions and greenhouse gases. [1] In addition, according to Andrew’s model consisted of a time period, a set of
the survey conducted by the European Economic Area, by the vehicles and the set of candidate locations. According to his
year 2050, 80 percent of the total vehicle fleet is expected to
be built by electric cars [2]. In other words, electric vehicles MODEL 1 – THE OPTIMAL PLACEMENT MODEL [3]
will affect humans’ life in the following years. min
, , , , (1)
Unlike the advantages, EVs have some problems in ( , , )
comparison to traditional transportation options such as range
problem. In recent years, battery technologies are developing . , , = 1 , ∀ ∈ (2)
( , , )
rapidly and EVs can travel more with a single charge.
However, still, this progress can’t enough for a long journey , , ≤ , ∀( , , ) (3)
without concerning the charging. Therefore, EVs must be
≤ , ∀ ∈ , ∈
charged regularly. Huge amounts of EVs will have an ( , , )∈
, ,
(4)
important impact on the total electricity consumption in the , ,

following years. When these scenarios are considered, there


should be many charging stations to meet the demand. The = (5)

locations should be properly planned and placed so that the
increasing number of electric vehicles will travel , , ∈ {0,1}, ∀( , , ) (6)
comfortably. When starting this project, the focus was to ∈ {0,1}, ∀ (7)
ensure that the distance of the electric vehicles from the
charging stations is at a minimum distance at any time.

978-1-7281-2301-1/19/$31.00 ©2019 IEEE


model data and constraints, a vehicle can charge a station
which is selecting one and the objective function is
minimizing the total distance traveled for the charging
stations. For minimizing the distance, he used the shortest
path algorithm. However, his model didn’t include any
charge of the state of any vehicle. Vehicles maximum and
remaining range weren’t indicated for charging situations.
Moreover, the investment cost of charging stations in
different locations within the city is not considered in the
objective of the model.
A. Indices & Sets
The proposed model is composed of time, specified
Figure 2: Distance of vehicles from each time zone to the next location
vehicle number, and possible charging station locations. A
single day is divided into three equal time zones: morning,
noon, and evening. These are indicated by the index ∈
{1,2,3} in the model and forms the set T. Each electric vehicle
is expressed with the index ∈ {1,2, … , ̅ } and forms the set V.
In addition, each possible charging location is expressed with
the index ∈ {1,2, … , }̅ to form the set J. The distance of the
vehicle v to the charging station j at time t is indicated by , ,
as shown in Fig. 1.

Figure 3: (Morning) Location of vehicles and distances to stations

Figure 1: Indication of the distances of the vehicles to each location

Each possible charging station has a maximum car


charging capacity, which is indicated by . In other words,
each station can charge a certain number of cars at the same
time. The minimum number of charging stations required for
the regular operation of the system is expressed by p.
Whether or not the charging station is running is controlled
in a binary (0,1) manner and indicated by . Likewise, the Figure 4: (Noon) Location of vehicles and distances to stations
situation of any vehicle being present in time t at the station j
is expressed by , , and it is binary (0,1). The distance of the
vehicles from their current location in that time period to the
next location in the next time period is shown in Fig. 2.
B. Case Study
A simple case study is adapted to clearly represent the
effectiveness of the model and to provide a test trial for the
interested audience. A simplistic map of the city of Istanbul
is shown in Fig. 3-4-5 with five possible charging stations.
These locations are in alphabetical order; Bakirkoy, Besiktas,
Cekmekoy, Kadikoy, and Tuzla. In addition, the number of
electric cars in this case study is four and the locations of
these vehicles are known in advance in the morning, noon and
evening to determine their routes. This data may be found by Figure 5: (Evening) Location of vehicles and distances to stations
following a probabilistic approach on the historical route MODEL 2 – THE PROPOSED OPTIMAL PLACEMENT
dataset of electric vehicles to create a realistic study. It is MODEL
assumed that the locations of the vehicles at three time
periods are as shown in Fig 3-4-5 for the morning, noon, and , , ∗ , , + ∗ (8)
evening respectively. The vehicles return to their location in , ,

the morning as their initial positions when starting each day. subject to
C. Objective Function & Constraints , , ≤ 1 , ∀ , ∀ (9)
The objective function of the proposed model as in (8) is
a multi-objective function. It minimizes the installation cost
of charging stations. Equation (9) states that a vehicle can , , ≥ 1 , ∀ (10)
,
only be in one station at any time. Equation (10) states that
the vehicles should be charged at least once per day. Equation , , ≤ , ∀ , ∀ , ∀ (11)
(11) states that the vehicles may only be charged at an
operational charging station. Equation (12) states that the ≤ ∗
, , , ∀ , ∀ (12)
number of vehicles charging at the same time in a station is
less or equal than the maximum capacity of the station.
Equation (13) indicates the minimum number of charging ≥ (13)
stations that should be opened. Equation (14) states that a
vehicle should go to its next destination with the current
amount of charge, and if it is required to go to a charging
station that is not on its way, the remaining distance should , − , + , − , , ∗ , , ≥ 0 , ∀ , ∀ (14)
be considered. Equation (15) states that the state of charge
cannot be greater than the maximum charge capacity of the
vehicle. Equation (16) states that the vehicle cannot discharge , + , − , , ∗ , , /2 ≤ _ , ∀ , ∀ (15)
but can only charge. Equation (17) states that the vehicle can
be charged up to its maximum capacity. Equation (18) states
that the state of charge or the remaining range must be more , ≥ 0 ∀ , ∀ (16)
or equal to the distance to the next destination the vehicle
goes to. Equation (19) states that each vehicle starts a day 0≤ , ≤ , , ∗ _ max , ∀ , ∀ (17)
with an initial state of charge. The last constraint (20) states
that the vehicle range should not be smaller than the distance
= + −
to a station in order to prevent any chargeless situation. ,( ) , , ,

D. Charging stations which are not on the route − , , ∗ , , , ∀ , ∀ : {1, . . (18)

The tensor , , indicates the extra distance that the − 1}


vehicle v should go in order to reach the charging station j at , = , ∀ , ∀ : 1 (19)
time t as shown in Fig. 6. This tensor holds important data to
select the possible charging stations in the current range of , ≥ , , ∗ , , , ∀ , ∀ , ∀ (20)
the vehicle before it runs out of battery. If a charging station
is already on the route to which the vehicle should follow, the charge the vehicle for future use, it must go to the nearest
extra distance is considered as 0 kilometers. This is because charging station which is not on the route. All these
the vehicle will go that way already, no need to spend the conditions allow the model to produce more successful
extra charge. However, if the vehicle does not have a results within the constraints.
sufficient range to arrive its new destination or if it needs to
III. COMPUTATIONAL RESULTS
The optimization problem with the objective function of
(1) subject to the constraints of (2-13) is developed in
MATLAB with YALMIP toolbox [5] and solved through
CPLEX.
A case study consists of separating a day into three even
time zones (morning, noon, evening), selecting five possible
charging stations and four electric cars. The model solves the
problem to optimally select the charging stations to be built
while satisfying the constraints. Because of the data and
restrictions entered, the model successfully showed the
charging status of the vehicles in each time period and the
results showing how far they traveled. It also helped
determine the charge quantity in any time zone. This
demonstrates which of the possible charging stations must be
built according to the received road and charge status
Figure 6: The Tensor of extra distance that the vehicle v should go in order information.
to reach the charging station j at time t
Fig. 8 shows that each vehicle starts the morning period
with the initial range. This range is showed by the red color
in the pie chart. If their initial range is more than their travel
to the next distance, the vehicle might skip charging
immediately; however, the following range requirements are
also important in the decision of selecting a charging station.
Yet, if the initial range is lower the travel, the vehicle must
charge the nearest charging stations. The figure clearly shows
that whenever a car needed to charge and how much it
charged.

Figure 9: Charging status at each time at each charging locations

ACKNOWLEDGMENT
Figure 8: The range of vehicles in each time period, the distance that The authors would like to thank Oguzhan Dalgic and Mert
must be taken when passing to the next time zone, and the amount of
charge necessary if the vehicle needs to be charged so that it can go this
Bektas who initiated this project for sharing their work.
way.
REFERENCES
If a vehicle is being charged in a station, the solution of [1] EU energy, transport and GHG emissions: Trends to 2050,
Variable x indicates “1” at any time period as shown in Fig. ‘PRIMES model for energy and CO2 emissions projections,
9. Through the solution of Variable y, the investment decision Reference scenario 2013’, European Commission, Directorate-
on charging stations are as follows: General for Energy,
[2] Electric mobility in Europe – Future impact on the emissions
• Kadikoy and the energy system, Oeko-Institut and Transport & Mobility
Leuven(TML), 2016
• Cekmekoy [3] M. Andrews, M. K. Dogru, J. D. Hobby, G. H. Tucci, and Y.
• Tuzla Jin,” Modeling and Optimization for Electric Vehicle Charging
Infrastructure,” Alcatel-Lucent Bell Labs, 2012.
IV. FUTURE WORK [4] A. Y. S. Lam, Y. W. Leung and X. Chu, “Electric Vehicle
Charging Station Placement: Formulation, Complexity, and
The authors performed other simulations with a higher Solutions”, IEEE Transactions on Smart Grid, vol. 5, no. 6,
level of data size, and the developed model achieves the pp. 2846-2856, Nov. 2014.
optimal result regardless of the data size. This model [5] J. Lorfberg, “YALMIP: A Toolbox for Modeling and
development constitutes the first part of a large project Optimization in MATLAB,” In Proceedings of the CACSD
consisting of five stages. With the success achieved at this Conference,2004
stage, the estimation of the daily energy demands, and
electric vehicles grid impact studies will be carried out in the
last stages, the overall combination will benefit when
working on the reliability of the national power grid.

You might also like