Optimizing the electric charge station network of EŞARJ

Gavranović, H., Barut, A., Ertek, E., Yüzbaşıoğlu, O.B., Pekpostalcı, O., Tombuş, O.,
Optimizing the electric charge station network of EŞARJ. In Proceedings of 2nd
International Conference on Information Technology and Quantitative
Management, ITQM 2014. Procedia Computer Science 31 ( 2014 ) 15 – 21.
Note: This is the final draft version of this paper. Please cite this paper (or this
final draft) as above. You can download this final draft from the following
websites:
http://research.sabanciuniv.edu
http://ertekprojects.com/gurdal-ertek-publications/

Gavranović et al. / Procedia Computer Science 00 (2013) 000–000
Optimizing the electric charge station network of EŞARJ
Haris Gavranovića, Alper Barutb, Gürdal Ertekc,*, Orkun Berk Yüzbaşıoğluc,
Osman Pekpostalcıc, Önder Tombuşd
aDepartment of Industrial Engineering, International University of Sarajevo,
Sarajevo, 71000, Bosnia and Herzegovina
bEşarj Electric Vehicles Charging Systems, Istanbul, 34903, Turkey
cFaculty of Engineering and Natural Sciences, Sabancı University, Istanbul, 34956,
Turkey
dDepartment of Industrial Engineering, Maltepe University, Istanbul, 34857,
Turkey
Abstract
In this study, we adopt the classic capacitated p-median location model for the solution of a
network design problem, in the domain of electric charge station network design, for a
leading company in Turkey. Our model encompasses the location preferences of the
company managers as preference scores incorporated into the objective function. Our model
also incorporates the capacity concerns of the managers through constraints on maximum
number of districts and maximum population that can be served from a location. The model
optimally selects the new station locations and the visualization of model results provides
additional insights.
© 2014 The Authors. Published by Elsevier B.V.
Selection and peer-review under responsibility of the Organizing Committee of ITQM 2014.
*
Corresponding author. Tel.: +90-216-483-9568 ; fax: +90-216-483-9550.
E-mail address: ertekg@sabanciuniv.edu.

Keywords: electric charge station network; electric vehicles; charge network; facility
location; information visualization.
1. Introduction
Widespread use of plug-in electric vehicles can significantly reduce greenhouse gas
emissions. Because plug-in electric vehicles do not emit greenhouse gases, they do not
directly contribute to the emissions of these gases. Thus, global greenhouse gas emissions
reduction goals can be achieved or at least approached through extensive adoption of plug-
in electric vehicles. Due to environmental and economical motivations, plug-in electric
vehicles are expected to become increasingly important in the upcoming decades1.
As plug-in electric vehicles enter the market, a huge demand for charging stations is
expected2. To this end, providing adequate charging station infrastructure becomes a
necessity for the success of electric vehicle technology in the market. If sufficient charging
infrastructure is provided, there will be a possible increase in public motivation for this
technology through reducing the plug-in electric vehicle owners’ current anxieties over the
mileage range. Ease of access to the charging stations will affect plug-in electric vehicles
adoption rates, petroleum demand and electricity consumption across the times of a day2.
In this paper, we aim to apply a facility location model to decide on how to install a
constrained number of charging stations for Eşarj3, a leading electric vehicles charge system
operator in Turkey.
2. Literature
Papers are available that consider various aspects of the electric charge station network
design. Frade et al.4 use a maximal covering location model and apply it to Lisbon, Portugal.
They aim to maximize the plug-in electric vehicle demand served by the charging station
infrastructure. Hanabusa and Horiguchi5 develop and solve a model to minimize plug-in
electric vehicle travel cost, assuming a minimum buffer distance between charging stations

as a constraint. They use travel time of the plug-in electric vehicles and waiting time at the
charging stations to develop a cost function for the travel cost. Chen et al.2 use a mixed-
integer optimization model to reduce costs as a function of walk distance between two
districts.
3. Company
Eşarj Electrical Vehicle Charging Systems (http://esarj.com) was established in 2008,
and is rapidly developing in this market since then. R&D operations of the company elevate
it to a prominent position in the electrical charging network business. In 2010, Eşarj signed
a partnership agreement with Efacec Engenharia e Sistemas, a Portuguese company that
leads Europe with significant operations in 65 countries. In 2011, Eşarj made an agreement
with Renault Turkey, and became a solution partner for Renault Electrical Vehicles (EVs).
Eşarj is constantly improving the electricity infrastructure for Renault EVs of Renault
dealers, while extending its charge station network through other means.
Eşarj has particular missions pinpointed to improve current infrastructure facilities in
Turkey. First, Eşarj contributes to electrical charging infrastructure through legislations,
management and engineering. Furthermore, installing an extensive supply network for
electrical charging stations is a priority of the company. Eşarj also provides service for
individual customers with home type charging stations. Additionally, the company works
towards becoming the national network operator of charging stations by providing charging
solutions and network management systems at every charging point.
Eşarj has specific primary missions: Improving service and product quality continually
for customers is the highest priority of the company. The next priority is providing extensive
options for products and solutions. Other visionary priorities include protecting natural
environment and reducing carbon emissions by encouraging customers to use EVs, and
contributing to the energy independence of Turkey in the electrical vehicle market and
electrical energy.
The study that we present in this paper looks into the strategic facility location decision of
Eşarj, following the principle agreement with the leading car-sharing company in Turkey,
YoYo (http://driveyoyo.com). YoYo is the Turkish counterpart of Zipcar, and rents cars for

short intervals, rather than for several days. The partnership guarantees benefits to both
parties. A primary strategy of YoYo is to operate a fleet of electric vehicles, besides fuel-
based vehicles. This will help in decreasing costs and enable YoYo to lead its industry as a
leader of sustainability. Meanwhile, the agreement will allow Eşarj to use the established
network of YoYo in shopping malls and erect vehicle charge stations in these locations,
without having to convince each of the malls, or having to go through long negotiation
cycles.
The strategic problem at hand is simple and classic: Which of the 32 potential new YoYo
locations should Eşarj prioritize, besides its 4 existing non-Renault stations, for erecting
charge stations?
Although the problem can be modeled as a pure p-median problem, some complicating
constraints call for an alternative model, namely a variant of the capacitated p-median
problem.
4. Model
P-median location model locates p charging stations in relation to a set of customers,
such that the sum of the weighted distances between the customers and charging stations is
minimized6. The capacitated p-median location model imposes a capacity constraint on the
locations7. In our study, we adopted the classic capacitated p-median location model with
additional constraints and the consideration of preference scores. The model is taken from
Yaghini et al.7, with the addition of constraint (constraint (4) in the model below).
The applied model is given and explained below. The modeling process begins with the
identification of the sets (the districts and potential locations) and the parameters:
Sets:
set of districts (demand points, customers) to be served
set of potential locations where electric charge stations can be erected
Parameters:
demand at district (population of district )

distance between district and candidate location
preference score for candidate location ; higher values denote higher preference;
{ }
number of charging stations to locate
maximum number of districts to be served by a location
maximum population to be served by a location
The decision variables tell where the stations are located and which districts each location
serves:
Decision Variables:
{
{
The full model, with all the constraints included and the preference scores embedded, is
as given below:
Mathematical Model:
∑ ∑ (1)
∑ (2)
∑ (3)
∑ (4)
∑ (5)
(6)
(7)
(8)

(9)
The objective function (1) is the summary of a set of cost components over i and j. Each
cost component is the multiplication of the demand of the district with the distance traveled
to the location to which it is assigned, divided by the preference score for that location. The
objective function penalizes larger demands traveled over longer distances, but rewards
higher values of the preference scores (provided by Eşarj management). Although there can
be several ways of reflecting the preference scores, this simplest form has been selected for
easy communication with Eşarj and Yoyo.
The constraints are: (2) Each district is assigned to a single location. (3) A maximum of P
locations are to be located. (4) A limit of MaxDistricts is placed on the number of districts
that a location can serve. (5) A limit of MaxPop is placed on the population that a location
can serve. Eşarj was particularly interested in this constraint, because they would prefer a
balanced distribution of traffic to different stations and because they would prefer to install
stations with similar capacities to different locations. (6) The first four locations have
existing Eşarj stations, already decided and located. (7) Linking constraint between the
binary variables. (8, 9) Binary variables.
5. Data
Data for the demand points (districts) was obtained for each district in Istanbul, Turkey
through a commercial database. Then (age 18+ only) population of each district was
obtained from the results of the 2011 general election through another commercial
database. Additionally, coordinates for the potential charging stations was obtained from
Eşarj (which obtained the data from Google Maps) as (lat, long) pairs where lat is the
latitude of the potential location and long is the longitude of this location.
First, all possible points were enumerated with unique IDs to ease the processing of data.
Besides IDs, coordinates of possible locations were also recorded in the spreadsheet to
calculate the distances among the charging points. Then the distance matrix was recorded in
a new sheet.

During data cleaning, all the districts and potential locations have been plotted in a
scatter plot to identify outliers. Fig. 1 displays the initial data as plotted. Because extreme
outlier points (on the outskirts of Istanbul) may distort the optimal solution substantially,
they were removed after communication with Eşarj management. The removed outlier
points are shown in Fig. 2. As can be observed in Figure 2, the outliers are very far from the
clusters of districts, and could result in stations being located near them, disrupting the
service to central districts. It should also be noted that these outlier districts are also
typically smaller villages of Istanbul, where electric cars have not yet penetrated. Thus it is
appropriate to remove them from the dataset.
Fig. 1. Scatter plot of geographical locations of districts Fig. 2. Outlier districts
that have been removed from the data
Finally, the distance matrix was created from obtained data using custom-developed
software (Fig. 3). The software acquires the road distances and puts them into the distance
matrix by automatically querying them from Microsoft Bing Maps. It was crucial for our
problem that the road distances be used instead of the Euclidean distances, especially
because of the Bosporus strait that flows through the middle of Istanbul. Any model built on
Euclidean distances for both the European and Asian sides of Istanbul at the same time
would be highly inaccurate because of the existence of the strait that flows through the two
sides. After its construction, the distance matrix (from the districts to the potential
locations) was used in the OPL optimization model by importing from the data file. After the

data was completed and placed in the model, the model was run using IBM ILOG OPL
Development Studio 6.3 modeling software, with CPLEX as the solver engine.
Fig. 3. Custom-developed Distance Calculator software
6. Analysis and Results
Eventually, Eşarj plans to establish stations to all the shopping malls where Yoyo is
located. However, the current problem is to select the most appropriate 15 locations. The
sensitivity of the objective function with respect to different values of MaxDistricts was
investigated, as in Figure 4. The objective function value itself does not have a particular
economic meaning, but rather is a measure of the incurred cost and the subjective
preferences of Eşarj managers regarding the locations. Finally, the model was run for P=15,
MaxDistricts=90, and MaxPop=100,000.
Figure 5 visualizes the optimal solution. The visualization shows the selected (filled
circles) and unselected locations (cross out signs) and districts (small hollow circles). The
visualization shows that although some selected locations are serving populated regions or
areas with many districts, many districts in the outskirts of the city are far from their

nearest station. Because these outskirt districts are not densely populated, the solution
seems a logical one. The solution, and other results, will be further discussed with Eşarj
managers to devise the most appropriate selections based on Eşarj’s business model and
strategic priorities.
Fig. 4. Objective function value as a function of MaxDistricts, for the model.
Fig. 5. Model solution, for scenario-model combination A4. Small circles show the districts,
cross out signs (x) show unused locations (yj=0), and filled circles show selected locations
(yj=1).

7. Conclusions and Future Work
P-median location model locates p charging stations in relation to a set of customers,
such that the sum of weighted distances between the customers and charging stations is
minimized. In this study, we adopted the capacitated p-median location model with an
additional constraint and the consideration of preference scores, for the solution of a
network design problem. The study presents the application of an optimization model for
the domain of electric charge station network design, although similar models have been
used in other domains. Our model encompasses the location preferences of the Eşarj
managers as preference scores incorporated into the objective function. Our model also
incorporates the capacity concerns of the Eşarj managers through constraints on maximum
number of districts and maximum population that can be served from a location. The
visualization of model results provide additional insights, although developing and solving
various models enables an ad hoc ranking for the prioritization of the locations to be
selected.
Considering the growing importance of sustainability around the globe, our study is
expected to be a benchmark for similar studies by other electric vehicle charging companies
throughout the world.
References
1. Durham City - County Electric Vehicle and Charging Station Plan, 2011. Available from:
http://bit.ly/1dzOoxS or
http://durhamnc.gov/ich/cmo/sustainability/Documents/Durham%20EV%20and%20EVC
S%20%20Plan.pdf. Accesed on March 28, 2014.
2. Chen D, Khan M, Kockelman K. The Electric Vehicle Charging Station Location Problem:
A Parking-Based Assignment Method for Seattle. Under review for presentation in: The
92nd Annual Meeting of the Transportation Research Board and for publication in: Transp
Res Rec. 2013. Available from

http://www.ce.utexas.edu/prof/kockelman/public_html/TRB13EVparking.pdf.
3. EŞARJ Elektrikli Araçlar Şarj Sistemleri - Elektrikli Araç Şarj İstasyonları, 2010.
Available from: http://esarj.com/en/about/esarj/.
4. Frade I, Ribeiro A, Goncalves G, Antunes AP. Optimal Location of Charging Stations for
Electric Vehicles in a Neighborhood in Lisbon, Portugal. In: Transportation Research
Record: Journal of the Transportation Research Board, 2011, 2252: 91-98.
5. Hanabusa H, Horiguchi R. A study of the analytical method for the location planning of
charging stations for electric vehicles. In: Knowledge-Based and Intelligent Information
and Engineering Systems. 2011, p. 596-605. Springer Berlin Heidelberg.
6. Kariv O, Hakimi SL. An algorithmic approach to network location problems. II: The p-
medians. SIAM J Appl Math, 1979, 37(3), 539-560.
7. Yaghini M, Karimi M, Rahbar M. A hybrid metaheuristic approach for the capacitated p-
median problem. Applied Soft Computing, 2013, 13(9), September 2013, 3922-3930.

Optimizing the electric charge station network of EŞARJ

More Related Content

What's hot (20)

Viewers also liked (7)

Similar to Optimizing the electric charge station network of EŞARJ (20)

More from ertekg (20)

Recently uploaded (20)

Optimizing the electric charge station network of EŞARJ