Renewable and Sustainable Energy Reviews 153 (2022) 111719

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

Too much or not enough? Planning electric vehicle charging infrastructure:

A review of modeling options
M.O. Metais a,b ,∗, O. Jouini a , Y. Perez a , J. Berrada b , E. Suomalainen b
a
Université Paris-Saclay, CentraleSupélec, Laboratoire de Génie Industriel, 91190 Gif-sur-Yvette, France
b
Institut VEDECOM, 23bis Allée des Marronniers, 78000 Versailles, France

ARTICLE INFO ABSTRACT

Keywords: While the transportation sector is responsible for a growing share of greenhouse gas emissions, electric
Electric vehicle vehicles (EVs) offer solutions for greener mobility. The proportion of electric vehicles in transportation fleets
Charging station is increasing, but wider adoption will not be possible without an appropriate charging infrastructure. The
Charging infrastructure planning
deployment of such infrastructure should follow a strategy that considers both the environment in which it is
Location problem
deployed and the behavior patterns of electric vehicle users. If these aspects are not taken into consideration,
Sustainable mobility
there is a risk of failing to meet users’ needs and generating additional costs. Here we review the literature
on location problems for electric vehicle charging stations. We aim to draw up a comparative overview of
approaches that have been used up to 2020 for optimizing the locations of charging infrastructure. We first
briefly review the issues raised by the deployment of charging infrastructure, namely technical, economic and
user acceptance concerns. We then look at the goals of the infrastructure location models in the literature.
Schematically, those goals fall into two categories: minimizing the cost of charging infrastructure for a given
level of service, or maximizing the service provided for a given cost. Finally, we focus on the approaches used
to achieve these goals. Three categories of approaches are identified: node, path, and tour- or activity-based
approaches. We then discuss these approaches in relation to technical, economic and user acceptance factors
in order to provide a comprehensive analysis for stakeholders involved in EV charging infrastructure planning.
Directions are given for future research to develop models that better reflect the real-world picture.

1. Introduction pollution and exposure to nitrogen oxides, volatile organic compounds,

and carbon monoxide in urban areas, and reduce particulate matter
The climate emergency requires a drastic and rapid reduction in emissions – and far less noise than ICE vehicles.
anthropogenic greenhouse gas (GHG) emissions, which are the cause Despite all these benefits, large-scale uptake of EVs is bottlenecked
of the fastest global warming ever observed [1]. The transportation by a number of different barriers [4]. A first major barrier is the high
sector is responsible for about 15% of global GHG emissions (27% in purchase price of EVs compared to ICE vehicles, although the purchase
the European Union), and this rate is expected to increase in the coming price impact is expected to diminish shortly. When considering total
years [2]. A transition from internal combustion engine (ICE) vehicles cost of ownership over the whole life cycle, an EV is already less ex-
to greener transportation could be a major lever for reducing global pensive than an ICE vehicle in countries such as Norway or France [5].
GHG emissions. Moreover, the purchase price of EVs is projected to drop below that of
For road transportation and individual mobility, which account for ICE vehicles by 2025 [6]. The second main barrier for users is tied to
the largest share of transportation-sector emissions, electric vehicles range anxiety. Most EVs have a lower driving range than ICE vehicles.
(EVs) emerge as a major alternative to ICE vehicles. Considering the
Even though the range offered by a full-charge battery is sufficient for
whole lifetime of the vehicle, EVs have a lower global warming poten-
daily use for a large majority of users, they fear that they will run out of
tial than ICE vehicles, especially if they are coupled with low-carbon
battery before being able to finish their trips or find a charging point.
electricity production systems [3]. Moreover, EVs have many other
User anxiety is thus the main problem to address to enable large-scale
benefits, such as no tailpipe emissions – which could help avoid air

Abbreviations: AC, Alternative Current; DC, Direct Current; EV, Electric Vehicle; EVSE, Electric Vehicle Supply Equipment; FCLM, Flow-Capturing Location
Model; FRLM, Flow-Refueling Location Model; GHG, Greenhouse Gas; ICCT, International Council on Clean Transportation; ICE, Internal Combustion Engine;
MCLM, Maximum Covering Location Model; PHEV, Plug-in Hybrid Electric Vehicle; SCLM, Set Covering Location Model
∗ Corresponding author at: Université Paris-Saclay, CentraleSupélec, Laboratoire de Génie Industriel, 91190 Gif-sur-Yvette, France.
E-mail address: [email protected] (M.O. Metais).

https://doi.org/10.1016/j.rser.2021.111719
Received 25 November 2020; Received in revised form 9 September 2021; Accepted 27 September 2021
Available online 23 October 2021
1364-0321/© 2021 Elsevier Ltd. All rights reserved.
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

Fig. 1. Overview of the charging infrastructure framework.

EV adoption. The way forward could be to increase battery capacity to literature proposing strategies and simulation models for deployment of
improve EV range or to provide an efficient charging infrastructure to electric charging infrastructures, considering the technical, economic
better cover charging needs. However, even with a larger range, the and user-side aspects of EVs.
fear of not being able to charge EVs when the battery is empty is still To identify the first relevant articles, the Google Scholar database
the same [7], so large-scale EV deployment cannot be achieved without was searched with combination of keywords : {EV, electric vehicles,
a prior appropriate charging infrastructure [8]. Furthermore, research charging infrastructure, charging stations} and {planning, location,
shows that investing in charging infrastructure is more efficient than model, optimization}. We kept a sample of 287 articles containing
subsidizing larger batteries as long as the investments in charging literature reviews and papers on infrastructure optimization and de-
infrastructure are not sufficient to cover the whole territory [9–11]. ployment models cited as references in this field. The articles cited in
However, deploying a charging infrastructure is hugely expensive these papers and the articles also citing them were then screened, and
and comes with several technical and economic constraints. The Energy we added 63 relevant articles to our review.
Transition for Green Growth act in France sets a target of 7 million The paper is structured as follows. Section 2 explains the different
EV charging stations (public and private) by 2030, which corresponds charging technologies and the issues involved in deploying charging
to a minimum cost of around 2 billion euros [12] while an ICCT infrastructure. Section 3 presents the objectives and targets of infras-
report projects an estimated 1 billion dollars in investment over the tructure deployment. Section 4 then covers the methods for locating
2019–2025 period for the USA to fill its public charging infrastructure and sizing infrastructure in a territory, and Section 5 highlights gaps in
gap [13]. These huge costs warrant a proper deployment strategy to the literature and avenues for future research.
efficiently locate and scale new charging stations in order to favor
large-scale EV adoption while avoiding resource waste or underinvest- 2. Background on charging infrastructure and the allied issues
ment for infrastructure investors. This deployment, with the costs it
entails, also faces a chicken-and-egg problem: drivers will be reluctant The issue of deploying charging infrastructure for EVs is set in
to buy an EV without adequate infrastructure, while operators will the following framework: EV users with limited autonomy travel the
refuse to invest in infrastructure until there is sufficient demand to road network. Making these trips consumes energy, which in turn
make it profitable. To ease this bottleneck, the first step must be taken decreases the state of charge of the EV battery and creates a need to
by operators [14]. charge, which can be met in two ways: either through home/office
Once the first step has been taken, the issue of optimal deployment charging, or through public (or semi-public) charging infrastructure.
of a vehicle refueling infrastructure is not a new challenge. Coverage This infrastructure needs to stay at a reasonable cost for operators,
and location models, such as those of Toregas [15] or Hodgson [16], who have limited investment capacity, while giving EV users the trans-
have been around for a relatively long time and are perfectly applicable portation network coverage they need. The goal is to enable drivers to
to gas refueling stations. However, EVs have different demands to ICE use their EVs with less range anxiety, knowing that they can rely on
vehicles (charging takes longer than refueling), which makes these public charging infrastructure when they need it. As public charging
coverage models incompatible with routine EV use. Models taking these infrastructure supplies energy from the grid, infrastructure deployment
specificities into account have thus been developed since the end of the needs to consider the constraints linked to power grid operation (see
2000s. Fig. 1).
Nevertheless, few of them seem to take advantage of the benefits In this framework, three main types of issues are to consider
offered by electric vehicle charging, which does not require the user to when deploying charging infrastructure : technical, economic and
be present during charging time. Moreover, the deployment of such an user-centered issues.
infrastructure does not happen all at once, partly because of the prob-
lem of the development costs it would generate without a guaranteed 2.1. Technical overview of charging devices
return on investment from a demand that will take a long time to come,
which brings us back to the previous chicken-and-egg problem. An Charging devices provide the link between electricity grid and EVs
incremental and over-time deployment must be considered, considering by converting AC power into DC power, which can charge a battery.
the early stages of the infrastructure already present in the territory. They can be on-board or off-board, depending on the type of charging.
This literature review aims to provide an overview of the timely The International Electrotechnical Commission (IEC) defines four
problem of EV charging infrastructure planning in terms of the op- charging modes [17]. In the first three modes, the EV is directly
timization models used to determine optimal locations of charging connected to the AC distribution network, and the conversion to DC
points, and sizing. It explores and compares a rapidly growing scientific is done through the vehicle’s onboard charger. The main difference

2
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

Fig. 2. Simplified architecture of EV charging.

between these three levels lies in the level of safety and charging have a single-family home with a private parking space where they can
control, which allow the vehicle battery to be charged with more or less install a charging point [22]. This illustrates that relying solely on home
power. For example, mode 1, which is used for low-power charging, is charging for the transition from ICE vehicles to EVs will leave important
equivalent to plugging the vehicle into an electrical outlet, while mode barriers to the adoption of electric mobility, justifying the need for an
3 allows advanced charging control and higher charging power. The appropriate public charging infrastructure.
mode 4 is mostly used for fast charging applications. Unlike the three The range and charging constraints of EVs make it illusory to
first modes, here the connection of EVs to the AC grid is not direct: the envisage the broad diffusion of EVs without sufficient charging in-
AC power is converted into DC power in an off-board charger, and then frastructure. At the same time, if there are not enough EVs on the
used to charge the EV’s battery. Fig. 2 gives a simplified illustration of road, there will not be enough interest in setting up an expensive,
EV charging. unprofitable charging infrastructure. But without this infrastructure, it
is illusory to envisage the democratization of EVs... The chicken-and-
2.2. Charging infrastructure and EV acceptance egg problem in this two-sided market has been studied by Delacretaz
et al. [23] who show that an initial infrastructure has little immediate
Charging EVs generally requires much more time than filling up an positive effect on EV adoption but that positive effect does increase
ICE vehicle gas tank. Charging times go from a few dozen minutes for over time. They also show a snowball effect: the demand elasticity for
the fastest chargers up to more than 20 h for the slow ones [18]. Charg- EVs relative to charging infrastructure provision increases with infras-
ing stations thus have different design and management imperatives to tructure development. In other words, the more charging stations there
conventional gas stations. EVs have different refueling behaviors due are, the greater the increase in EV demand with further investment in
to different required charge-times and charging locations, especially charging infrastructure.
when taking into account one of the major conveniences offered by This raises the question of marginal – or incremental – infrastructure
EVs, i.e. that EV batteries can charge while the vehicle is not in use development. An infrastructure is deployed in a spatial context, but
for mobility purposes (while parked at home, the workplace, in mall also in a temporal one, and it is unrealistic to consider instantaneous
parking, etc.). Home EV charging does not require any effort from deployment of a complete set [24]. It is therefore important to define
the driver other than plugging in the EV. It also does not require a temporal deployment sequence along with a spatial set of locations
any specific installation – at least not for slow charging [17] – even to determine the most cost-effective investments [25]. Otherwise, the
if the majority of EV drivers install specific equipment to increase risk is to end up with an infrastructure unsuited to driver needs at the
charging speed and for safety aspects. Moreover, in 90% of cases, trips beginning, which would not allow the diffusion of EVs to start and thus
do not exceed 80 km, whereas the typical range for an EV is about discourage additional investments in infrastructure, and so on (again, a
200 km1 [20,21]. Thus, home charging should be sufficient for a large chicken-and-egg paradigm). In addition, even though charging stations
majority of users: with a fully charged battery when leaving home, they are often deployed without a global vision, they nevertheless already
could complete their daily trips and charge their EV once back home, exist in the territory, and it would be a mistake not to consider this
ready for the next day. existing resource. We must therefore think about the problem of placing
However, home charging has some limits. First, if trips – or a ‘one more charging station’ and the value of this station when there is
succession of trips – exceed the EV range, then drivers need to be already a set of operational stations, while almost all the models focus
able to charge their EV elsewhere than at home. If this is not possible, on optimizing the final charging infrastructure without considering the
then EVs will remain as a second car for the wealthiest percentiles process to get there.
of the population, since users will not be able to use it to make
occasional long journeys and will therefore prefer an ICE vehicle [8].
2.3. Economical issues
Moreover, in many countries, a large part of the population do not
A naive approach would be to consider the best option is to put fast
1
Note : This is valid for the European market, where the population
chargers everywhere, as people value the option to charge quickly [26].
densities are rather high and the distances to be traveled relatively small. In However, a DC fast charging station costs much more than a slower
the case of the US or similar markets, the distances involved may be higher. one. The average cost for a level 2 public charging station is $3000,
However, they remain well below 200 km, which is already a pessimistic while the average cost of a DC fast charging station is nine times more
estimate of the range of a standard electric vehicle [19]. (Table 1).

3
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

Table 1 2.5. Summary

Electric vehicle supply equipment (EVSE) purchase and installation costs in the U.S.
[13].
In summary, the charging infrastructure for EVs needs to address
EVSE type Average public Average home
installation cost installation cost technical issues linked to the technology used and the constraints it
Level 1 $4000 $400–$900
places on existing grid infrastructures. It also meets financial chal-
Level 2 $6000 $680–$4100 lenges: the costs related to charging infrastructure are relatively high,
DC fast charging (50 kW) $73,000 Not available so it is important for operators to avoid making unnecessary invest-
DC fast charging (150 kW) $120,000 Not available ments and to be assured of a return on their investments. Finally,
DC fast charging (350 kW) $205,000 Not available
charging infrastructure needs to respond to user demand in order to
garner user acceptance of EVs.
The financial stakes and public acceptance of EVs are closely linked:
Since more expensive infrastructure should lead to more expensive insufficient coverage of the territory, i.e. underinvestment or unwise
charging service for users, a poor choice of electric vehicle supply investment, will discourage users from buying and using an EV. This
equipment (EVSE) penalizes not just the consumer but also the operator in turn will have consequences on return on investment, as would
for whom a charging station adapted to local needs guarantees a prohibitively high utilization cost of the infrastructure. The technical
better return on investment. Let us explain this with a simple example. constraints linked to the charging station energy supply can lead to
Suppose a charger able to fully charge an EV in three hours is placed in significant additional costs linked to the electrical network. Finally,
a site where parking times are usually eight hours. A person who leaves the adoption of EVs requires a charging infrastructure technology that
their EV parked and plugged in will charge for a maximum of three meets users’ expectations. Users expect to have at their disposal an
hours but then unnecessarily occupy the terminal for the remaining infrastructure that suits their needs in a convenient way, and that they
time. However, for the same budget, several slower charging stations can rely on.
could have been installed which would maximize the profit for the
operator and the level of service for users.
3. Overview and scope of planning simulation models
Finally, it is important in the case of several operators that they
coordinate with each other to ensure interoperability and good global
An appropriate EV charging infrastructure has to satisfy technical,
coverage. But it is also important to put in place regulations to prevent
economic and acceptability constraints. The infrastructure must address
the creation of local private monopolies in public parking areas, which
would be harmful to users [27]. a threefold issue: its location, i.e. its distribution on the transport
network, its capacity, i.e. the demand it can serve, and its users. In
2.4. Power grid issues addition, infrastructure deployment can serve different goals depending
on the interests of those deploying it on the transportation network.
Another issue in charging station location concerns the power grid. To describe a transportation network in a location problem, we
Level 1 infrastructure only requires about 3 kW from the power net- decompose it into nodes and paths (or links). The simplest strategy is
work, which is no more than common household appliances. This to define one in relation to the other: a path or route is a link between
should not have a big impact on the wider grid, even when several two nodes, and a node is the intersection point of two paths, or can be
EVs are simultaneously charging, or at least not one that a small tariff the end of a path too.
incentive could not solve. However, current fast chargers can require Users make trips in the transportation network, i.e. they travel
up to 150 kW from the grid, which is not necessarily scaled for that, between two nodes. They also make tours, i.e. series of trips. During
especially if there are several fast chargers at the same place, as is the those trips, EV users use energy from their batteries, and sometimes
case with charging hubs. Placing chargers requiring too much power in need to charge their EVs with charging infrastructure in public space.
non-adapted locations can stress the existing infrastructure and lead to
the need for grid reinforcement, which can be very expensive [28]. 3.1. Users and charging infrastructure utilization
The choice of charging station type and placement can therefore
be a source of cost inefficiencies. To control the total cost of the
Location and sizing of the charging infrastructure must meet user
infrastructure, this choice must be considered in relation to the real
demand. The literature mainly focuses on three types of charging-
needs of users, as well as the capacity of the power grid.
infrastructure users: buses, taxis, or private vehicles. A classification
Charging EVs is not simply a source of grid stress but also a potential
has been made in the table in Appendix.
source of grid stability if combined with smart grid management to
Charging infrastructure is easier to design for buses, as buses have
exploit positive synergy with renewable energy production. Renewable
fixed tours with (more or less) precise time schedules, so uncertainties
energies are a source of stress for the electricity grid, as they are not
about their state of charge, availability or itinerary is quite low. For this
or only partially controllable. Therefore, they sometimes produce too
much energy with respect to the needs, and this surplus of energy has problem, there are two options. If the buses have enough autonomy to
to be used. The batteries of EVs can then store this surplus energy run all day long without being charged, they can simply be charged at
produced by renewables to smooth out excess power output. Scheduling the end of their shift at charging stations installed at the depot. The
the charging of vehicles according to the constraints of the electrical second option is to place fast charging stations at bus stops to allow
grid is often called ‘‘smart charging’’. all buses to complete their tours, as described by Wang et al. [32]. The
EVs can also provide additional power to the grid during grid stops at charging points do not even have to be longer than at other
stress episodes by injecting electrical power from their battery into the bus stops, as current flash charging technology is able to charge two or
grid, like a generator [29]. This bi-directional mechanism is commonly three kWh into bus batteries in a couple of seconds. In this case, the
called Vehicle-to-Grid (V2G). Through this mechanism, EVs can notably choice could even be made to place a fast charging point at each bus
flatten consumption peaks and play a role in regulating grid incidents stop, allowing the buses to be equipped with low-capacity batteries.
by providing ancillary services to the grid, such as frequency regula- These two options are not mutually exclusive, and it is possible to
tion, resource adequacy, network deferral, energy arbitrage, spinning charge buses at night and add charging stations for buses that are
reserve, etc. [30] They can also directly fast-charge other vehicles, unable to complete their tours.
avoiding power demand peaks from fast charging on the distribution For taxis, as fuel is a large part of their costs and they mainly
network [31]. make short trips, EVs could be an excellent option, and taxis could

4
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

become a good showcase for the usefulness of EVs,2 but the charging global distance they can travel (Wang et al. [49]), which is almost the
infrastructure has to meet specific requirements. First, electric taxis same. This prevents many vehicles being covered by a single station,
cannot charge during trips with a customer: they have to charge during as can be the case with the previous objective. However, in this case
idle time. However, these downtimes must be as short as possible. This charging 10 kWh in a single EV is the same as charging 1 kWh in
is why it is critical here to consider the time spent at the charging ten vehicles, regardless of whether the intended trip is feasible for the
station (waiting time and charging time), as it is idle time for the vehicles. This is why some papers aim to build a charging infrastructure
driver. Taxis already have many idle time situations, typically when that minimizes failed – or maximize feasible – trips (Asamer et al. [50],
waiting for customers. Charging time should not be an additional heavy Micari et al. [51]).
constraint. Ideally, it should be available where and when taxis have These optimization objectives are implemented using various loca-
idle time. Then, because it is common for taxi drivers to share a vehicle, tion methods, as detailed in Section 4.
home charging is not always an option: the charging infrastructure for
taxis must allow them to operate continuously. The fast charging option 3.3. Sizing charging infrastructure
is therefore often preferred for taxis.
Most of the literature focuses on private vehicles, which account With the problem of location comes the problem of sizing the
for the biggest share of the vehicle fleet, or at least considers that an charging stations at the chosen locations. This is mainly a matter of
infrastructure can be developed for all light vehicles. Private vehicle answering two questions: how many charging points should be placed
owners have a wide variety of uses for their vehicles depending on their at a location, and which charging speed should be chosen. Locations are
environment (rural, urban), travel habits (distance from their main also dependent on station capacity, i.e. the number of vehicles that can
points of interest, frequency of travel), and many other factors. The be served within a certain period. For example, if a station with a large
different ways of looking at the case of private vehicles are detailed in service capacity is installed at one location, there is limited interest in
the rest of the paper. placing another station near it.
Some studies only focus on the problem of locating charging sta-
3.2. Optimization goals tions, sometimes considering an infinite capacity [43,52] which does
not represent a real situation where charging points can only accommo-
The literature has considered several optimization goals to effec- date a limited number of vehicles. But once locations have been found
tively meet user charging demand. without considering this limited capacity, charging stations can be sized
Many studies aim to minimize the infrastructure costs for meeting a according to the demand at each station, as in Micari et al. [51]. The
given demand, thus taking demand as a constraint. Like infrastructure sizing can be done simply with the number of EVs likely to need each
costs, some papers only take the installation costs into account. These station, or by more sophisticated models such as queuing models that
can be a simple fixed cost for any charging station, which can be can consider the randomness of charging demand, as in Zhu et al. [53].
actualized considering its life-cycle as in Dong et al. [33]. In this case, However, not considering the capacity of charging stations in a first
the objective narrows down to finding the configuration that allows step of charging station location planning can lead to sub-optimal
to have as few stations as possible. The cost of charging infrastructure results, as the size of the stations influences their distribution over a
can also be made more complex if we consider the different costs of territory.
chargers and the construction costs, land costs as in Mehar et al. [34], Some models directly take into account limited capacity of their
or network reinforcement costs as in Rajabi-Ghahnavieh et al. [35]. charging stations as a constraint, like the models proposed by Upchurch
Others take into account both investment and operation costs, such as et al. [54] or Gavranovic et al. [55]. By doing so, it is possible to
maintenance costs or cost of electricity (Jia et al. [36]). consider disparities in demand and avoid, for example, an area with
With a view to achieving profitability, several papers also aim to a high concentration of demand being covered by only one station
maximize the utilization of chargers (Cai et al. [37], Pevec et al. [38]). that will not be able to satisfy all the demand in its area. In addition,
Other studies choose to deal not with the infrastructure cost but multiplying the number of stations in areas of high demand reduces
with minimizing the user’s costs. User costs are mainly tied to time the impact of a failure of one of them, which is important for the
spent waiting at charging stations (Hanabusa et al. [39], Tu et al. [40]), reliability of the infrastructure. Unlike the previous method, however,
and the trip – or the deviation from their original path – they have to this approach leaves little flexibility in terms of the size of each station,
make to charge their vehicle (Ge et al. [41], Xu et al. [42]). since this parameter must be set beforehand.
Some papers choose to focus on maximizing the number of EVs that Sizing the charging infrastructure is not just a matter of deciding the
could be charged at the station. In other words, the objective is to number of vehicles that can be accommodated, but also the time spent
maximize EV flow at the charging station, based on the rationale that at the station. It is not always inconvenient that the charging process
the more people have access to the infrastructure, the more useful it is. takes several hours, but this is not always acceptable, such as during
Some models only consider a location problem and provide a geograph- long journeys requiring a quick charge to reach the destination. That is
ical coverage of the demand (Wang and Wang [43], Motoaki [44]). why it is also important to wisely choose the power level of charging
In this case, the objective is to have a maximum number of EVs stations based on the use case, and many models incorporate power
with access to a potentially available station, and the charging station sizing (You et al. [56], Wang et al. [57]). This sizing can also be done
locations are uncorrelated to the charging station sizes. Other works with each type of station chosen according to the type of targeted route,
consider the availability of the station, by introducing charging time which allows fast charging stations to be placed where a quick charge
during which the station is unavailable (Sun et al. [45]), or queuing is most useful. Indeed, even if increasing the charging speed of a station
models (Yang et al. [46]). This allows to address the question of sizing also increases its capacity as it serves EVs faster, slow charging stations
the infrastructure. are a more cost-effective option to meet the needs of a whole territory
An alternative to maximizing EV flow is to maximize the amount of (Sun et al. [45]).
energy charged by the EVs (Chen et al. [47], Csizar et al. [48]) or the Finally, charging stations must be sized by considering grid capacity
at the location of the charging points. As explained earlier, a large
number of charging points at the same place or high power charging
2
Note: In Amsterdam Airport Schiphol, where there is the largest Tesla points cannot be installed where the electrical grid is too weak, at
taxi fleet, taxis are massively using the free infrastructure provided by the car the risk of causing instabilities due to excessive power demand [58].
company, making it the most intensively used charging infrastructure in the Some studies choose to take the characteristics of the electrical grid as
world. a constraint (Zhu et al. [53], Zhang et al. [59]), and a few consider the

5
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

Fig. 3. Simplified principle of a node-based approach.

possibility of reinforcing the electrical grid (Sadeghi-Barzani et al. [60], method based on vehicle-routing logics with the aim of making all
Guo & Zhao [61]). Other grid-related issues, such as peaks in demand transportation network nodes accessible to each other. Later, Wang and
or power quality, may also arise because of charging infrastructure Wang [43] used an SCLM to cover the maximum demand for both intra-
that does not take the power grid into account [62] or because of a and inter-city trips while minimizing cost, assuming that the capacity
power grid that does not take the charging infrastructure into account, of each station is unlimited.
depending on the point of view. Another node-based approach is the Maximum Covering Location
Model (MCLM) [66]. Its objective is to locate a given number of
4. Location methods facilities to maximize coverage of the demand, considering a critical
distance as the SCLM does: a facility covers a demand node if the
Several methods to locate charging infrastructure have been devel- distance from facility to node is under this critical distance. Unlike the
oped, and most can be grouped into three main categories: node, path, SCLM, the MCLM allows some demand nodes to not be covered, so can
or tour-based approaches [63]. be used when resources are insufficient to cover all the demand nodes,
The node-based approach is the most popular method for locating as is often the case in reality. However, both SCLM and MCLM consider
charging stations. It deals with the location problem as a facility the distance to determine if the demand node is geographically covered
location problem, which has been extensively studied for many ap- by the facility, without taking into account the impact of that distance:
plications [64]. The problem to be solved is formulated as follows. placing a plant at a demand node or at the node’s critical distance is
Given candidate locations which are the nodes, the objective is to place the same thing. Frade et al. [67] used the MCLM in a case study in
facilities, i.e. the charging stations, to meet the demand at the nodes. Lisbon, Portugal to determine the locations of charging stations and
Even if it seems a simple formulation, this problem belongs to the then sized the stations according to the demand in each zone covered.
NP-hard class, meaning that we are not able to find exact solutions Sun et al. [45] used a node-based maximum coverage model to locate
in a reasonable time because the corresponding resolution algorithms slow charging stations in competition with fast charging stations placed
have an execution time that increases exponentially in the problem with a flow-capturing model (see later). Wagner et al. [68] used a
dimension. Heuristic methods are often used to provide approximate maximum coverage optimization and quantified the value of putting
solutions in a reasonable computing time. The principle of the method a charging station at points of interest such as schools or stores.
is illustrated in Fig. 3. The 𝑝-median model first introduced by S. Hakimi [69] is now
A second approach that has been considered is the path-based one of the most widely-used models in facility location problems. The
approach, introduced by Hodgson [16] and illustrated in Fig. 4. This objective of a 𝑝-median problem is to determine where to place 𝑝
approach relies on a flow-capturing model: the objective is to place facilities among candidate locations to minimize the transportation
charging stations along paths with the highest flows of vehicles, con- cost (or weighted distance) between customers and facilities, with
sidering origin–destination trips, in order to serve as many users as each customer assigned to a facility. The problem can be capacitated,
possible. It considers effects that only emerge from the demand em- meaning that the facilities have capacity restrictions on the amount of
anating from vehicle flows, whereas the node-based approach offers a demand they can serve, and so the demand from customers assigned to
relatively static view of demand. this facility cannot exceed this capacity. In the case of charging stations,
Last, the tour-based approach, illustrated in Fig. 5, does not just this means that only a limited number of cars can be served within
consider individual origin–destination trips but the entire activity of an a certain period, and therefore the availability of the station depends
agent and its vehicle during a period. It considers origin, destination, directly on its capacity. Gavranovic et al. [55] used this model on a
distance traveled, vehicle paths and dwell times, to choose the best subset of potential locations in Turkey, considering the demand and
places to put charging infrastructures according to users’ behavior. the preferences of local stakeholders. Jia et al. [70] separated the need
for fast and slow charging, and used the 𝑝-median model to locate
4.1. Node-based approach fast-charging stations. Jung et al. [71] also used the 𝑝-median in a bi-
level problem to locate charging stations for taxis, while minimizing
The Set Covering Location Model (SCLM) is a facility location model both distance to travel to the station and queue at the station. He
that aims to minimize the number of facilities while covering all the et al. [72] estimated charging demand through socio-demographic data
demand from the customers [15]. In this model, facilities are located in Beijing and used this estimation as an input for all three node-based
in such a way that all demand points are not further from a plant models (SCLM, MCLM and 𝑝-median). They found that the 𝑝-median
than a certain determined distance. It assures all the consumers that model outperform SCLM and MCLM, and gives more stable solutions.
they can find a facility under this distance, but does not consider the An et al. [73] developed a two-stage optimization framework that
demand: all the demand points have the same weight, they just have considers the disruptions that could lead to charging demand changes.
to be covered. Wang and Lin [65] adapted this method and proposed Table 2 gives an overview of the node-based methods applied to EV
a refueling-station-location model using a mixed integer programming charging stations location.

6
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

Fig. 4. Simplified principle of a path-based approach.

Fig. 5. Simplified principle of a tour-based approach.

4.2. Path-based approach 4.3. Tour-based approach

The third method is the tour-based approach, sometimes also called

Instead of dealing with demand at nodes, Hodgson [16] introduced activity-based. Jia et al. [36] proposed a model with the estimation
a path-based version of the MCLM called the Flow-Capturing Location of vehicle charging demand based on parking demand, measured in
Model (FCLM) with the hypothesis that traffic in a network can be vehicle-hours. They assumed that the more occupied the parking slots
served by several facilities located on common paths. The FCLM con- are, the more charging demand there will be, regardless of turnover.
siders origin–destination pairs and aims to maximize the flow captured Chen et al. [47] developed a parking-based model that considers the
on the shortest path between origins and destinations. In this model, duration of parking time but excludes home parking. Cavadas et al. [92]
the path is considered covered if it passes through at least one node also considered the possibility of demand transference between charg-
with a charging station. ing sites for users, meaning that the charging demand on distinct
places can be transferred between those sites according to the users’
The FCLM was later extended. Kuby and Lim [82] developed the
activities. You et al. [56] adopted a strategy based on missed trips in
Flow-Refueling Location Problem (FRLM) especially for alternative-fuel
tours. Their optimization model tries to minimize the number of tours
vehicles that considers the limited range of the vehicles, as a vehicle
that could not be done due to a lack of charging stations. Andrews
may have to stop at more than one refueling station in order to com-
et al. [93] adopted a similar approach on missed trips but considering
plete a path. They found that placing charging stations only at nodes the available charging infrastructure. They developed a ‘user charging
would not be sufficient to provide total coverage, and then developed model’ that determines where and how EV users need to charge given
a method to locate stations on links [83]. Then, with Upchurch [54], the available charging methods. If a vehicle fails its trip due to a lack
they went on to develop the CFRLM, which is a FRLM with capacity of infrastructure, it is taken as an input in an optimization program
constraints on the refueling stations. Wang et al. [57] used this model to place new charging stations. Cai et al. [37] proposed a data-driven
to place different kinds of stations, as previous models only take into method based on taxi data to put charging stations in existing gas
account one type of charging stations. Kim and Kuby [84] then devised stations. They extracted stop events to find charging opportunities at
an optimization model that considers the deviations from the shortest the different stations and estimated the potential charging demand for
path that drivers should have to make to refuel their vehicle, and stop points in gas stations by evaluating state of charge according to
Huang et al. [85] proposed a model with the possibility of multiple previous tours. Shahraki et al. [94] used a similar method but focused
on plug-in hybrid electric vehicles (PHEV). They looked at dwelling
deviation paths. Li et al. [25] proposed a ‘multi-period multi-path’
time between trips and estimated the state of charge of batteries
FRLM with the objective to minimize the total cost of installations
after each trip, then placed charging stations to minimize the distance
while making each trip feasible via at least one path between origin
traveled by PHEV in combustion-engine mode. Gonzalez et al. [95]
and destination within a reasonable tolerance compared to the shortest adopted a similar approach from simulation data, with an optimization
path, and considering the dynamics of the network over time. Further, concerning vehicles not able to complete their daily trips without
Wu and Sioshansi [86] developed a stochastic FCLM model that takes modifying their initial behavior to charge their EV while considering
into account the uncertainty of EV charging demand as soon as the electricity price fluctuations in order to minimize charging cost. He
infrastructure is built in anticipation of future EV adoption. Table 3 et al. [52] determined a bi-level tour-based model with traffic network
gives an overview of path-based methods. equilibrium considering interactions between trips and charging needs

7
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

Table 2
Summary of articles using the node-based approach.
Method Problem Main optimization goal Paper
MCLM Location Maximize the number of EVs charged Frade et al. (2011) [67]
MCLM Location Maximize the number of EVs charged Guo & Zhao (2015) [61]
MCLM Location and sizing (capacitya ) Maximize the number of EVs charged Wang et al. (2013) [57]
MCLM Location and sizing (capacity) Maximize the number of EVs charged Gopalakrishnan et al. (2016)
[74]
MCLM Location and sizing (powerb Minimize the infrastructure cost for a given Yang et al. (2017) [46]
and capacity) demand
MCLM Location and sizing (power) Maximize the amount of energy charged Wagner et al. (2013) [68]
MCLM Location and sizing (power) Maximize the number of EVs charged Liu, J. (2012) [75]
MCLM Location and sizing (power) Minimize the infrastructure cost for a given Deb et al. (2019) [76]
demand
p-median Location Minimize the distance (or deviation) to a charging Xu et al. (2013) [42]
station
p-median Location Minimize the distance (or deviation) to a charging Gavranović et al. (2014) [55]
station
p-median Location Minimize the infrastructure cost for a given Jia et al. (2014) [70]
demand
p-median Location and sizing (capacity) Minimize the distance (or deviation) to a charging Ge et al. (2011) [41]
station
p-median Location and sizing (capacity) Minimize the infrastructure cost for a given Mehar et al. (2013) [34]
demand
p-median Location and sizing (capacity) Minimize the infrastructure cost for a given Sadeghi-Barzani et al. (2014)
demand [60]
p-median Location and sizing (capacity) Minimize the infrastructure cost for a given Zhu et al. (2017) [53]
demand
SCLM Location Maximize the number of EVs charged Wang & Lin (2009) [65]
SCLM Location Maximize the number of EVs charged Wang & Wang (2010) [43]
SCLM Location and sizing (capacity) Maximize the amount of energy charged Csiszár et al. (2019) [48]
SCLM Location and sizing (capacity) Minimize the infrastructure cost for a given Andrenacci et al. (2016) [77]
demand
SCLM Location and sizing (capacity) Minimize the infrastructure cost for a given Ghamami et al. (2016) [78]
demand
SCLM Location and sizing (capacity) Minimize the infrastructure cost for a given Davidov & Pantoš (2017) [79]
demand
SCLM Location and sizing (capacity) Minimize the infrastructure cost for a given Vazifeh et al. (2019) [80]
demand
SCLM Location and sizing (power) Minimize the infrastructure cost for a given Li et al (2011) [81]
demand
Unclassified Location Maximize charger utilization Pevec et al. (2018) [38]
(node-based)
Unclassified Location Minimize the infrastructure cost for a given Rajabi-Ghahnavie &
(node-based) demand Sadeghi-Barzani (2017) [35]
Unclassified Location and sizing (capacity) Maximize the number of EVs charged He et al. (2016) [72]
(node-based)
Unclassified Location and sizing (power) Maximize the distance traveled Wang et al. (2019) [49]
(node-based)
a The sizing in capacity refers to number of EVs that can be served per unit of time.
b
The sizing in power refers to charging speed (higher charging power means higher charging speed.

in the lower level and aiming to maximize social welfare in the upper principle. This can be valuable in the case of data-greedy tour-based
level. Xi et al. [96] adopted a lower-resolution model, dividing a models. Multi-agent models make it possible to track each agent in
region into sub-regions for which the trip data between sub-regions is a studied population individually, and therefore carry out analyses in
available. Their aim was to maximize the number of EVs that charge, relation to the activities of that population, and provide explicit repre-
or the amount of battery charged, with a trade-off between level 1 sentations of tours [102]. Moreover, modeling tools like the MATSim
and 2 infrastructures under a budget constraint. They found that the project [103] have been developed to simulate populations’ behavior
efficiency of privileging level 1 or 2 infrastructure depends on the with regard to the transport system, and they can be used to model
objective chosen, but that level 1 chargers are more cost-efficient if energy demand [104].
sufficient funds are unavailable.
An overview of the tour-based literature is given in Table 4. The 4.4. Discussion
tour-based methods are not really categorized, so the ‘‘Method’’ column
does not appear contrary to the two previous tables. To sum up, Fig. 6 gives an overview of the methods previously
Tour-based models often require a lot of data, which is often dif- discussed.
ficult to access for privacy reasons. Agent-based models – or multi The main advantage of the node-based approach is that it needs
agent models – can informatively simulate data and analyze traffic little data, only requiring population density, which is relatively ac-
dynamics [49]. Chen et al. [101] used an agent-based model with cessible. This makes the node-based approach an easy first estimate
autonomous EVs to place charging stations. This kind of model can be of charging station locations. However, there are limits to this type
built from real travel data (travel surveys, etc.) and be used to compare of coverage. For instance, the uncapacitated models only deal with
users’ behaviors among different charging infrastructure deployment coverage without considering the amount of demand. Second, this
strategies. Agent-based models can also be built to scenarios for study, resolution pathway offers a static vision of the charging demand,
which can be useful if there is insufficient data to validate a model which is not the case in reality: as previously stated, one of the main

8
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

Table 3
Summary of articles using the path-based approach.
Method Problem Main optimization goal Paper
FCLM Location Maximize number of EVs charged He et al. (2018) [87]
FCLM Location Maximize number of EVs charged Motoaki, Y (2019) [44]
FCLM Location Maximize number of EVs charged Riemann et al. (2015) [88]
FCLM Location Maximize number of EVs charged Wu & Sioshansi (2017) [86]
FCLM Location Minimize the infrastructure cost for a given Li et al (2016) [25]
demand
FCLM Location Minimize waiting time at the station Hanabusa & Horiguchi (2011)
[39]
FCLM Location and sizing Minimize failed trips (or maximize number of Micari et al. (2017) [51]
(capacity) possible trips)
FCLM Location and sizing Minimize the infrastructure cost for a given Dong et al. (2016) [33]
(capacity) demand
FCLM Location and sizing Minimize the infrastructure cost for a given Xiang et al. (2016) [89]
(capacity) demand
FRLM Location Maximize number of EVs charged Kuby et al. (2005) [82]
FRLM Location Maximize number of EVs charged Kuby et al. (2007) [83]
FRLM Location Minimize the infrastructure cost for a given Huang et al. (2015) [85]
demand
FRLM Location Minimize the infrastructure cost for a given Li & Huang (2014) [11]
demand
FRLM Location and sizing Maximize number of EVs charged Upchurch et al. (2009) [54]
(capacity)
FRLM Location and sizing Maximize number of EVs charged Zhang et al. (2018) [59]
(capacity)
Hybrid approach: node and path-based Location Minimize failed trips (or maximize number of Upchurch & Kuby (2010) [90]
possible trips)
Hybrid approach: path-based (fast charging) Location Minimize the infrastructure cost for a given Huang et al. (2016) [91]
and node-based (slow charging) demand
Hybrid approach: path-based (fast charging) Location and sizing Maximize number of EVs charged Sun et al. (2018) [45]
and node-based (slow charging) (power)

Table 4
Summary of articles using the tour-based approach.
Problem Main optimization goal Paper
Location Maximize distance traveled Shahraki et al. (2015) [94]
Location Maximize number of EVs charged He et al. (2015) [52]
Location Minimize the distance (or deviation) to a charging station Andrew et al. (2013) [93]
Location Minimize failed trips (or maximize number of possible trips) Asamer et al. (2016) [50]
Location Minimize the infrastructure cost for a given demand Wang et al. (2017) [32]
Location Minimize waiting time at the station Jung et al. (2014) [71]
Location Minimize waiting time at the station Tu et al. (2016) [40]
Location and sizing (capacity) Maximize charger utilization Cai et al. (2014) [37]
Location and sizing (capacity) Maximize number of EVs charged Cavadas et al. (2015) [92]
Location and sizing (capacity) Minimize failed trips (or maximize the number of possible trips) Dong et al. (2012) [97]
Location and sizing (capacity) Minimize the infrastructure cost for a given demand Han et al. (2016) [98]
Location and sizing (capacity) Minimize the infrastructure cost for a given demand Jia et al. (2012) [36]
Location and sizing (power and capacity) Maximize the amount of energy charged Chen et al. (2013) [47]
Location and sizing (power and capacity) Maximize number of EVs charged or maximize the amount of energy charged Xi et al. (2013) [99]
Location and sizing (power) Minimize failed trips (or maximize number of possible trips) You & Hsieh (2014) [56]
Location and sizing (power) Minimize waiting time at the station Kameda & Mukai (2011) [100]

advantages of a flow-based model over a nodal approach is that it for slow charging stations where EV batteries can take several hours
can take into account issues that only emerge from the description of to charge. Thus, the flow-based approach is not a substitute for the
vehicle flows. Another issue is that node-based coverage can lead to node-based approach, but complementary to it, depending on objective,
a poor representation of charging needs. According to Hodgson [16], territory, type of charging stations, etc. However, many studies only
the demand in a network is not always expressed at nodes, as people use one or the other category. Sun et al. [45] used a mixed-method ap-
generally will not make a trip from home to the charging station just proach, with location of fast charging stations for vehicle interception
to charge their vehicle. Furthermore, a node-based approach fails to and a node-based approach to place slow charging stations in places
deal with issues emerging from traffic flows such as cannibalization, where a long charging time is acceptable. However, flow-capturing
meaning that charging stations cut into each other’s coverage areas. models often fail to capture the uncertainty of EV charging demand,
In addition, Upchurch et al. [90] found that the flow-based method is which can lead to less robust locations [86].
more stable as the number of charging stations to place increases, which Given the issues with the flow-based approach, the tour-based ap-
is really important when planning over time. That is why many studies proach is based not only on user driving patterns but more generally
explicitly integrate the effect of flows into the location of charging on user behaviors. This type of approach is sometimes also referred
stations [25]. to as ‘activity-based’. By considering events around the details of the
However, this flow-based approach is not suitable for all cases. sequence of trips, it allows a better representation of drivers’ charging
Flow-based methods consider that EV charging will be done quickly needs than the two previous approaches. By using real and individual
before continuing the trip to the primary destination, just as any ICE data, the tour-based approach captures the randomness in the behavior
vehicle user would do. While this solution is possible with fast charging of users, and allows to serve all users, which cannot be done with
stations, which can refuel an EV in a dozen minutes, it is not possible aggregated data, as illustrated in Fig. 7. In this case, both green and

9
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

Fig. 6. Overview of location methods.

criteria chosen are the right ones. The high cost of the infrastructure
makes large-scale testing unfeasible. To overcome this problem, multi-
agent models can help, as explained above. However, these models may
be subject to simulation bias, and may therefore give an erroneous view
of user behavior.

5. Conclusion

This paper analyzed models for deploying charging infrastructure

and discussed the allied technical, economic, and user behavior-related
issues.
The wide diffusion of EVs is a step towards greener mobility, which
is one of today’s big challenges. This transition from ICE vehicles
Fig. 7. Example of three paths.
to EVs cannot take place without infrastructure that greatly reduces
early users’ range anxiety and reassures potential future users that
Table 5
EVs are capable of providing the same services as ICE vehicles. For
Main points of comparison between location methods.
the time being, infrastructures have been developed with a limited
Criteria Method
real coherent overarching strategy. However, the underlying costs of
Node-based Path-based Tour-based
necessary infrastructure to meet the needs of a large number of EV
Urban territory + −− ++ users, as well as the physical limitations of the electricity grid, make it
Highways − ++ + imperative to coordinate and optimize the large-scale deployment of an
Representation of charging needs +
− /+ + ++ electric charging infrastructure, failing which there is a risk of wasting
User behavior − +
− /+ ++ valuable resources and of ending up with an infrastructure that is not
adapted to user needs.
Data requirements Very low Low Very high
The scholarship has used several approaches for optimizing the de-
ployment of charging infrastructure. These approaches can be collapsed
into three categories: node-based, path-based, and tour-based. Although
yellow paths pass through nodes 1 and 2, and the red path passes not specific to EV charging infrastructure planning, these approaches
through node 3. If two stations were placed based on aggregated data, can readily adapt to consider the specificities of EVs instead of copying
they would be at nodes 1 and 2 that have the most traffic passing the gas station model, and facilitate the transition from ICE vehicles to
through, but the green and yellow vehicles would be served twice and EVs easier by minimizing the constraints of using EVs.
the red one would not be served at all, which could have been avoided The node-based approach is easy to implement and suitable for
if using individual data. certain areas such as residential neighborhoods, but it fails to capture
However, as noted in most of the tour-based works, this method is the problems arising from vehicle flows. The path-based approach can
often data-driven, with real or at least simulated data. It requires a large address this gap, but it is better suited for highway use-cases and has
amount of highly detailed data, drilling down to at least the detail of the downside of leading to time-consuming infrastructure, which may
individual trips and stops for a sufficiently large sample size to make prove a barrier for users to make the transition from ICE vehicles to
the model realistic. This data can be hard to obtain. The mains points EVs. The tour-based approach requires a lot of data and is therefore
of comparison between approaches are summarized in Table 5 more difficult to implement, but it is able to consider user activities
To conclude this section, note that many studies have been con- in order to get the best-adapted and least-restrictive infrastructure
ducted for the purpose of planning the best possible charging infrastruc- possible for users. With data on the activities of users, points of interest
ture. They have been carried out with different criteria to be optimized can be exploited to provide charging solutions at locations where there
according to the desired objective. However, while it is easy to check is demand, without users having to change their behavior [105].
whether chosen criteria have been optimized, it is harder to measure The methods adopt different response strategies, regardless of the
the impact of this model on the population, in other words whether the approach used. Some focus on maximizing served demand for a fixed

10
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

Table A.1

Paper Approach Problem Limited Charging Main Use case Target Home Electric grid Previous
resources type optimization environment charging aspect existing
goal available infrastructure
Wang & Lin Mixed Location and Yes Mixed Minimize failed Private Mixed Unspecified No No
(2013) [57] approach sizing (power (constraint) trips (or vehicle environment
and capacity) maximize the
number of
possible trips)
Upchurch & Node and Location Yes Unique Minimize failed Private Highway Unspecified No No
Kuby (2010) path-based (constraint) trips (or vehicle
[90] maximize the
number of
possible trips)
He et al. (2016) Node-based Location and Yes Mixed Maximize the Private Urban Yes, not for No No
[72] sizing (capacity) (constraint) number of EVs vehicle environment all
charged
Rajabi- Node-based Location No (budget Fast charging Minimize the Private Urban Yes, for all Yes, as a No
Ghahnavieh & to optimize) infrastructure vehicle environment parameter
Sadeghi-Barzani cost for a given
(2017) [35] demand
Pevec et al. Node-based Location Yes Unique Maximize the Private Mixed Yes, not for No Yes
(2018) [38] (constraint) charger’s vehicle environment all
utilization
Wang et al. Node-based Location and Yes Mixed Maximize the Taxis and Urban Yes, not for No No
(2019) [49] sizing (power) (constraint) distance traveled private environment all
vehicles
Frade et al. Node-based Location Yes Slow Maximize the Private Urban Yes, not for No No
(2011) [67] (MCLM) (constraint) charging number of EVs vehicle environment all
charged
Liu, J. (2012) Node-based Location and Yes Mixed Maximize the Private Urban Yes, not for Yes, as a No
[75] (MCLM) sizing (power) (constraint) number of EVs vehicle environment all constraint
charged
Wagner et al. Node-based Location and Yes Mixed Maximize the Private Semi-urban Unspecified No No
(2013) [68] (MCLM) sizing (power) (constraint) amount of vehicle environment
energy charged
Wang et al. Node-based Location and Yes Mixed Maximize the Private Unspecified Unspecified Yes, as a No
(2013) [57] (MCLM) sizing (capacity) (constraint) number of EVs vehicle constraint
charged
Guo & Zhao Node-based Location Yes Unique Maximize the Private Urban Unspecified Yes, as a No
(2015) [61] (MCLM) (constraint) number of EVs vehicle environment parameter
charged
Gopalakrishnan Node-based Location and Yes Unique Maximize the Private Unspecified Unspecified No No
et al. (2016) (MCLM) sizing (capacity) (constraint) number of EVs vehicle
[74] charged
Yang et al. Node-based Location and Yes Unique Minimize the Taxi fleet Semi-urban Unspecified No Yes
(2017) [46] (MCLM) sizing (power (constraint) infrastructure environment
and capacity) cost for a given
demand
Deb et al. Node-based Location and No (budget Mixed Minimize the Private Urban Unspecified Yes, as a No
(2019) [76] (MCLM) sizing (power) to optimize) infrastructure vehicle environment parameter
cost for a given
demand
Ge et al. (2011) Node-based Location and Yes Unique Minimize the Private Mixed Unspecified No No
[41] (p-median) sizing (capacity) (constraint) distance (or the vehicle environment
deviation) to a
charging station
Xu et al. (2013) Node-based Location Yes Unique Minimize the Private Urban Unspecified No No
[42] (p-median) (constraint) distance (or the vehicle environment
deviation) to a
charging station
Mehar et al. Node-based Location and No (budget Unique Minimize the Private Unspecified Unspecified Yes, as a Yes
(2013) [34] (p-median) sizing (capacity) to optimize) infrastructure vehicle constraint
cost for a given
demand
Gavranović et al. Node-based Location Yes Unique Minimize the Private Mixed Unspecified No No
(2014) [55] (p-median) (constraint) distance (or the vehicle environment
deviation) to a
charging station

(continued on next page)

11
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

Table A.1 (continued).

Paper Approach Problem Limited Charging Main Use case Target Home Electric grid Previous
resources type optimization environment charging aspect existing
goal available infrastructure
Jia et al. (2014) Node-based Location No (budget Fast charging Minimize the Private Semi-urban Yes, for all No No
[70] (p-median) to optimize) infrastructure vehicle with environment
cost for a given home
demand charging
Sadeghi-barzani Node-based Location and No (budget Fast charging Minimize the Private Unspecified Unspecified Yes, as a No
et al. (2014) (p-median) sizing (capacity) to optimize) infrastructure vehicle parameter
[60] cost for a given
demand
Zhu et al. Node-based Location and No (budget Unique Minimize the Private Unspecified Unspecified Yes, as a No
(2017) [53] (p-median) sizing (capacity) to optimize) infrastructure vehicle parameter
cost for a given
demand
Wang & Lin Node-based Location Yes Unique Maximize the Private Unspecified Unspecified No No
(2009) [65] (SCLM) (constraint) number of EVs vehicle
charged
Wang & Wang Node-based Location Yes Unique Maximize the Private Unspecified Unspecified No No
(2010) [43] (SCLM) (constraint) number of EVs vehicle
charged
Li et al. (2011) Node-based Location and No (budget Mixed Minimize the Private Unspecified Unspecified No No
[81] (SCLM) sizing (power) to optimize) infrastructure vehicle
cost for a given
demand
Andrenacci et al. Node-based Location and No (budget Unique Minimize the Private Semi-urban Yes, for all No No
(2016) [77] (SCLM) sizing (capacity) to optimize) infrastructure vehicle environment
cost for a given
demand
Ghamami et al. Node-based Location and No (budget Unique Minimize the Private Urban No No No
(2016) [78] (SCLM) sizing (capacity) to optimize) infrastructure vehicle environment
cost for a given
demand
Davidov & Node-based Location and No (budget Unique Minimize the Private Unspecified Unspecified No No
Pantoš (2017) (SCLM) sizing (capacity) to optimize) infrastructure vehicle
[79] cost for a given
demand
Csiszár et al. Node-based Location and Yes Unique Maximize the Private Unspecified Yes, not for No No
(2019) [48] (SCLM) sizing (capacity) (constraint) amount of vehicle all
energy charged
Vazifeh et al. Node-based Location and No (budget Unique Minimize the Private Semi-urban Yes, for all No No
(2019) [80] (SCLM) sizing (capacity) to optimize) infrastructure vehicle environment
cost for a given
demand
Huang et al. Path-based Location No (budget Mixed Minimize the Private Semi-urban Unspecified No No
(2016) [91] (fast to optimize) infrastructure vehicle environment
charging) and cost for a given
node-based demand
(slow
charging)
Sun et al. (2018) Path-based Location and Yes Mixed Maximize the Private Mixed Yes, for all No No
[45] (fast sizing (power) (constraint) number of EVs vehicle environment
charging) and charged
node-based
(slow
charging)
Hanabusa & Path-based Location Yes Unique Minimize the Private Unspecified Unspecified Yes, as a No
Horiguchi (FCLM) (constraint) waiting time at vehicle constraint
(2011) [39] the station
Riemann et al. Path-based Location Yes Unique Maximize the Private Mixed Yes, for all No No
(2015) [88] (FCLM) (constraint) number of EVs vehicle environment
charged
Dong et al. Path-based Location and No (budget Fast charging Minimize the Private Highway Unspecified No No
(2016) [33] (FCLM) sizing (capacity) to optimize) infrastructure vehicle
cost for a given
demand
Li et al. (2016) Path-based Location No (budget Unique Minimize the Private Mixed Yes, for all No Yes
[25] (FCLM) to optimize) infrastructure vehicle environment
cost for a given
demand

(continued on next page)

12
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

Table A.1 (continued).

Paper Approach Problem Limited Charging Main Use case Target Home Electric grid Previous
resources type optimization environment charging aspect existing
goal available infrastructure
Xiang et al. Path-based Location and No (budget Unique Minimize the Private Unspecified Unspecified Yes, as a No
(2016) [89] (FCLM) sizing (capacity) to optimize) infrastructure vehicle constraint
cost for a given
demand
Wu & Sioshansi Path-based Location Yes Fast charging Maximize the Private Mixed Yes, for all No No
(2017) [86] (FCLM) (constraint) number of EVs vehicle environment
charged
Micari et al. Path-based Location and Yes Unique Minimize failed Private Highway Unspecified No No
(2017) [51] (FCLM) sizing (capacity) (constraint) trips (or vehicle
maximize the
number of
possible trips)
He et al. (2018) Path-based Location Yes Fast charging Maximize the Private Unspecified Yes, for all No No
[87] (FCLM) (constraint) number of EVs vehicle
charged
Motoaki, Y. Path-based Location Yes Fast charging Maximize the Private Unspecified Yes, for all Yes, as a Yes
(2019) [44] (FCLM) (constraint) number of EVs vehicle constraint
charged
Kuby et al. Path-based Location Yes Unique Maximize the Private Unspecified Unspecified No No
(2005) [82] (FRLM) (constraint) number of EVs vehicle
charged
Kuby et al. Path-based Location Yes Unique Maximize the Private Unspecified Unspecified No No
(2007) [83] (FRLM) (constraint) number of EVs vehicle
charged
Upchurch et al. Path-based Location and Yes Unique Maximize the Private Unspecified Unspecified No No
(2009) [54] (FRLM) sizing (capacity) (constraint) number of EVs vehicle
charged
Li & Huang Path-based Location No (budget Unique Minimize the Private Unspecified Yes, for all No No
(2014) [11] (FRLM) to optimize) infrastructure vehicle
cost for a given
demand
Huang et al. Path-based Location No (budget Unique Minimize the Private Unspecified Unspecified No No
(2015) [85] (FRLM) to optimize) infrastructure vehicle
cost for a given
demand
Zhang et al. Path-based Location and Yes Fast charging Maximize the Private Mixed Yes, for all Yes, as a No
(2018) [59] (FRLM) sizing (capacity) (constraint) number of EVs vehicle environment constraint
charged
Kameda & Mukai Tour-based Location and Yes Fast charging Minimize the Bus or public Urban No No No
(2011) [100] sizing (power) (constraint) waiting time at transport environment
the station
Dong et al. Tour-based Location and Yes Mixed Minimize failed Private Semi-urban Yes, for all No No
(2012) [97] sizing (capacity) (constraint) trips (or vehicle environment
maximize the
number of
possible trips)
Jia et al. (2012) Tour-based Location and No (budget Fast charging Minimize the Private Urban Unspecified No No
[36] sizing (capacity) to optimize) infrastructure vehicle environment
cost for a given
demand
Chen et al. Tour-based Location and Yes Mixed Maximize the Private Unspecified Yes, not for No No
(2013) [47] sizing (power (constraint) amount of vehicle all
and capacity) energy charged
Xi et al. (2013) Tour-based Location and Yes Mixed Maximize the Private Mixed Yes, for all No No
[99] sizing (power (constraint) (medium and number of EVs vehicle environment
and capacity) slow charged or
charging) maximize the
amount of
energy charged
Andrew et al. Tour-based Location Yes Medium Minimize the Private Semi-urban Yes, for all No No
(2013) [93] (constraint) charging distance (or the vehicle environment
deviation) to a
charging station
Cai et al. (2014) Tour-based Location and Yes Fast charging Maximize the Taxi fleet Semi-urban Yes, for all No No
[37] sizing (capacity) (constraint) charger’s environment
utilization

(continued on next page)

13
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

Table A.1 (continued).

Paper Approach Problem Limited Charging Main Use case Target Home Electric grid Previous
resources type optimization environment charging aspect existing
goal available infrastructure
You & Hsieh Tour-based Location and Yes Mixed Minimize failed Private Unspecified Unspecified No No
(2014) [56] sizing (power) (constraint) trips (or vehicle
maximize the
number of
possible trips)
Jung et al. Tour-based Location Yes Fast charging Minimize the Taxi fleet Urban Unspecified No No
(2014) [71] (constraint) waiting time at environment
the station
Shahraki et al. Tour-based Location Yes Unique Maximize the Taxi fleet Semi-urban Yes, for all No No
(2015) [94] (constraint) distance traveled environment
Cavadas et al. Tour-based Location and Yes Slow Maximize the Private Urban Yes, not for No No
(2015) [92] sizing (capacity) (constraint) charging number of EVs vehicle environment all
charged
He et al. (2015) Tour-based Location Yes Medium Maximize the Private Unspecified Yes, for all No No
[52] (constraint) charging number of EVs vehicle
charged
Asamer et al. Tour-based Location Yes Mixed Minimize failed Taxi fleet Urban No No No
(2016) [50] (constraint) (medium and trips (or environment
fast charging) maximize the
number of
possible trips)
Han et al. Tour-based Location and No (budget Fast charging Minimize the Taxi fleet Semi-urban Yes, for all No No
(2016) [98] sizing (capacity) to optimize) infrastructure environment
cost for a given
demand
Tu et al. (2016) Tour-based Location Yes Fast charging Minimize the Taxi fleet Semi-urban Unspecified No No
[40] (constraint) waiting time at environment
the station
Wang et al. Tour-based Location No (budget Fast charging Minimize the Bus or public Urban Unspecified No No
(2017) [32] to optimize) infrastructure transport environment
cost for a given
demand

budget, which can be expressed in terms of the number of vehicles to Declaration of competing interest
be charged, volume of energy to be charged, time saved or number of
feasible trips. Others consider charging demand as the primal condition The authors declare that they have no known competing finan-
and try to minimize the budget needed to satisfy it. While early work cial interests or personal relationships that could have appeared to
focused on the geographical placement of charging stations to meet influence the work reported in this paper.
charging demand, more recent models now also integrate the service
capacity of the stations, introducing station sizing into the results. Acknowledgments
Charging speed used is rarely considered: many models consider only
one type of charging station, thus defining only the number (and not This work was financially supported by Institut VEDECOM, a French
quality) of charging points needed. Public–Private research institute and one of the Institutes for the Energy
Few of the models other than node and parking-based models look Transition. The authors claim no conflict of interest. The authors thank
to take advantage of the benefits offered by EV charging, which does the Editor in Chief Dr Foley and three anonymous reviewers for several
not require the user to be present during charging time. This key ad- useful suggestions that helped to improve this paper. The authors also
vantage should be considered in order to plan a charging infrastructure thank Olivier Massol, Bassem Haidar, Icaro Freitas-Gomes and Felipe
that matches charging opportunities, to make EV use as unrestrictive as Gonzales for their helpful discussions.
possible and thus encourage EV diffusion.
To conclude, the optimization models reviewed do not consider Appendix. Literature table
any temporality in deployment: for a given budget, the infrastructure
is optimized as if all the stations were placed simultaneously. How- See Table A.1.
ever, this kind an infrastructure does not get deployed all in one go,
partly because of the development costs it would generate without a
guaranteed return on investment from a demand that will take a long References
time to come. Charging infrastructure deployment will take place over
[1] Cook John, Nuccitelli Dana, Green Sarah A, Richardson Mark, Winkler Bärbel,
a period that may last several years, and this factor should now be
Painting Rob, Way Robert, Jacobs Peter, Skuce Andrew. Quantifying the
explored in order to have an infrastructure that provides acceptable consensus on anthropogenic global warming in the scientific literature. Environ
coverage from the very beginning of its deployment, and not just once Res Lett 2013;8(2):024024.
the last charging points have been installed. An incremental ‘over-time’ [2] European Environment Agency. Greenhouse gas emissions from transport in
Europe. 2019.
deployment must therefore be considered, factoring in the early-stage
[3] Hawkins Troy R, Singh Bhawna, Majeau-Bettez Guillaume, Strømman An-
infrastructure already present in the territory, which very few models ders Hammer. Comparative environmental life cycle assessment of conventional
do (see Appendix), and the action of ‘adding one more station’. and electric vehicles. J Ind Ecol 2013;17(1):53–64.

14
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

[4] Berkeley Nigel, Jarvis David, Jones Andrew. Analysing the take up of battery [36] Jia Long, Hu Zechun, Song Yonghua, Luo Zhuowei. Optimal siting and sizing
electric vehicles: An investigation of barriers amongst drivers in the UK. Transp of electric vehicle charging stations. In: 2012 Ieee international electric vehicle
Res D 2018;63:466–81. conference, Vol. 36(3). 2012, p. 54–9.
[5] Yan Shiyu. The economic and environmental impacts of tax incentives [37] Cai Hua, Jia Xiaoping, Chiu Anthony SF, Hu Xiaojun, Xu Ming. Siting public
for battery electric vehicles in Europe. Energy Policy 2018;123(August electric vehicle charging stations in Beijing using big-data informed travel
2017):53–63. patterns of the taxi fleet. Transp Res D 2014;33:39–46.
[6] Soulopoulos Nikolas. When will electric vehicles be cheaper than conventional [38] Pevec Dario, Babic Jurica, Kayser Martin A, Carvalho Arthur, Ghiassi-
vehicles? Bloom New Energy Financ 2017. Farrokhfal Yashar, Podobnik Vedran. A data-driven statistical approach
[7] International Energy Agency. Global EV outlook 2019. 2019, p. 1–40. for extending electric vehicle charging infrastructure. Int J Energy Res
[8] France Stratégie. Panorama des politiques publiques en faveur des véhicules à 2018;42(9):3102–20.
très faibles émissions - Note de synthèse. 2018, p. 140. [39] Hanabusa Hisatomo, Horiguchi Ryota. A study of the analytical method for the
[9] Springel Katalin. Network externality and subsidy structure in two-sided location planning of charging stations for electric vehicles. In: Lecture notes in
markets : Evidence from electric vehicle incentives resources. 2017, p. 1–63. computer science (including subseries lecture notes in artificial intelligence and
[10] Zhou Yiyi, Li Shanjun. Technology adoption and critical mass: The case of the lecture notes in bioinformatics). Lnai(Part 3), vol. 6883, 2011, p. 596–605.
U.S. electric vehicle market. J Ind Econ 2018;66(2):423–80. [40] Tu Wei, Li Qingquan, Fang Zhixiang, Shaw Shih-lung, Zhou Baoding. Optimizing
[11] Li Shanjun, Tong Lang, Xing Jianwei, Zhou Yiyi. The market for elec- the locations of electric taxi charging stations : A spatial – temporal demand
tric vehicles: Indirect network effects and policy impacts. SSRN Electron J coverage approach. Transp Res C 2016;65(3688):172–89.
2014;4(1). [41] Ge Shaoyun, Feng Liang, Liu Hong. The planning of electric vehicle charging
[12] Ministère de l’environnement de l’énergie et de la mer. La loi de transition station based on grid partition method. In: 2011 international conference on
énergétique pour la croissance verte. 2016, p. 4. electrical and control engineering, icece 2011 - proceedings. IEEE; 2011, p.
[13] Nicholas Michael. Estimating electric vehicle charging infrastructure costs across 2726–30.
major U.S. metropolitan areas. ICCT 2019. [42] Xu Hao, Miao Shihong, Zhang Chunyong, Shi Dongyuan. Optimal placement of
[14] Melaina Marc, Bremson Joel. Refueling availability for alternative fuel vehicle charging infrastructures for large-scale integration of pure electric vehicles into
markets: Sufficient urban station coverage. Energy Policy 2008;36(8):3233–41. grid. Int J Electr Power Energy Syst 2013;53(1):159–65.
[15] Toregas Constantine, Swain Ralph, ReVelle Charles, Bergman Lawrence. The [43] Wang Ying Wei, Wang Chuan Ren. Locating passenger vehicle refueling stations.
location of emergency service facilities. Oper Res 1971;19(6):1363–73. Transp Res E 2010;46(5):791–801.
[16] Hodgson M John. A flow-capturing location-allocation model. Geogr Anal [44] Motoaki Yutaka. Location-allocation of electric vehicle fast chargers-research
1990;22(3):270–9. and practice. World Electr Veh J 2019;10(1):10–6.
[17] International Electrotechnical Commission. Norme européenne IEC 61851-1, 33, [45] Sun Zhuo, Gao Wei, Li Bin, Wang Longlong. Locating charging stations for
(0), 2014. electric vehicles. Transp Policy 2018;(May):1–7.
[18] Office of Energy Efficiency and Renewable Energy. Vehicle charging. 2020. [46] Yang Jie, Dong Jing, Hu Liang. A data-driven optimization-based approach
[19] Pearre Nathaniel S, Kempton Willett, Guensler Randall L, Elango Vetri V. for siting and sizing of electric taxi charging stations. Transp Res C
Electric vehicles: How much range is required for a day’s driving? Transp Res 2017;77(2):462–77.
C 2011;19(6):1171–84. [47] Chen T Donna, Kockelman Kara M, Khan Moby. The electric vehicle charging
[20] Armoogum Jimmy, Hubert Jean-paul, Roux Sophie, Le Jeannic Thomas. Enquête station location problem: A parking-based assignment method for seattle. In
nationale transports et déplacement 2008 - introduction. Rev - Commis Gén Dév Proceedings of the 92nd annual meeting of the transportation research board in
Durable – Serv Obs Statist 2010;(Décembre):5–24. washington dc, january 2013, and forthcoming in transportation research record,
[21] Pasaoglu Guzay, Zubaryeva Alyona, Fiorello Davide, Thiel Christian. Analysis of 2013, p. 22.
European mobility surveys and their potential to support studies on the impact [48] Csiszár Csaba, Csonka Bálint, Földes Dávid, Wirth Ervin, Lovas Tamás. Urban
of electric vehicles on energy and infrastructure needs in Europe. Technol public charging station locating method for electric vehicles based on land use
Forecast Soc Change 2014;87:41–50. approach. J Transp Geogr 2019;74(November 2018):173–80.
[22] Eurostat. Housing statistics. 2020. [49] Wang Hua, Zhao De, Meng Qiang, Ong Ghim Ping, Lee Der Horng. A four-step
[23] Delacrétaz Nathan, Lanz Bruno, Van Dijk Jeremy. The chicken or the egg: method for electric-vehicle charging facility deployment in a dense city: An
technology adoption and network infrastructure in the market for electric empirical study in Singapore. Transp Res A 2019;119(July 2018):224–37.
vehicles. Irene working papers, 20 (08), 2020. [50] Asamer Johannes, Reinthaler Martin, Ruthmair Mario, Straub Markus,
[24] Wesolowsky George O. Dynamic facility location. Manage Sci Puchinger Jakob. Optimizing charging station locations for urban taxi providers.
1973;19(11):1241–548. Transp Res A 2016;85:233–46.
[25] Li Shengyin, Huang Yongxi, Mason Scott J. A multi-period optimization model [51] Micari Salvatore, Polimeni Antonio, Napoli Giuseppe, Andaloro Laura, An-
for the deployment of public electric vehicle charging stations on network. tonucci Vincenzo. Electric vehicle charging infrastructure planning in a road
Transp Res C 2016;65:128–43. network. Renew Sustain Energy Rev 2017;80(May 2016):98–108.
[26] Nicholas Michael, Tal Gil, Davies Jamie, Woodjack Justin. DC fast as the only [52] He Fang, Yin Yafeng, Zhou Jing. Deploying public charging stations for electric
public charging option? Scenario testing from GPS tracked vehicles. Transp Res vehicles on urban road networks. Transp Res C 2015.
Board Conf 2012;1–20. [53] Zhu Jie, Li Yixin, Yang Jun, Li Xianglong, Zeng Shuang, Chen Yanxie.
[27] San Román Tomás Gómez, Momber Ilan, Abbad Michel Rivier, Sánchez Mi- Planning of electric vehicle charging station based on queuing theory. J Eng
ralles Álvaro. Regulatory framework and business models for charging plug-in 2017;2017(13):1867–71.
electric vehicles: Infrastructure, agents, and commercial relationships. Energy [54] Upchurch Christopher, Kuby Michael, Lim Seow. A model for location of
Policy 2011;39(10):6360–75. capacitated alternative-fuel stations. Geogr Anal 2009;41(1):127–48.
[28] Green Robert C, Wang Lingfeng, Alam Mansoor. The impact of plug-in hybrid [55] Gavranović Haris, Barut Alper, Ertek Gürdal, Yüzbaşioǧlu Orkun Berk,
electric vehicles on distribution networks: A review and outlook. In: Ieee Pes Pekpostalci Osman, Tombuş Önder. Optimizing the electric charge station
General Meeting, Pes 2010. 2010, p. 1–8. network of EŞARJ. Procedia Comput Sci 2014;31:15–21.
[29] Freitas Gomes Icaro Silvestre, Perez Yannick, Suomalainen Emilia. Coupling [56] You Peng Sheng, Hsieh Yi Chih. A hybrid heuristic approach to the problem of
small batteries and PV generation: A review. Renew Sustain Energy Rev the location of vehicle charging stations. Comput Ind Eng 2014.
2020;126(March):109835. [57] Wang Ying Wei, Lin Chuah Chih. Locating multiple types of recharging stations
[30] Thompson Andrew W, Perez Yannick. Vehicle-to-everything (V2X) energy for battery-powered electric vehicle transport. Transp Res E 2013;58:76–87.
services , value streams , and regulatory policy implications. Energy Policy [58] Dharmakeerthi CH, Mithulananthan N, Saha TK. Impact of electric vehicle fast
2020;137(July 2019):111136. charging on power system voltage stability. Int J Electr Power Energy Syst
[31] Sehimi Yacine, Suomalainen Emilia, Almaksour K, Robyns B. Mitigating the 2014;57:241–9.
impact of fast charging on distribution grids using vehicle-to-vehicle power [59] Zhang Hongcai, Moura Scott J, Hu Zechun, Song Yonghua. PEV Fast-charging
transfer : a Paris case study. In: 4th e-mobility power integration symposium. station siting and sizing on coupled transportation and power networks. IEEE
2020. Trans Smart Grid 2018;9(4):2595–605.
[32] Wang Xiumin, Yuen Chau, Hassan Naveed Ul, An Ning, Wu Weiwei. Electric [60] Sadeghi-barzani Payam, Rajabi-ghahnavieh Abbas, Kazemi-karegar Hosein.
vehicle charging station placement for urban public bus systems. IEEE Trans Optimal fast charging station placing and sizing. Appl Energy 2014;125:289–99.
Intell Transp Syst 2017;18(1):128–39. [61] Guo Sen, Zhao Huiru. Optimal site selection of electric vehicle charging
[33] Dong Xiaohong, Mu Yunfei, Jia Hongjie, Wu Jianzhong, Yu Xiaodan. Planning station by using fuzzy TOPSIS based on sustainability perspective. Appl Energy
of fast EV charging stations on a round freeway. IEEE Trans Sustain Energy 2015;158:390–402.
2016;7(4):1452–61. [62] Shareef Hussain, Islam Md Mainul, Mohamed Azah. A review of the stage-of-
[34] Mehar Sara, Mehar Sara, Senouci Sidi Mohammed. An optimization location the-art charging technologies, placement methodologies, and impacts of electric
scheme for electric charging stations. 2013, (May). vehicles. Renew Sustain Energy Rev 2016;64:403–20.
[35] Rajabi-Ghahnavieh Abbas, Sadeghi-Barzani Payam. Optimal zonal fast-charging [63] Deb Sanchari, Tammi Kari, Kalita Karuna, Mahanta Pinakeswar. Review of
station placement considering urban traffic circulation. IEEE Trans Veh Technol recent trends in charging infrastructure planning for electric vehicles. Wiley
2017;66(1):45–56. Interdiscip Rev Energy Environ 2018;7(6):1–26.

15
M.O. Metais et al. Renewable and Sustainable Energy Reviews 153 (2022) 111719

[64] Owen Susan Hesse, Daskin Mark S. Strategic facility location: A review. [85] Huang Yongxi, Li Shengyin, Qian Zhen Sean. Optimal deployment of alternative
European J Oper Res 1998;111(3):423–47. fueling stations on transportation networks considering deviation paths. Netw
[65] Wang Ying Wei, Lin Chuah Chih. Locating road-vehicle refueling stations. Spat Econ 2015;15(1):183–204.
Transp Res E 2009;45(5):821–9. [86] Wu Fei, Sioshansi Ramteen. A stochastic flow-capturing model to optimize the
[66] Church Richard L, Revelle Charles. The maximal covering location problem. location of fast-charging stations with uncertain electric vehicle flows. Transp
Pap Reg Sci Assoc 1974;32:101–18. Res D 2017;53:354–76.
[67] Frade Inês, Ribeiro Anabela, Gonçalves Gonçalo, Antunes António. Optimal [87] He Jia, Yang Hai, Tang Tie Qiao, Huang Hai Jun. An optimal charging station
location of charging stations for electric vehicles in a neighborhood in Lisbon, location model with the consideration of electric vehicle’s driving range. Transp
Portugal. Transp Res Rec 2011;(2252):91–8. Res C 2018;86(December 2016):641–54.
[68] Wagner Sebastian, Götzinger Markus, Neumann Dirk. Optimal location of [88] Riemann Raffaela, Wang David ZW, Busch Fritz. Optimal location of wireless
charging stations in smart cities: A point of interest based approach. In: charging facilities for electric vehicles: Flow capturing location model with
International conference on information systems (icis 2013): reshaping society stochastic user equilibrium. Transp Res C 2015;58(Part A):1–12.
through information systems design, Vol. 3. 2013, p. 2838–55. [89] Xiang Yue, Liu Junyong, Li Ran, Li Furong, Gu Chenghong, Tang Shuoya.
[69] Hakimi SL. Optimum locations of switching centers and the absolute centers Economic planning of electric vehicle charging stations considering traffic
and medians of a graph. Oper Res 1964;12(3):450–9. constraints and load profile templates. Appl Energy 2016;178:647–59.
[70] Jia Long, Hu Zechun, Liang Wenju, Lang Wenzuo, Song Yonghua. A novel [90] Upchurch Christopher, Kuby Michael. Comparing the p-median and flow-
approach for urban electric vehicle charging facility planning considering com- refueling models for locating alternative-fuel stations. J Transp Geogr
bination of slow and fast charging. In: POWERCON 2014 - 2014 international 2010;18(6):750–8.
conference on power system technology: towards green, efficient and smart [91] Huang Kai, Kanaroglou Pavlos, Zhang Xiaozhou. The design of electric vehicle
power system, proceedings, (powercon). 2014, p. 3354–60. charging network. Transp Res D 2016;49:1–17.
[71] Jung Jaeyoung, Chow Joseph Y J, Jayakrishnan R, Young Ji. Stochastic dynamic [92] Cavadas Joana, de Almeida Correia Gonçalo Homem, Gouveia João. A MIP
itinerary interception refueling location problem with queue delay for electric model for locating slow-charging stations for electric vehicles in urban areas
taxi charging stations. Transp Res C 2014;40:123–42. accounting for driver tours. Transp Res E 2015.
[72] He Sylvia Y, Kuo Yong Hong, Wu Dan. Incorporating institutional and spatial [93] Andrew Matthew, Dogru Mustafa K, Hobby John D, Jin Yue, Tucci Gabriel H.
factors in the selection of the optimal locations of public electric vehicle Modeling and optimization for electric vehicle charging infrastructure. 2013.
charging facilities: A case study of Beijing, China. Transp Res C 2016. [94] Shahraki Narges, Cai Hua, Turkay Metin, Xu Ming. Optimal locations of electric
[73] An Yu, Zeng Bo, Zhang Yu, Zhao Long. Reliable p-median facility location public charging stations using real world vehicle travel patterns. Transp Res D
problem: Two-stage robust models and algorithms. Transp Res B 2014;64:54–72. 2015;41(2015):165–76.
[74] Gopalakrishnan Ragavendran, Biswas Arpita, Lightwala Alefiya, Vasude- [95] González Jairo, Alvaro Roberto, Gamallo Carlos, Fuentes Manuel, Fraile-
van Skanda, Dutta Partha, Tripathi Abhishek. Demand prediction and placement Ardanuy Jesús, Knapen Luk, Janssens Davy. Determining electric vehicle
optimization for electric vehicle charging stations. In: Ijcai international joint charging point locations considering drivers’ daily activities. Procedia Comput
conference on artificial intelligence, Vol. 2016-January. 2016, p. 3117–23. Sci 2014;32:647–54.
[75] Liu Jian. Electric vehicle charging infrastructure assignment and power grid [96] Xi Xiaomin, Sioshansi Ramteen, Marano Vincenzo. Simulation-optimization
impacts assessment in Beijing. Energy Policy 2012;51:544–57. model for location of a public electric vehicle charging infrastructure. Transp
[76] Deb Sanchari, Tammi Kari, Kalita Karuna, Mahanta Pinakeswar. Charging Res D 2013;22:60–9.
station placement for electric vehicles: A case study of Guwahati City, India. [97] Dong Jing, Liu Changzheng, Lin Zhenhong. Charging infrastructure plan-
IEEE Access 2019;7:100270–82. ning for promoting all-electric vehicle market: An activity-based assess-
[77] Andrenacci N, Ragona R, Valenti G. A demand-side approach to the optimal ment using multiday travel data jing dong (corresponding author). Phone
deployment of electric vehicle charging stations in metropolitan areas. Appl 2012;(865):946–1308.
Energy 2016;182:39–46. [98] Han Daehee, Ahn Yongjun, Park Sunkyu, Yeo Hwasoo. Trajectory-interception
[78] Ghamami Mehrnaz, Nie Yu (Marco), Zockaie Ali. Planning charging infras- based method for electric vehicle taxi charging station problem with real taxi
tructure for plug-in electric vehicles in city centers. Int J Sustain Transp data. Int J Sustain Transp 2016;10(8):671–82.
2016;10(4):343–53. [99] Xi Xiaomin, Sioshansi Ramteen, Marano Vincenzo. Simulation – optimization
[79] Davidov Sreten, Pantoš Miloš. Planning of electric vehicle infrastructure based model for location of a public electric vehicle charging infrastructure. Transp
on charging reliability and quality of service. Energy 2017;118:1156–67. Res D 2013;22:60–9.
[80] Vazifeh Mohammad M, Zhang Hongmou, Santi Paolo, Ratti Carlo. Optimizing [100] Kameda Hisashi, Mukai Naoto. Optimization of charging station placement by
the deployment of electric vehicle charging stations using pervasive mobility using taxi probe data for on-demand. 2011, p. 606–15.
data. Transp Res A 2019. [101] Chen T Donna, Kockelman Kara M, Hanna Josiah P. Operations of a
[81] Li Yanqing, Li Ling, Yong Jing, Yao Yuhai, Li Zhiwei. Layout planning of shared, autonomous, electric vehicle fleet: Implications of vehicle & charging
electrical vehicle charging stations. 2011. infrastructure decisions. Transp Res A 2016;94:243–54.
[82] Kuby Michael, Lim Seow. The flow-refueling location problem for alternative- [102] Hunt JD, Stefan KJ. Tour-based microsimulation of urban commercial
fuel vehicles. Soc-Econ Plan Sci 2005;39(2):125–45. http://dx.doi.org/10.1016/ movements. Transp Res B 2007;41(9):981–1013.
j.seps.2004.03.001. [103] Axhausen Kay W, Horni Andreas, Nagel Kai. The multi-agent transport
[83] Kuby Michael, Lim Seow. Location of alternative-fuel stations using the flow- simulation MATSim. 2016.
refueling location model and dispersion of candidate sites on arcs. Netw Spat [104] Novosel T, Perković L, Ban M, Keko H, Pukšec T, Krajačić G, Duić N. Agent
Econ 2007;7(2):129–52. based modelling and energy planning - utilization of MATSim for transport
[84] Kim Jong Geun, Kuby Michael. The deviation-flow refueling location model energy demand modelling. Energy 2015;92:466–75.
for optimizing a network of refueling stations. Int J Hydrogen Energy [105] Wolbertus Rick, Van den Hoed Robert. Electric vehicle fast charging needs in
2012;37(6):5406–20. cities and along corridors. World Electr Veh J 2019;10(2):45.

16