Optimal location of EV charging stations in a neighborhood considering a multi-objective approach

Highlights

• Presents a methodology to locate semi-fast charging stations in distribution system.

• Includes constraints of CS service zones delimited by a hierarchical clustering.

• Considers the distribution system and the EV's demand in a multi-objective approach.

• Evaluate CS location for mid and long-term planning based on the CS occurrence rate.

• Presents a real case distribution network including both MV and LV networks.

Abstract

Despite the environmental and economic benefits of Electric Vehicles (EVs), distribution network operators will need to understand the location where the charging infrastructure will be placed to ensure EV users’ needs are met. In this sense, this work proposes a methodology to define the optimal location of EV semi-fast charging stations (CS) at a neighborhood level, through a multi-objective approach. It applies a hierarchical clustering method to define CS service zones, considering both technical and mobility aspects. Besides, it considers uncertainties related to the EV load profile to determine the CS capacity, based on the user's charging behavior. A Pareto Frontier method is deployed to support the decision-making process on CS optimal location, considering utility and EV users’ preferences. The results indicate that the best CS locations for mid-term EV penetration can also fit into long-term planning, with higher EV charging demand. Thus, these locations would be good candidates for the power utility to make initial investments, regarding both planning horizons. A real distribution system case is used to demonstrate the applicability of the results.

Introduction

The electric vehicles (EVs) sales have increased over the years. While this brings multiple environmental and economic benefits, their deployment requires comprehensive studies to optimally plan the location of charging stations (CS) that ensures the EV charging demand will be met. As EV charging duration is still high [1], planning the location of semi-fast CSs to serve a substantial amount of EV users has become a major issue [2]. Therefore, CSs must be installed in strategic locations, where people often spend sufficient time, such as shopping malls, supermarkets and office buildings [3].

The CS location problem has been widely addressed in the literature through the delimitation of CS service zones [4]. The zones can be arranged using different clustering methods, usually divided between hierarchical and non-hierarchical algorithms. Non-hierarchical methods, such as the k-means and fuzzy c-means method, divide the database into distinct groups, while hierarchical methods categorize data in a tree with a hierarchical structure called dendrogram [5]. On the other hand, the utility operator should consider the EV's flexibility to make optimal decisions. This flexibility is affected by the driving behavior and vehicle parking patterns that are inherently stochastic [6]. These uncertainties related to the EV nature could be properly modeled through stochastic simulation. However, a comprehensive simulation can lead to impractical computation time due to the high number of scenarios. Therefore, scenario reduction techniques such as backward probability distance algorithm [7] and clustering methods [8] can be applied to lower the computational burden while preserving the stochastic information from the original scenarios.

Public charging infrastructure can encourage EV users to drive longer distances, as the availability of public CSs reduces range anxiety [9]. Locating public CSs in a neighborhood can also reduce the EV total cost of ownership and increase the EV penetration, since some users do not need to purchase a home charger.

This paper deals with an up-to-date problem that will become even more relevant as EVs sales increases. In mid-term planning, EV penetration is expected to be low, as consumers are starting to buy the region's first EVs. Although, it is expected to increase in long-term planning, as the technology becomes more accessible and widespread among the population. According to the Costa Rica Decarbonization Plan, mid- and long-term planning refers to 2023–2035 and 2035–2050, respectively [10]. Developing countries are expected to experience a similar growth in EV sales, as the price is still a major barrier. So, it is expected that this planning horizon would be similar in other countries.

The main approaches in the literature in this topic consider aspects related to economics, impacts on distribution power systems and EV mobility [11]. However, the problem is often formulated to meet the objectives of the CS operator only. For the power utilities perspective, optimal solutions should consider locations that minimize grid reinforcements cost, while maintaining the grid operation at a standard level [[12], [13], [14]]. For a private agent operator, capturing more clients to maximize the CS profit may be the priority [15, 16].

Although most of the problems associated with CS planning may include different aspects, they are often approached in a single objective model. However, in practice problems in this field of knowledge are better evaluated considering the existing conflicting objectives. For example, authors in [17] minimize the total cost (construction and charging) and maximize coverage. Another key factor for planning CSs is the user's behavior. The uncertainties associated with the time of arrival, departure, and state of charge of the batteries directly affect the size of the CS [18, 19].

Other topics often evaluated in CS planning include multi type chargers [20], integration with the transport network [21] and Vehicle-to-grid (V2G) [22]. A relevant subject that is rarely discussed in this area is the influence of EV penetration in CS position, as addressed in [23]. However, the CS planning problem in [23] is formulated as a mono-objective (maximize power utility profit) and does not consider the uncertainties in EV user's behavior. Thus, to the best of our knowledge, none of the existing works considers CS siting optimization in long term-planning with a multi-objective approach and EV user's uncertainties.

The present work aims to evaluate the impacts caused by the location of semi-fast CSs in a real case example in Costa Rica. A multi-objective approach comprising a metaheuristic algorithm is performed as a possible decision-making tool for the location of semi-fast CSs. The CSs are located in the distribution transformers, which are selected using the Bat Algorithm. The optimization problem considers the constraints of CS service zones, which are delimited by a hierarchical clustering method. The proposed objectives encompass both technical and mobility aspects. The technical aspect includes power losses minimization to reduce costs. The mobility aspect includes consumer preferences represented in this work as the distance that EV users have to reach the closest CS (shorter distances are expected to better meet user's needs).

In specific, this work contributes to the literature in:

• Performing an analysis that considers both the distribution power system and the EV's charging demand in a multi-objective approach.

• Evaluating the difference in CS location for mid and long-term planning based on the CS occurrence rate, i.e., the number of times a transformer is selected to allocate a CS.

• Consider uncertainties related to EV user's behavior.

• Presenting a holistic approach where decision-makers can balance between EV users and power utilities conflicting objectives.

The paper is organized as follows: Section II presents a literature review, with typical methods used to create zones, as well as the most used objective functions in CS location problems. Section III presents the methodology as well as the zone construction using a hierarchical clustering method of the zone and the formulation of the problem. Section IV presents the case study, with information about the planning horizons and the distribution network. In section V the results are discussed and the conclusion is developed in section VI.