Elektrikli Araç Sarj İstasyonlarının Verimli Konumlandırılması için Kümeleme ve Matematiksel Optimizasyon Yaklaşımı ile Yer Tespiti, Clustering and Mathematical Optimization Approaches for Efficient Estimation of Electric Vehicles Charging Stations' Locations

• Main Authors: Ekmekci, Yilmaz Can, Demirors, Dilan, Rassad, Nimet Aylin, Polat, Zeynep Ayca, Akkaya, Efe Berke, Bayturk, Enes, Pehlivan, Merve
• Format: Conference Proceeding
• Language: English
• Subjects: Analytical models; Charging Station Positioning; Charging stations; Classification algorithms; Climate change; Clustering algorithms; Data analysis; Electric Vehicle; Electric vehicle charging; Estimation; K-means Clustering; Machine Learning; Mathematical Modeling; Mathematical models; Optimization; Prediction algorithms; Prototypes; Surveys
• ISSN: 2770-7946
• DOI: 10.1109/ASYU62119.2024.10756982

The increasing effects of global warming have led to a shift to more environmentally friendly fuels. As electric vehicles become more popular in Türkiye, the demand for charging stations has also increased. However, charging stations are not able to meet demand, hence there is no strategically located charging network. In this study, a prototype for the optimal placement of electric vehicle charging stations is developed using analytical and mathematical approaches such as clustering and mathematical modeling, and Kocaeli province of Türkiye is selected as the prototype city. A preliminary survey was designed to better understand the needs and preferences of electric vehicle users. Supported by an extensive literature review, this research collected critical data on the most important criteria for the construction of EV charging stations and created a dataset by applying a systematic and iterative selection process. Various clustering methods were applied to this dataset and the KMeans algorithm achieved the highest score. With the K-Means algorithm, the data were divided into three clusters and classified as good, medium and poor according to the survey results and distribution. Using the developed clustering model, predictions were made for 50 coordinates where charging stations are planned to be installed. The 22 coordinates that were rated as good and medium by the estimation were selected for further mathematical analysis. The mathematical model with the most critical constraints aimed to maximize the number of users. The solution consists of three phases, with each phase allowing only one installation per region. At each stage, locations from previous stages were removed from the model and rerun with updated utilization scores. With the mathematical model, the most suitable charging station locations were determined within 22 coordinates.