Data-Driven Classification of Public Electric Vehicle Charging Station Power Using K-Means Clustering: A Case Study in Jakarta, Indonesia

Abstract

The rapid growth of electric vehicles requires public charging infrastructure that is not only widely available but also properly classified according to its service capability. In Jakarta, Indonesia, the rated power of public electric vehicle charging stations varies widely from 7 kW to 480 kW, indicating diverse charging service levels. However, the existing fixed-threshold classification may not fully represent the actual distribution of charger power, particularly in the high-power segment where all chargers above 50 kW may be grouped into a single Ultra-Fast category. This study aims to classify the rated power of public EV charging stations in Jakarta using K-Means Clustering and compare the results with the existing power classification scheme. The analysis was conducted using 731 public charging station points, with rated power as the main clustering variable. The optimal number of clusters was evaluated using the Elbow Method and Silhouette Score. The results show four main clusters: Low Power at 7–40 kW, Mid Power at 47–74 kW, High Power at 100–120 kW, and Ultra High Power at 160–480 kW. The Low Power cluster dominates the dataset with 517 points or 70.7% of the total data. The findings indicate that chargers above 50 kW are not homogeneous and can be divided into more specific power segments. Therefore, data-driven classification can provide a more representative basis for charging station nomenclature, infrastructure mapping, and future charging infrastructure planning.