Clustering methodologies for flexibility characterization of electric vehicles supply equipment

Abstract

The continuous growth of electric vehicles (EVs) poses new challenges to power systems planning and operation due to the need to meet society's decarbonization goals. In this context, clustering has emerged as a powerful tool to help understand and categorize the uncertain behavior of EV users and the electric vehicle supply equipment (EVSE) needs. However, previous studies lack empirical European EV charging data and relevance for practical applications.Therefore, to address such issues, this study evaluates different clustering techniques to identify typical EV charging profiles and, mainly, usage flexibility. The defined methodology comprises three major stages: data preprocessing, clustering application, and validation of results. We conduct benchmarking based on EV energy consumption, arrival, and sojourn times, using K-means, Gaussian mixture model, and Hierarchical clustering. This method allows greater applicability to various datasets from different regions, producing more comprehensive profiles that can provide empirical flexibility data in a visual, intuitive, and relevant approach.A use case considering EV charging data from Caltech University and Greece is utilized to test the proposed methods, demonstrating the versatility of our methodology.Specifically, Caltech features highly flexible prolonged charging sessions, while Greece exhibits quick-stay sessions with less flexibility potential. Both contexts offer opportunities to use the available flexibility for coordination with renewable energy sources and help balance the grid. This information unlocks the potential for future studies, enabling distribution system operators and charge point operators to intelligently and successfully integrate EVs into the energy system.