Clustering-based EV suitability analysis for grid support services

Abstract

The literature extensively discusses the benefits of electric vehicles (EVs) for grid support services. However, not all EVs are suitable for these services due to variations in service requirements (duration and frequency) and EV driver behavior (available energy, parking duration, and battery degradation). This study proposes a three- step EV classification approach to assist aggregators in selecting appropriate EVs for specific services. First, using data from the National Household Travel Survey and a commercial EV database, various EV parameters are estimated, including available energy for grid support services, the duration of parking at home and at work, and the battery degradation factor. Second, the K-means clustering method is applied to categorize EVs based on each parameter, chosen for its superior performance and lower complexity compared to other clustering methods. Finally, suitability indices are proposed for each service, taking into account the service requirements and EV parameters of different clusters. Each EV is then ranked to help the aggregators select the best-suited EVs for each service. The performance of the proposed method is evaluated for two ancillary services (frequency regulation and ramping) and two operating reserve services (contingency spinning and supplemental reserves). Simulation results demonstrate that more EVs are suitable for home services due to longer parking hours, while those with low parking duration are unsuitable for workplace services, despite low degradation scores. Additionally, the proposed method consistently shows higher poolable energy compared to the random selection method, with differences reaching 56 kW (10.1%) for 10 EVs, 383 kW (26.8%) for 25 EVs, and 895 kW (31.2%) for 50 EVs.