Behaviour analysis of electric vehicle charging flexibility: A data-driven approach

This paper delves into the unexplored domain of electric vehicle (EV) flexibility in charging behaviour, employing an analysis of 3365 EV charging sessions. The findings underscore the significance of extracting valuable insights from the aggregated U.S. Department of Energy's (DOE) dataset, emphasizing the untapped potential of data analysis in power systems. The innovative methodology involves clustering arrival and departure times to identify distinct charging behaviours. Through a detailed examination of each behavioural cluster, encompassing factors such as arrival, departure, and stay duration times, the study proposes measures and an algorithm to quantitatively assess flexibility utilization. The research explores the applicability of data-driven solutions in real-world scenarios, introducing a behaviour analysis method capable of efficiently categorizing large datasets into user groups, including those utilizing spontaneous charging outlets (SCOs), office charging outlets (OCOs), and home charging outlets (HCOs). Addressing the challenge of obtaining precise datasets, particularly with maximum flexibility, machine learning methods are applied. The paper demonstrates the integration of EVs and different machine learning-based clustering techniques which are compared, utilizing clustering for refining extensive datasets and extracting meaningful data.