Electric Vehicle Charging Behavior Prediction using Machine Learning Models

Abstract

The global spread of the COVID-19 pandemic has significantly impacted the electric vehicle (EV) industry. The lockdown restriction has resulted in a significant shift in the use of public charging infrastructures. This paper investigates the effects of COVID-19 on electric vehicle users' charging behavior before, after, and during COVID-19 lockdown restrictions, using the data from a public charging facility from the City of California. In this study, we performed data visualization using K-means and hierarchical clustering analysis. This work uses the vehicle's connection and disconnection time to identify common charging pattern identification and charging behavior where K-means clustering outperforms the hierarchical clustering for all three different scenarios modelled. In addition, prediction of collective charging session duration is achieved using Machine Learning Models, Random Forest and XgBoost. We achieved a mean absolute percentage error (MAPE) of 0.146 and 0.151 percent for XgBoost and Random Forest respectively.