Research on Charging Behaviour Strategies for Electric Vehicles Based on Machine Learning

Abstract

With the popularity of electric vehicles, accurately predicting charging demand in the workplace has become the key to optimising the grid load and improving the operational efficiency of charging stations. To cope with the uncertainty of users' charging behaviours, this study is based on real charging records from the ACN-Data platform and uses the Random Forest Model to systematically identify the core features that affect charging energy. The analysis results show that charging duration is a decisive factor in predicting charging energy, with a contribution of more than sixty percent. The amount of charging requested by the user is also crucial, with a contribution of twenty-one per cent. This suggests that the user's own charging behavioural characteristics are central to driving demand. This study further provides key insights and data to support the development of high utility and low computational cost charging management strategies. In the future, lightweight random forest models can be explored to be embedded in the edge computing units of charging stations for intelligent scheduling decisions. Building an adaptive charging network with user behaviour prediction as the core will be a key direction to achieve grid interaction optimization and enhance user experience.