Optimizing Electric Vehicle Charging Considering Driver Satisfaction Through Machine Learning

Abstract

Electric vehicles (EVs) are essential to the modernization of transportation systems. However, optimizing EV charging to align with grid stability and renewable energy availability remains a challenge. To address this challenge, this study introduces a machine learning-based framework to optimize EV charging by considering driver satisfaction - a novel approach quantifying this multidimensional construct through socio-demographic attributes, State of Charge (SoC), proximity to charging stations, and variable charging fees. Driver satisfaction is defined as the extent to which the EV charging experience aligns with drivers' expectations, integrating these key factors to influence decision-making and overall happiness with the charging service. Trained on a dataset from Hungarian EV users, the developed model predicts outcomes with high accuracy (87.9%), leading to an optimization algorithm that maximizes driver satisfaction while minimizing grid power purchase costs. Our results from a simulated smart grid demonstrate the model's effectiveness, achieving an average charging satisfaction score of 98.5% compared to 69.54% from a traditional method. Additionally, the proposed method maintained the SoC of the EV fleet at a stable average around 50%, optimizing energy use and grid stability. By dynamically assigning EVs to charging stations and leveraging photovoltaic sources, our solution not only boosts driver satisfaction but also aids in the sustainable growth of smart grids. This research marks a significant step forward in the smart management of EV charging by introducing a driver-centric optimization model, filling a critical gap in current literature and offering insights into its application in enhancing urban mobility solutions.