Machine Learning Based Electric Vehicle Drivers Charging Satisfaction Analysis and Prediction

In this paper, we develop a prediction model to assess electric vehicle (EV) drivers’ charging satisfaction based on their socio-demographic characteristics. Our main focus was on the human side factors of charging behavior, but not the technical aspects, such as network related topics. We examined and predicted EV drivers’ charging behavior based on socio-demographic factors, vehicle and charging station characteristics, and charging patterns. To understand the charging preferences and habits of EV drivers, we conducted a survey with 225 participants in Hungary. The effect of Several factors including age, driving experience, year of EV adoption, gender, education level, income level, state of charge (SoC), charging fees, and distance from charging stations on EV charging satisfaction was studied. A significant correlation was found between some of these factors and EV charging satisfaction. In addition, we used a feedforward neural network (FFNN) model based on TensorFlow and Keras frameworks to predict future EV drivers’ charging satisfaction levels. We found that the findings of our study have practical implications for the design and planning of EV charging infrastructure and planning of EV charging sessions. In addition to providing insight into the factors affecting EV owners’ charging behavior, they can also advise on the optimal design and placement of charging stations, as well as the best incentives for EV owners. Keywords- Electric vehicle, prediction, machine learning, data analysis, driver behavior, driver satisfaction