Understanding User Behaviour and Predicting Charging Costs: A Machine Learning Approach to Support Electric Vehicle Adoption Decisions

Authors:
Open Access
Publisher website
Google Scholar
Abstract
The increasing adoption of electric vehicles (EVs) necessitates a comprehensive understanding of charging patterns and user behaviour to enable future transportation infrastructure to be planned and designed to meet user needs. This study uses machine learning to predict the costs of EV charging sessions and analyse user behaviour to support strategic planning and decision‐making. We examined data that included factors such as total energy consumption and charging duration, and compared three models: linear regression, random forest, and gradient boosting. The gradient boosting model performed the best, with a mean squared error of 0.041 and an ‐squared () of 0.91. Additionally, the analysis of user behaviour revealed peak charging times between 6:00 PM (18:00) and 9:00 PM (21:00), with the majority of sessions occurring on weekdays, particularly Wednesdays. Most users preferred charging infrastructures within a 10‐mile radius. These insights not only enhance the understanding of current EV charging behaviours but also provide valuable information for local authorities and decision‐makers in transportation planning and infrastructure development. By integrating predictive modelling and behavioural analysis, this research offers a novel and robust framework for designing EV charging networks, addressing user needs, and advancing the sustainability of urban transportation systems. This approach not only supports the efficient deployment of charging infrastructures but also introduces the concept of charging comfort by aligning infrastructure development with real user needs. Unlike traditional methods that overlook user preferences and waiting times, our model integrates behavioural analysis to improve the overall user experience. By quantifying when, where, and how users prefer to charge their vehicles, this framework supports not only infrastructure optimisation but also enhances user satisfaction, a key factor in accelerating EV adoption and reducing the environmental burden of urban mobility.

This publication has 48 references indexed in Scilit: