Use of Machine Learning in Predicting Electric School Bus Battery Range for Optimized Routing

Abstract

The transition to electric school buses (ESBs) promises significant environmental and economic benefits. However, optimizing their operations remains a challenge due to the limited and variable range of their batteries. This paper contributes to addressing this challenge by introducing a machine learning (ML)-based framework for accurately predicting ESB battery range under diverse operational conditions. By leveraging historical and real-time data on energy consumption, traffic patterns, weather conditions, and charging infrastructure, this study develops predictive models that enhance routing efficiency, reduce operational costs, and improve fleet reliability. Our approach integrates advanced ML techniques such as regression models, ensemble learning, and neural networks to create robust range predictions. The study's key contributions include (1) the development of a comprehensive ML-driven predictive model tailored for ESB fleets, (2) the integration of real-time environmental and operational data for dynamic decision-making, and (3) the demonstration of the model's effectiveness through numerical experiments using both simulated and real-world datasets. The findings illustrate the potential of ML in optimizing ESB routing and reducing energy wastage, paving the way for more sustainable student transportation systems.