Unraveling energy demand in battery electric bus operations through an explainable machine learning approach using real-world cold-climate data

Battery electric buses (BEBs) are increasingly deployed as a low-emission solution in public transit systems, yet their energy performance under cold climate conditions remains largely understudied. Seasonal changes in temperature and road conditions introduce significant variability in energy consumption, while empirical data from cold regions are limited. This study addresses this gap by analyzing over 66,000 real-world trip records from 29 electric buses operating in Montreal between June 2022 and October 2023. Key variables such as motion time, average speed, auxiliary heating demand, and regenerative braking efficiency were extracted and examined across seasons. Four models were tested for trip-level energy prediction, including multiple linear regression, random forest, XGBoost, and multilayer perceptron. The XGBoost model achieved the best performance (R squared equals 0.96, RMSE equals 1.33 kW h). Results indicate that energy consumption increases by up to 26 % in winter, driven by heating loads and adverse driving conditions, while regenerative braking efficiency declines from 53.4 % in summer to 32.2 % in winter. Speed also plays a critical role, with optimal energy recovery observed at 30–40 km/h. Despite seasonal variations in performance, BEBs maintain strong economic advantages over diesel alternatives. Findings underscore the need for adaptive operational strategies, such as temperature-aware scheduling, route optimization, and integrated charging planning, to optimize BEB deployment in cold regions. This research offers practical insights for transit agencies aiming to expand electrified fleets under variable climate conditions.