In the pursuit of sustainable urban transportation, electric buses (EBs) have emerged as a promising solution to reduce emissions. The increasing adoption of EBs highlights the critical need for accurate energy consumption prediction. This study presents a comprehensive methodology integrating traction modeling with a Light Gradient Boosting Machine (LightGBM)-based trip-level energy consumption prediction framework to address challenges in power system efficiency and passenger load estimation. The proposed approach combines transmission system efficiency evaluation with dynamic passenger load estimation, incorporating temporal, weather, and driving pattern features. The LightGBM model, hyperparameter tuned through Bayesian Optimization (BO), achieved a mean absolute percentage error (MAPE) of 3.92% and root mean square error (RMSE) of 1.398 kWh, outperforming traditional methods. SHAP analysis revealed crucial feature impacts on trip-level energy consumption predictions, providing valuable insights for operational optimization. The model’s computational efficiency makes it suitable for real-time IoT applications while establishing precise parameters for future optimization strategies, contributing to more sustainable urban transit systems.
Battery electric bus (BEB) fleets are pivotal to urban decarbonization policies, yet their operational reliability is critically undermined by stochastic charging-infrastructure failures. This paper addresses a key policy-relevant question: How should transit agencies allocate limited maintenance budgets to balance operational cost and system resilience under uncertainty? We propose a two-stage robust optimization framework that jointly coordinates preventive maintenance investments and adaptive charging scheduling, which explicitly considers the cost-resilience trade-off and provides a decision-support tool for fleet managers. Numerical experiments on real-world-based transit networks show that integrated optimization reduces worst-case operational costs by 26.66% with only about 10% additional maintenance spending. These findings offer quantifiable evidence for resilience-oriented infrastructure investment policies. We further derive managerial insights regarding the value of proactive maintenance under different failure severity scenarios, the impact of uncertainty set weights on resource allocation, and the role of vehicle-to-charger ratios in systemic vulnerability. The proposed framework helps transit operators and policymakers design cost-effective, resilience-guaranteed BEB systems without over-investing in redundancy.