An online energy consumption estimation method for different types of battery electric buses based on incremental learning

Accurately estimating the real-time energy consumption (EC) of battery electric buses (BEBs) is critical for optimizing eco-driving strategies and energy management. Such estimations can significantly enhance the energy efficiency and economic viability of urban public transportation systems. However, research on online estimation models remains relatively limited. This study proposes a model employing incremental learning for real-time EC estimation, enabling continuous parameter updates to adapt to dynamic environments. Natural driving data collected from eight BEBs was analyzed to evaluate the influence of dynamic parameters on EC and to investigate variations in EC characteristics among different buses. Based on these insights, personalized input features were identified for the model, tailored to specific bus types using vehicle dynamics and correlation analysis. An online EC estimation model based on incremental learning was subsequently developed, and experiments were performed to optimize its structure and fine-tuning strategy. The results demonstrated an average mean absolute percentage error (MAPE) of 5.992%, validating the model's robustness across diverse conditions. These findings highlight the practical applicability of the proposed model for real-time EC estimation in BEBs, contributing to enhanced energy efficiency in public transportation systems.