Decadal analysis of spatial and temporal dynamics in energy consumption: Insights from the Beijing Subway

Urban subway systems are essential for sustainable transportation, yet their energy consumption patterns remain insufficiently understood, particularly in the context of rapid urbanization and network growth. This study examines traction (TEC) and auxiliary (AEC) energy consumption on Beijing Subway from 2013 to 2022, considering operational, structural, network topology, and weather dynamics. We introduce an adaptive framework that combines CatBoost forecasting with SHAP-based interpretability, addressing the critical need for both high predictive accuracy and transparent attribution of key factors influencing energy consumption. Our analysis uncovers that TEC is dominated by service metrics (e.g., travel kilometers), whereas AEC is primarily driven by line structure and climate. Spatially, central and loop lines exhibit stronger factor sensitivities than peripheral ones; temporally, structural influence grew by + 6.7 % (TEC) and + 5.4 % (AEC). By offering a transparent, data-driven view of subway energy use, this work informs strategies for low-carbon operation in rapidly growing metropolises, enabling energy-led network planning, demand-responsive energy allocation, and condition-adaptive scheduling.