Understanding electric vehicle charging network resilience: Resilience curves and interpretable machine learning

Abstract

Understanding and enhancing the resilience of urban infrastructure, particularly electric vehicle (EV) charging networks becomes crucial as extreme weather events intensify. While extensive research exists on EV charging infrastructure planning, the resilience of charging networks under extreme weather remains largely unexplored. This study presents a comprehensive framework combining resilience curves with interpretable machine learning to assess and explain charging network resilience during extreme weather events. Utilizing data from 18,061 public charging piles during Typhoon Chaba in Shenzhen, China, we uncover significant spatial–temporal variability in network resilience. XGBoost modeling with SHAP analysis identifies charging station count, commercial land ratio, and raining as primary determinants of resilience patterns. The integration of traditional resilience assessment with modern data-driven analysis provides new insights into charging network dynamics, offering urban planners quantitative evidence and spatially-explicit strategies for developing climate-resilient charging infrastructure tailored to different urban contexts.