A novel approach to evaluating the accessibility of electric vehicle charging infrastructure via dynamic thresholding in machine learning

Abstract

The spatial deployment of urban public electric vehicle charging stations (PEVCSs) plays a pivotal role in the widespread adoption of electric vehicles (EVs). However, with the rapid advancements in EV technology and battery capabilities, substantial improvements in both range and charging efficiency have emerged and are expected to continue experiencing sustained growth. This situation underscores the urgent necessity of establishing dynamic metrics to reconsider the existing static charging infrastructure, aiming to ameliorate the current severe spatial imbalances and supply–demand disparities encountered in the deployment of PEVCSs. In this study, we harnessed and analyzed 84,152 sets of authentic data, fine-tuned through geospatial-aggregation technology, and ensured anonymity. Our findings bridged users’ residential and occupational patterns with their charging propensities. Comparing these with the spatial distribution of current charging stations revealed that Beijing and Shenzhen’s infrastructure aligned with the cities' economic, educational, and residential zones, epitomizing a synergy in provisioning. However, certain areas experienced either a demand–supply imbalance or an oversupply. To address these challenges, we introduced the Charging Access Reachability Index (CARI) using machine learning techniques. This dynamic metric serves as a tool for quantifying the effective coverage range of charging facilities. Its adaptive threshold holds potential as a crucial indicator enabling the dynamic transition towards more efficient and resilient charging infrastructure.