融合多源地理数据与机器学习的新能源汽车充电站选址规划研究——以郑州中心城区为例

Integrating Multi-Source Geospatial Data and Machine Learning for Electric Vehicle Charging Station Siting—With a Case Study of Zhengzhou’s Central Urban Area

摘要: 为解决新能源汽车充电站紧缺问题，提出一种融合多源地理数据与机器学习的充电站选址方法。网格化郑州中心城区，通过多阶段统计从2025年POI中筛选出7类与充电站分布具有显著协同效应的类别。运用SHAP方法，解析决策树模型中各类POI的特征贡献并进行空间协同分析。构建空间扩散模型，从空间邻域视角评估候选点的综合得分并排序。制定一种分层选址策略，从1440个网格中筛选出297个有效候选点，同时覆盖繁华区域与非繁华区域。为城市充电基础设施规划提供了数据驱动的决策支持。

Abstract: To address the shortage of electric vehicle charging stations, the paper proposes a site selection method integrating multi-source geospatial data and machine learning. First, the central urban area of Zhengzhou is gridded, and 7 types of POI with significant synergistic effects on the distribution of charging stations are screened out from the 2025 POI data through a multi-stage statistical analysis. Then, the SHAP method is applied to interpret the feature contribution of each POI type in the decision tree model and conduct spatial synergy analysis. A spatial diffusion model is constructed to evaluate and rank the comprehensive scores of candidate sites from the perspective of spatial neighborhoods. Finally, a hierarchical site selection strategy is formulated, and 297 valid candidate sites are selected from 1440 grids, covering both bustling and non-bustling areas, providing data-driven decision support for the planning of urban electric vehicle charging infrastructure.