Research on Site Selection and Capacity Determination of Electric Vehicle Public Charging Stations by Integrating K-Means++ and Improved RODDPSO

Abstract

To address the suboptimal spatial distribution and low comprehensive utilization of existing electric vehicle (EV) public charging infrastructure, this study proposes an innovative charging station placement and capacity determination methodology integrating K-Means++ clustering with an enhanced RODDPSO variant. Building upon conventional K-Means and RODDPSO frameworks, we develop an improved hybrid algorithm incorporating three critical advancements: 1) an adaptive mutation mechanism within the RODDPSO architecture to enhance global search capabilities and prevent premature convergence; 2) synergistic optimization of K-Means++ cluster centroids through the enhanced RODDPSO operator; and 3) a novel cluster validation metric based on real-world utilization patterns. The proposed methodology effectively resolves the inherent limitations of conventional K-Means approaches, particularly their sensitivity to initial centroid selection and tendency toward local optima. Empirical validation through a case study of Nanjing's charging infrastructure demonstrates the algorithm's superior performance: stations sited using the proposed hybrid method exhibit 63.8% greater spatial correlation with high-utilization zones (>15% operational utilization) compared to baseline K-Means implementations. The advancements provide both methodological contributions to spatial optimization algorithms and practical insights for urban EV infrastructure planning.