Comparative analysis of optimization techniques for optimal EV charging station placement

Abstract

The optimal placement of electric vehicle (EV) charging stations plays a crucial role in improving accessibility, reducing travel distances, and minimizing infrastructure costs in smart urban planning. This study presents a comparative analysis of traditional optimization techniques-such as linear programming (LP), particle swarm optimization (PSO), k-means clustering, and greedy heuristic methods-alongside a machine learning-based approach using genetic algorithms (GA). A machine learning framework is implemented to simulate EV charging demand, optimize station deployment, and incorporate real-world constraints like cost, grid capacity, and user travel penalties. The results demonstrate that GA achieves superior performance in balancing cost-efficiency and user convenience, outperforming traditional techniques in solution quality under dynamic demand conditions. PSO and LP provide faster convergence but are less adaptive to changing parameters. The study highlights the potential of integrating machine learning into infrastructure planning and provides actionable insights for urban planners and policymakers in developing resilient and intelligent EV charging networks.