Resilient EV charging station network design using AI algorithms

Abstract

This paper proposes an AI-driven resilient network design framework for optimal electric vehicle (EV) charging station placement under stochastic demand and dynamic grid constraints. The proposed approach uniquely integrates long short-term memory (LSTM) based spatiotemporal demand forecasting with a hybrid genetic algorithm-particle swarm optimization (GA-PSO) model for multi-objective station placement. In addition, a deep reinforcement learning (DRL) agent is incorporated to enhance adaptive resilience under real-time grid disturbances. The framework minimizes installation cost, reduces user travel distance, and improves grid stability while ensuring equitable accessibility. The model is evaluated under multiple scenarios, including peak demand, station outages, renewable intermittency, and grid capacity reduction. Results demonstrate that the proposed hybrid AI framework achieves a resilience index of 0.92, reduces travel distance by 54%, and lowers installation cost by up to 16% compared to conventional approaches such as linear programming (LP) and K-means clustering. The integration of renewable energy further reduces peak grid dependency by 18%. The proposed methodology provides a scalable and practical solution for designing sustainable and resilient EV charging infrastructure in smart urban environments.