Dynamic Selection of Electric Vehicle Charging Stations Using Deep Reinforcement Learning

This paper presents an innovative Electric Vehicle Navigation System (EVNS) designed to optimize the selection of charging stations for electric vehicles (EVs). By incorporating customer preferences, such as travel time and wait times at charging stations, and addressing uncertainties in traffic conditions and dynamic EV charging demands, we propose an optimal charging station selection system based on Deep Reinforcement Learning (DRL). Utilizing the Deep Q-Network (DQN) algorithm, our system is engineered to minimize both travel and charging times by adapting to fluctuating traffic patterns and predicting future charging demands. The approach involves processing data from EV charging requests, charging station availability, and traffic conditions, framing the selection dilemma as a Markov Decision Process with uncertain transition probabilities. Our simulations reveal substantial benefits over conventional deterministic approaches that merely aim to reduce travel time. This advantage is particularly pronounced during peak periods characterized by high volumes of EV charging requests. The findings underscore the superiority of our DRL-based system in navigating the complexities of real-world EV charging scenarios.