Towards Decentralized Coordination of the Charging Demand of Connected Electric Vehicles: Game Theory vs. Reinforcement Learning

With the growing global interest towards sustainable mobility, electric vehicles (EVs) are gaining increasing popularity as a zero-emissions transportation solution. Nonetheless, the growing penetration of EVs translates into an increase in the overall electricity demand of the transportation sector, which imposes challenges on the power and transportation networks alike. This calls for optimal EV charging coordination strategies to address the growing demand. Decentralized coordination algorithms can be utilized to encourage EVs to optimize their charging decisions towards their self-interests while acknowledging community-level advantages. This work presents a comparative analysis between two decentralized charging coordination approaches, namely game theory (GT) and reinforcement learning (RL). Based on the conducted simulations, superior performance is obtained using the GT-based model in comparison with a Q-learning RL model and a benchmark random selection algorithm, in terms of the individual and aggregate utilities of all involved players. This preliminary study paves the way for further investigation into utilizing advanced GT- and RL-based models for decentralized EV charging coordination.