Dynamic Reactive Power Support through Mobile Charging Stations Using Reinforcement Learning

This paper proposes a novel method to optimize Electric Vehicle (EV) charging and reactive power support for distribution grids using Mobile Charging Stations (MCS). By leveraging the mobility of MCS, we tackle the combined problem of routing and scheduling MCS for EV charging while providing dynamic reactive power support to stabilize the grid. The problem is modeled as a Markov Decision Process (MDP) and solved with Deep Q-Networks (DQN), a reinforcement learning algorithm for real-time decision-making. The state space includes grid voltage, MCS battery status, power injections, and EV charging requests, while the action space covers routing decisions and MCS power outputs. The goal is to maximize the cumulative reward by balancing successful EV charging and grid stability. The reward function penalizes voltage violations and operational costs, with incentives for efficient charging. Simulations show the effectiveness of the DQN-based approach in optimizing EV charging and reactive power support, reducing voltage deviations, and enhancing grid performance.