The rapid increase in electric vehicle (EV) usage has led to an urgent need for coordinating EV charging activities with power distribution networks (PDNs) to accommodate the resulting redistributed electrical demand and implement greater operational flexibility for PDNs. However, current efforts relying on the implementation of EV charging price strategies to influence the decision-making activities of EV users at charging stations face challenges arising from the uncertainty of EV charging behavior and data privacy concerns. This work addresses these issues by developing a robust reinforcement learning approach that requires no personal EV user information to determine charging schedules and pricing strategies at charging stations, while the uncertainty in EV behavior is addressed by applying a robust reward function. The loss function is relaxed using Holder’s and Cauchy-Schwarz inequalities, which yields an upper bound for the loss caused by the worst case scenario and therefore enhances the computational efficiency of the solution process. Numerical results demonstrate that the proposed method contributes to balancing the PDN load and improving the utilization of fast charging stations (FCSs). Compared to deterministic and traditional robust methods, the proposed method reduces computation time in uncertain environments while ensuring moderate revenue of FCSs.
Toward Efficient Coordination of Power Distribution Network and Electric Vehicles: Deep Reinforcement Learning With Robust Reward Function
IEEE Transactions on Transportation Electrification ; 11 , 3 ; 8164-8176
2025-06-01
2513917 byte
Article (Journal)
Electronic Resource
English
Reward Function Optimization of a Deep Reinforcement Learning Collision Avoidance System
ArXiv | 2022
|Aero-engine acceleration control using deep reinforcement learning with phase-based reward function
SAGE Publications | 2022
|