Data Driven Matching Protocol for Vehicle to Vehicle Energy Management Considering Privacy Preservat

To address the increasing concern on environmental issues, electric vehicles (EVs) have been attracting more attention. On the one hand, the high penetration level of EVs will bring potential risks to the electricity grids. On the other hand, EVs can enhance the flexibility of the grids via vehicle-to-grid (V2G) technology. With the advancement of digitalization and communication technologies, the concept of vehicle-to-vehicle (V2V) has emerged, which can provide an alternative charging way for EVs to relieve the charging overload problem in the power systems. In this article, we present a data-driven matching protocol for V2V energy management. First, deep reinforcement learning (DRL) is utilized to learn the long-term reward of the matching action based on the formulated Markov decision process (MDP) at an offline stage. To protect the private information of the EV owners, a federated learning framework is proposed where the cooperation between different EV aggregators can be realized without sharing sensitive information. At the online matching stage, a matching optimization model is established and is converted into a bipartite graph problem to enhance the computation efficiency. The proposed methodology is verified in case studies. Simulation results show that the proposed methodology can help EV owners save costs and increase revenues while hedging the trading risks. Besides, the local energy self-sufficiency is increased, indicating that less burden is brought to the electricity network.