Interaction Framework and Method of Electric Vehicles Aggregator and Distribution Transformer Area Based on Federated Learning

With the large - scale adoption of electric vehicles (EVs), their interaction with the power grid has become crucial for maintaining electricity balance in distribution networks. However, data privacy protection in the exchange between EV aggregators and grid operators faces increasingly severe challenges. To address this, this paper proposes an interactive framework based on federated learning, aiming to achieve collaborative optimization between EVs and distribution networks while ensuring data privacy and security. This framework innovatively establishes a many - to - many federated learning architecture that enables multiple charging stations (EV aggregators) to collaborate efficiently with multiple distribution transformer areas. Raw data remains local, and only model update information (gradients and parameters) is transmitted, effectively safeguarding the privacy of distribution network data. Simulation results demonstrate that the proposed framework achieves model performance comparable to centralized training without compromising privacy, with no significant increase in the number of iterations, and training time primarily influenced by communication efficiency. Furthermore, through reasonable model design and parameter optimization, the framework can effectively balance EV charging demands and distribution network loads, and promote the accommodation of distributed renewable energy.