Secure and Efficient V2G Scheme through Edge Computing and Federated Learning

Since the road transport accounts for 15% of total carbon emissions, electric vehicles (EVs) have made great strides and large-scale uncoordinated EV charging will greatly increase the load pressure of power system. The vehicle-to-grid (V2G) technology can optimize the idle EVs to charge during the grid load valley period and feeding the grid as a power source during the load peak period. This bidirectional energy flow technology builds a bridge between the power grid and EVs. However, the user privacy-preserving and data asset protection have always been ignored in previous works. In this paper, the secure and efficient V2G scheme through edge computing and federated learning from the double layer has been proposed. Firstly, the edge computing unit is set in the data source, viz., the charging point, to avoid sensitive data leakage. And then, the desensitized charging data will be stored in charging station. Next, the federated learning is utilized among charging stations to jointly train a global model without breaching data asset. Finally, the real dataset is applied to the experiment, and the effectiveness of the proposed architecture is verified.