Differential Privacy Federated Learning based Privacy-Preserving Transmission Scheme of Distributed PV Stations

As a renewable and sustainable energy source, solar energy is attracting more and more attention, where photovoltaics being one of the primary ways. However, the distributed nature of the PV stations raises serious privacy issues, as the data for each PV station is owned by different utilities that are unwilling to share their data. In order to effectively ensure data security and privacy, this paper proposes a novel federated learning (FL)-based distributed PV station transmission framework. Since the private information can still be leaked by analyzing the uploaded parameters of PV stations. Differential privacy (DP) algorithm is added to the proposed FL-based PV station transmission framework by inserting artificial noise before the aggregation process of FL, so that the attacker cannot tell whether a particular customer is included in the dataset or not. The simulation results evaluate the performance of the proposed algorithm with different data distribution, privacy leakage and DP mechanisms.