Asynchronous Decentralized Federated Learning for Collaborative Fault Diagnosis of PV Stations

Due to the different losses caused by various photovoltaic (PV) array faults, accurate diagnosis of fault types is becoming increasingly important. Compared with a single one, multiple PV stations collect sufficient samples, but their data not allowed be shared directly due potential conflicts interest. Therefore, federated learning can exploited train collaborative model. However, modeling efficiency seriously affected model update mechanism since each station has computing capability and amount data. Moreover, for safe stable operation system, robustness must guaranteed rather than simply being processed on central server. To address these challenges, novel asynchronous decentralized (ADFL) framework proposed. Each only trains its local also participates in exchanging parameters improve generalization without losing accuracy. The global aggregated distributedly avoid node failure. By designing scheme, communication overhead training time are greatly reduced. Both experiments numerical simulations carried out verify effectiveness proposed method.