Wind Turbine Main Bearing Fault Detection for New Wind Farms with Missing SCADA Data

The installed capacity of wind turbines has been continuously increasing over the past two decades, but it is hard to implement existing bearing fault detection methods to new wind farms since the lack of fault data. To detect main bearing faults for wind turbines installed in new wind farms without relying on their SCADA data, this paper proposed an across-wind-farms fault detection method named IIFDA-V based on the domain generalization method Information Induced Feature Decomposition and Augmentation (IIFDA). The proposed IIFDA-V optimizes the fault decoder additionally by minimizing the risk differences of source domains. Finally, five fault detection tasks are conducted with 8 operational 2 MW wind turbines in 4 different real wind farms, the results indicate the superiority of the proposed IIFDA-V.