Fault Prediction of Wind Turbine Rolling Bearing Considering Multi-time Scale Information

The fault of wind turbine rolling bearing will cause long-term downtime. To accurately predict the fault of wind turbine rolling bearing, this paper proposes a fault prediction method of wind turbine rolling bearing considering multi-time scale information. First, successive variational mode decomposition(SVMD) is used to adaptively extract the multi-dimensional characteristics of bearing health data. Next, the decomposed IMFS is input into the Informer model to extract multi-scale time information for training, and the nonstandard Bayesian optimization algorithm based on tree structure Parzen density estimation (TPE) is used to optimize the super parameters of the Informer model. Then, the fault index based on residual is constructed, and the fault early warning threshold is determined by kernel density estimation(KDE). Finally, the running data is input into the trained Informer model for fault prediction. In this paper, the bearing temperature data of a wind farm are selected for fault prediction. The simulation results show that the SVMD-TPE-Informer model considering multi-time scale information has higher prediction accuracy and calculation efficiency. The proposed method can predict faults with a lead time of 15.5 hours and 10 hours respectively in two fault cases, without any false alarm, which verifies the effectiveness and stability of the proposed model.