Fault warning study of gearbox based on SOM-ASTGCN-BiLSTM and mutual diagnosis of same clustered wind turbines

Accurate warning of the low-speed bearing temperature of a wind turbine gearbox is the basis for ensuring its healthy and stable operation. Therefore, a gearbox fault warning method based on self-organizing map (SOM)-attention-based spatiotemporal graph convolutional network (ASTGCN)- bidirectional long short-term memory network (BiLSTM) and mutual diagnosis of the same clustered wind turbines was proposed. This method utilizes the SOM clustering algorithm to cluster wind turbines with similar external environments and operation states into one cluster, which provides support for the mutual diagnosis of the operation states of the same clustered wind turbines. An ASTGCN was used to deeply mine the spatiotemporal correlation characteristics between the operating state data of the wind turbine and the temperature value of the gearbox low-speed bearing. A BiLSTM was used to bidirectionally mine the temporal correlation between the operating state data of the wind turbine and the temperature value of the gearbox low-speed bearing, and a forecasting model of the gearbox low-speed bearing temperature based on ASTGCN-BiLSTM was constructed. The temperature of the gearbox low-speed bearings of the same clustered wind turbines exhibited a similar dynamic change process. By comparing and analyzing the distribution characteristics of the forecasted temperature values of the gearbox low-speed bearings of the same clustered wind turbines, it is possible to accurately identify wind turbines with abnormal gearbox operating states. Taking a certain wind farm as the calculation object, the calculation results show that the forecasting accuracy of the proposed SOM-ASTGCN-BiLSTM model is higher than that of other models such as ASTGCN, Reformer, Transformer, Informer, Pyraformer, QR-LSTM, and PSO-ELM, proving the superiority of the algorithm proposed in this study. The mutual-diagnosis strategy for the same clustered wind turbines can accurately identify wind turbines with abnormal gearboxes.