Early Warning Model of Wind Turbine Front Bearing Based on Conv1D and LSTM

The generator bearing is the key component of wind turbine energy transfer, and its health condition has an important impact on the good operation of the unit. The bearing temperature is an important indicator to reflect the health of the generator bearing. In this paper, the main bearing of direct-drive permanent magnet wind turbine is taken as the research object, and the temperature warning model of generator front bearing is established based on the time-series data of SCADA. Firstly, according to the correlation and wind turbine operation mechanism, five key attributes are selected to participate in the front bearing temperature prediction; Then, the one-dimensional convolution neural network (Conv1D) is used to fuse the information of multiple features at the same time, and the long short-term memory network (LSTM) is used to predict the temperature of the front bearing of the wind turbine in the healthy state; Finally, the statistical threshold of fault warning is designed according to the residual of the validation set. The method is verified by the real historical data of a wind turbine in Hubei Province, China. The experimental results show that the method can capture the abnormal front bearing of wind turbine three months in advance, and reserve enough time for equipment fault handling.