I. Introduction
With the development of wind technology, the power of a single fan has increased more than threefold in the past 30 years. It is expected that the maximum power of wind turbines on land and at sea will reach 15 GW and 35 GW by 2050 respectively[1]. Considering the increase of the capacity of a single fan, its structure becomes more and more complex. Most of them are distributed in the outer suburbs and offshore areas with complex environment, which leads to the higher failure rate and operation and maintenance costs of the fan[2]. According to the investigation, it is found that the failure rate of bearings in the fan gearbox and generator is as high as 80%, and the bearing failure is the main cause of fan shutdown[3]. If the bearing fault is not diagnosed in time, the maintenance cost of the fan will be increased by about 5 times. It may even cause social and economic losses such as power failure[4]. In order to realize the safe and stable operation of wind turbines and reduce maintenance costs, it is urgent to conduct intelligent fault diagnosis research on wind turbine bearings.