The invention relates to a wind generating set bearing fault prediction method based on a neural network, which relates to the field of wind generating set fault diagnosis. The method comprises the steps of: S1, collecting historical operation data of a wind driven generator bearing, carrying out data preprocessing, and carrying out the data standardization and data missing value filling of initial data, S2, carrying out frequency conversion by using improved stationary wavelet packet conversion so as to carry out frequency bandwidth separation, and extracting a fault characteristic frequencyvalue, S3, using an Elman artificial neural network and using the training set to train the Elman artificial neural network to obtain a neural network model, and S4, performing fault prediction on theinput real-time data. According to the method, the accuracy of wind driven generator bearing fault prediction is improved to a large extent, and the operation speed is obviously improved.