Short-Term Prediction of Generator Blade Ice Fault Based on Multi-AN

In order to improve the problem that the accuracy of the existing methods are normally low when predicting icing faults of wind turbine blades, in this paper, a prediction method of fan blade icing fault based on multi-artificial neural network algorithm is presented. Firstly, the key features are extracted by dynamic principal component analysis (PCA) algorithm, and then the eigenvalues of all kinds of features in the near future are predicted by Elman artificial neural network (EANN). Finally, the back propagation (BP) artificial neural network is trained with a large number of working condition label data, and self-clustering judgment is made to predict the trend of wind turbine blade icing fault. Compared with the typical algorithm of support vector machines for regression (SVR) and other methods verifies that the prediction error of Elman artificial neural network is smaller, which illustrates the effectiveness of this method.