Feature extraction method of wind turbine unbalance fault based on Hilbert marginal spectrum and information entropy

Abstract

In view of the difficulty in feature extraction of wind turbine unbalance fault by traditional methods, and the difficulty in determining the proportion of fault type when a fault occurs, a method was proposed to preprocess the signal containing fault information by five-point cubic smoothing filtering, and then to extract features of wind turbine unbalance fault by combining Hilbert marginal spectrum and information entropy. Combined with wind turbine simulation software GH Bladed, the wind turbine unbalance model of 3MW wind turbine was built, and the fault types were identified by comparing the amplitude changes of 1 times turbine frequency in the marginal spectrum diagram under different working conditions, and then the proportion of each fault type was determined by the value of marginal spectrum entropy. The results show that this method is an effective feature extraction method for impeller unbalance faults, and the marginal spectrum has obvious fault type stratification. The proportion of two kinds of unbalance faults in the occurrence of faults is calculated by combining the theory of information entropy, which provides a theoretical basis for practical engineering applications.