A method for classification and detection of faults in wind turbine components

Abstract

The invention discloses a method for classifying and detecting faults of wind power generator components, which belongs to the technical field of wind power generator state monitoring and includes the following steps: Step S1 performs wavelet decomposition on each data input signal collected by a wind power generator monitoring control and data acquisition system , to obtain the local signal of the original input signal; Step S2 inputs the original signal and the local signal decomposed into the long short-term memory neural network for feature learning; Step S3 connects the output of each sub-network of the global feature and the local feature, and performs The dynamic weighting process realizes the adaptive dynamic fusion of the global feature and the local feature; Step S4 passes the classification result through the method of sliding window and majority voting to generate the final detection result. The invention can effectively improve the accuracy rate of failure classification of wind power generator components, thereby timely processing and maintaining the wind power generator faulty parts, and avoiding deep damage to wind power generator components.

Description

technical field [0001] The invention relates to a fault classification and detection method for wind turbine components, belonging to the technical field of wind turbine state monitoring. Background technique [0002] Wind energy is a kind of clean and renewable energy, which has developed rapidly in recent years, and the installed capacity of wind turbines has continued to increase. However, with the continuous and rapid development of wind power and the continuous increase in the scale of wind turbine assemblies, the failures of in-service wind turbines continue to emerge, such as unit collapse, blade icing, blade cracking and other accidents, which lead to the operating efficiency of wind power equipment units. Low, short life, high failure rate, poor reliability and other problems are becoming increasingly prominent, and maintenance costs remain high. Factors such as downtime losses caused by faults not only hinder the economic development of wind farms, but also have a...

AI Technical Summary

Problems solved by technology

[0004] At present, there is a classification method for component fault detection by processing SCADA data of wind turbines. However, due to the large amount of SCADA monitoring data and high dimensionality, the existing component fault detection classification methods generally have clear correlations due to the close correlation of data time series. Insufficient accuracy, which affects the accuracy and accuracy of fault prediction

Images