Wind turbine generator gearbox fault diagnosis method based on wavelet packet decomposition and convolutional neural network

Abstract

The invention discloses a wind power generation gearbox end-to-end fault diagnosis method based on wavelet packet decomposition (WPD) and a convolutional neural network (CNN). The method comprises thesteps of data preprocessing, multi-scale vibration decomposition, multi-scale feature extraction and classification, adaptive decomposition is carried out on vibration signals by utilizing WPD, thensignal components are input into a hierarchical structure, multi-scale features are adaptively extracted by utilizing the convolutional neural network (CNN) of the hierarchical structure, and faults are effectively classified. According to the invention, the added WPD layer can reasonably process the multi-scale adaptive features of nonlinear and non-stationary vibration data acquisition components, and allows the CNN to extract the multi-scale features; the WPD layer directly sends multi-scale components to the layered CNN, rich fault information is effectively extracted, and useful information loss caused by manual feature extraction is avoided; and the frame universality is good.

Description

technical field [0001] The invention belongs to the field of fault diagnosis, and specifically relates to a fault diagnosis method for wind turbines based on wavelet packet decomposition (WPD) and convolutional neural network (CNN). The real-time diagnosis of faults can provide reliable safety information for the operation and maintenance management of wind farms. Background technique [0002] According to the Global Wind Energy Council (GWEC), wind power is growing fastest, with a significant increase in cumulative installed capacity and is one of the most reliable renewable energy sources. Many wind turbines (WTs) are built at sea, on mountains and in remote locations to obtain abundant wind resources. In these places, the temperature difference between day and night is large, and the weather changes greatly, which will cause great environmental interference to the fan and make the load unstable. Therefore, WTs are prone to failure, seriously affecting the reliability an...

AI Technical Summary

Problems solved by technology

[0007] Existing fault diagnosis methods have their own advantages and disadvantages. Although WPD has strong signal processing capabilities, it is usually combined with shallow machine learning methods for classification, which is not an end-to-end method, and the recognition effect is limited. Although the CNN framework has strong feature learning ability, but lacks the ability to learn multi-scale features of signals

Images