Fault Diagnosis for Tidal Power Generation Systems: An SWT-VAE-SVM Based Approach

To address fault diagnosis issues for tidal power generation systems, this paper proposes a novel method that integrates Stationary Wavelet Transform (SWT) with a Variational Autoencoder (VAE) for feature extraction, followed by the application of a Support Vector Machine (SVM) for fault classification. The developed method involves three key steps: firstly, current signals are collected over a finite time horizon, whose time-frequency features are extracted via multi-scale SWT; secondly, SWT coefficients are converted into a bank of two-dimensional images, which are used as input of fault diagnosis system with the training and test sets being determined according to a predefined ratio; finally, a VAE-SVM model is employed to identify fault patterns, thereby achieving comprehensive fault diagnosis for tidal power generation systems. A simulation study has been performed, and the results show that satisfactory fault classification accuracy can be achieved by using the proposed scheme.