On-load tap-changer fault diagnosis method and device based on convolutional neural network

The invention provides an on-load tap-changer fault diagnosis method and device based on a convolutional neural network, and the method comprises the steps: collecting a vibration signal generated in the switching process of an on-load tap-changer in real time, carrying out the denoising processing of the collected vibration signal, can effectively remove a noise signal generated in the collection process of the vibration signal, and improves the fault diagnosis precision of the on-load tap-changer. And the influence of noise components on fault diagnosis is reduced. By utilizing the characteristics in the spectral kurtosis curve corresponding to the vibration signals, the defects in the aspect of non-stationary signals can be well described, and the time variance can be measured. And the fault diagnosis model trained by using the convolutional neural network and the SE attention mechanism is used for fault classification, so that redundant features in the image can be reduced, and the recognition efficiency