Transformer fault identification method based on waterfall spectrogram and convolutional neural network

The invention discloses a transformer fault identification method based on a waterfall spectrogram and a convolutional neural network. The method comprises the steps of S1, transformer sound signal denoising: providing a combined denoising method based on wavelet transform and independent component analysis; s2, transformer sound signal feature extraction: analyzing a sound signal time-frequency domain by using short-time Fourier transform to obtain a waterfall spectrogram; and S3, transformer sound signal feature identification: classifying the transformer sound signals by using a convolutional neural network Le Net-5. The main purpose of the invention is to realize state monitoring and fault diagnosis of a transformer based on a sound signal, and provides a combined denoising method based on wavelet transform and independent component analysis to carry out denoising processing on the sound signal, and then a waterfall spectrogram is extracted as feature information. And inputting the obtained waterfall spec