Transformer fault diagnosis method based on acoustic signal and recurrent neural network

The invention relates to a transformer fault diagnosis method based on an acoustic signal and a recurrent neural network, and provides a detection method for performing fault diagnosis on a transformer by using the acoustic signal. According to the method, after two sound signal features are extracted, the two features are input into the recurrent neural network in a time sequence mode for training, and a better recognition and diagnosis effect is achieved. The method comprises the following steps: acquiring a transformer sound signal; classifying the samples; preprocessing the sample sound signals; extracting sound signal features; sorting two features of the sound signals; constructing a fault model; training a fault model; importing a transformer sample into the model; and obtaining a result. The fault identification accuracy is higher, and effective support is provided for operation fault identification of the transformer. The invention relates to a transformer fault diagnosis method based on acoustic sig