Transformer fault diagnosis method and device based on asynchronous federated learning and privacy protection

The invention discloses a transformer fault diagnosis method and device based on asynchronous federated learning and privacy protection. The method comprises the following steps: a client initializes first parameters of respective local models; the local data and the shared data are mixed and preprocessed, and a local data set is obtained; inputting the local data set into a local model for training to obtain a second parameter of the local model; calculating an updated difference value of the second parameter of the local model, and introducing Gaussian noise into the difference value to carry out adaptive differential privacy processing to obtain a third parameter of the local model; and aggregating the third parameters to obtain updated parameters of the global model, issuing the updated parameters of the global model to the local model to obtain an optimal local model, and performing fault diagnosis on the transformer by adopting the optimal local model. According to the method, the differential privacy t