Abstract: Due to the complex operating state of the new power system, traditional methods are difficult to accurately diagnose relay protection circuit faults. Therefore, this article proposes a new fault diagnosis method for power system relay protection circuits based on improved wavelet neural networks. Collect and preprocess three-phase electrical data of the new power system relay protection circuit, design a multi-layer compact wavelet neural network structure, and determine the number of nodes in each layer. Adopting genetic algorithm to improve the training method of wavelet neural network, inputting preprocessed electrical quantity data. The experimental results show that the fault diagnosis accuracy of the new power system relay protection circuit under the design method is as high as 97.5%, and the diagnostic effect is good.