As the information source of relay protection and automation devices, the operation performance of electronic transformer will directly affect the grid. Influenced by environment and electromagnetic radiation, the stability of its output signal becomes the bottleneck of popularization and application for intelligent substation. By means of classifying the fault types of electronic transformer in operation, the different fault types with the piecewise function are simulated and fault signal with wavelet theory analysis is denoised. Then, by means of fault detection and location of the denoised signal with wavelet decomposition, a method of wavelet-neural network for fault diagnosis analysis is deduced which can directly distinguish the drift deviation fault, fixed deviation fault and transformation ratio deviation fault. It will be a prospective significance for improving the operation reliability of the intelligent substation.