A systematic method for fault diagnosis of steam-turbine generator sets based on the combination of wavelet neural networks and particle swarm optimization is presented. Using the model of wavelet neural networks, we can not only extract the features of system but also predict the development of the fault.The features are applied to the proposed wavelet neural network and the fault patterns are classified. Unlike conventional back propagation training algorithms, the particle swarm optimization does not require gradient information and can provide a stochastic optimal search.It can improve the train speed of the wavelet neural network and can increase the real-time performance of the system.At last ,the hybrid method is applied in the fault diagnosis of steam turbine generators.