A Study on Fault Diagnosis of Hydroelectric Generator Based on D-S Evidence Theory

The fault of hydroelectric generator can be reflected by different characteristic signal from different side, due to complexity of fault reason in hydroelectric generator. In order to improve the accuracy of fault diagnosis, a data fusion method using multi-neural network and D-S evidence theory is presented in this paper. The combination of multi-neural network with D-S evidence theory can take full advantage of one's own merit. The output value of every diagnosis subsystem is regarded as input value of decision fusion level module. In this paper, the feature level module adopts three different fault diagnosis subsystems BP neural network, radial basis function (RBF) network and the fuzzy neural network (FNN). Its function is to do local fault diagnosis and to get basic probability assignment (BPA) of D - S evidence theory. An example of mechanical vibration fault in hydroelectric generator is used for simulation experiments, the simulation results verify the method for hydroelectric generator fault diagnosis based on D-S evidence theory has better accuracy.