Fault Diagnosis of Generator Based on D-S Evidence Theory

It is difficult to identify the fault type with the signal gathered from the sensors. In this paper, a new fusion algorithm based on the Dempster-Shafer theory of evidence and neural networks is brought forward. This method combines the advantages of D-S evidence theory and the BP neural network. Neural networks are used to pretreated the data gathered from the embedded sensors in the monitoring system of hydropower plant. Compared with the approaches that only adopt D-S evidence theory or neural networks, the accuracy of diagnostic results is obviously improved, and the signals analysis proved this conclusion. This method has been applied in the monitoring system of JiLin FengMan hydropower plant successfully.