Simulative Tests And Pattern Recognition Of Stator Winding Partial Discharges In Large Generators

The partial discharge (PD) of large generator winding is an important reason for insulation damages. The PD contains such sufficient characteristic infonnation that it can be used to inspect the stator winding insulation condition to avoid sudden failure specially for on-line monitoring. This paper made Five types of different simulative physical insulation models, which reflect stator winding PDs in large generators, a lots of simulative PD tests were done on the model bars. The characteristics of different types of models were obtained from wave of PDs and 4 graphs of PDs. This paper applied the method based on an artificial neural network (ANN) and the method of fractal to recognize PD patterns. In the method based ANN, the ANN with back propagation algorithm was designed to identify the types of PDs; a statistic method was used to analysis the extent distributions of PDs, which made it very practical to identify the extent of PDs. The research shows that the method based on ANN can recognize PD patterns very accurately and the fractal method can recognize PD patterns in some extent. The presented pattern recognition methods can also be used for PD monitoring and insulation diagnosis of other electrical apparatus.