Research On The On-line Monitoring Of Partial Discharges On Large Generator Stator Windings

The on-line monitoring of partial discharge on large generator stator windings is an effective way to diagnose the insulation condition of stator windings. In this paper the time features and the frequency features of three different types of partial discharge (PD) occurred on generators, such as delamiantion PD, end winding PD and slot PD, were analyzed based on the laboratory test of generator stator winding model bars. The frequency band of the detector for on-line monitoring partial discharges on generators was discussed. A high frequency setup (HF) and a new low frequency (LF) partial discharge detector named PDD-4 were used to investigate the features of single pattern partial discharges and composite pattern partial discharges on large generators. The test results showed that the three different types of partial discharge on generator stator windings had different statistical features and the features under the two different detectors were similar. The spectrums and their statistical parameters of different types of partial discharge were used to study on the pattern recognition of partial discharges on generators by a BP neural network. The research shows that the three different types of partial discharge on generators have very different features and we can recognize their patterns using the spectrums and their statistical parameters by neural network quite well.