Pattern recognition of single and composite partial discharge on generator stators

Partial discharge (PD) can indicate faults in insulation of generator stator windings, it is important to identify the PD pattern for later inspection of generator stator windings. In this paper, some generator model bars were made and some typical PD phenomena in generator stator windings were simulated. PD tests were tried to find the features of composite PD. The data gotten from the test were analyzed using statistical parameters and the spectrums of different types of PD were plotted to reveal the difference of different patterns of PD. A BP neural network was also used to recognize the PD patterns using some new input parameters.