Feature extraction for multi source partial discharge pattern recognition

Partial discharge (PD) tests and its analysis plays vital role in insulation quality assessment. The amount of insulation degradation depends upon the type of discharge. Classification of discharge sources becomes necessary to know the type of discharge occurring in the power equipment. In this paper six types of discharges have been addressed for classification, out of which three are single source and remaining three are two source discharge types. Wavelet transform is effectively utilized for de-noising and as well as feature extraction from phi-q-n pattern. The effectiveness of neural network (NN) system for multi source PD pattern recognition is investigated.