Partial discharge pattern recognition based on 2-D wavelet transform and neural network techniques

Partial discharge (PD) pattern recognition is an important tool in HV insulation diagnosis. A PD pattern recognition approach based on the two-dimensional (2-D) wavelet transform and a neural network is proposed in this paper. The approach uses the 2-D wavelet transform to highlight the detailed characteristics of a three-dimensional (3-D) PD pattern. The feature vectors are then extracted from the seven sub-patterns derived by a 3-level wavelet transform and input to a neural network (NN) that implements classification. The recognition rate and reliability are extremely high as compared to the results presented in the literature. It is also suitable for identifying discharges with multiple sources. The capability of the approach was demonstrated by classification of the patterns measured in laboratory experiments.