Recognition of partial discharge using wavelet entropy and neural network for TEV measurement

Abstract

Partial discharge (PD) is caused by the deterioration of insulation materials. Its detection and accurate measurement are very important to prevent insulation breakdown and catastrophic failures. Detection of PDs in metal-clad apparatus via TEV method is a promising approach in non-intrusive on-line tests. However, the electrical interference from background environment is the major barrier of improving its measuring accuracy. The combination of wavelet analysis that reveals local features and entropy that measures disorder can just fulfill the requirements of PD signal analysis and is thus investigated in this paper. Then a wavelet-entropy based PD recognition method is proposed. The pulse features that are characterized by wavelet entropy are employed as the input pattern of a classifier constructed with feed-forward back-propagation neural network. Finally, some PD groups with noisy interferences are tested by trained network. The recognition rate of real PD pulses demonstrates the proposed wavelet-entropy based method is effective in PD signal de-noising.