Wavelet de-noising of partial discharge signals based on genetic adaptive threshold estimation

Abstract

Wavelet shrinkage is efficient for de-noising the partial discharge (PD) detection. An improved wavelet de-noising approach for PD online measurement is presented. The wavelet de-noising approach is based on a genetic adaptive threshold estimation (GATE) scheme. The thresholding functions with continuous derivatives are used for the GATE scheme. A genetic algorithm is used to obtain global optimum thresholds of the GATE, and to improve the robustness and computation speed of the adaptive threshold estimation. De-noising experiments of simulative high-frequency PD signals, actual PD ultra-high-frequency (UHF) signals, and a field detected PD signal are presented. The GATE generates significantly smaller waveform distortion and magnitude errors than the Donoho's soft threshold estimation.