Based On Complex Wavelet Transform Partial Discharge Signal Extraction Algorithm Research

The preliminary problem in transformer's partial discharge on-line monitoring system is to get rid of the interferences from the PD signal. White noise and periodic narrow bandwidth noise, which are the main interferences, are usually suppressed by digital filtering method. There exist some difficulties in the filtering algorithms in the past:for white noise, the toughest problem is how to extract the PD signal from the high level background noise with less distortion; as for narrow bandwidth noise,1) the adaptability,2) the performance when there exist numbers of different interference frequencies in site which overlap badly with spectrum of PD signal,3)the performance when under high level noises, are for the algorithm to maintain. Aiming at these problems, this paper proposes corresponding algorithms using complex wavelet transform and combining information, and to verify which, makes latitudinal and longitudinal comparisons which arrive at some meaningful results.(1) Find through the comparisons of white noise de-noising under different combining information, that the Iwtim has best performance, and through further comparisons, find that the algorithm based on complex wavelet has better SNR performance all the way, and better NCC performance under high level noises.(2) Propose the complex wavelet transform-adaptive filter algorithm. This algorithm makes good use of the natural ability of adaptive filter in suppressing narrow bandwidth noise; utilizes complex wavelet transform's ability of dividing frequency domain to remedy the flaw of adaptive filter; improves the extraction ability with the help of combining information. Given fixed parameters, the complex wavelet transform-adaptive filter algorithm maintains good and stable performances when the features of narrow bandwidth noise changes, it proves this algorithm has a wide range of adaptability.(3) Through some well designed comparisons to two other often used algorithm, the complex wavelet transform-adaptive filter algorithm shows an overwhelming superiority over other algorithms on both SNR and NCC performance and has a potential of practical application prospect.