Research On The Method Of Partial Discharge Signalanalysis Based On Time-frequency Refinement

Transformer, Gas Insulated Switchgear(GIS) and other high voltage power equipment is located in the important part of power production, the insulation performance of power equipment directly affects the stability of the system operation. The analysis of partial discharge signal can detect the insulation defects of equipment in time, prevent the further expansion of insulation failure, and is an effective means to measure the insulation performance of equipment.Partial discharge is a non-stationary random signal, the discharge characteristics of signal in both time domain and frequency domain can not be described by using time domain or frequency domain analysis method separately. In addition, the partial discharge signal often contains noise interference, which affects the accuracy of the frequency analysis of partial discharge signals.Based on the study of the time-frequency analysis method of partial discharge signals, the time-frequency information of the signal is integrated into the suppression of noise interference,meanwhile, the simulation and the measured partial discharge signal are used to study in depth.(1) The frequency slice wavelet transform, which is a new time-frequency analysis method,is applied to the analysis of partial discharge signal. Firstly, according to the Heisenberg uncertainty principle, the optimal frequency slice function is determined, the time-frequency distribution of the partial discharge signal is obtained by using the frequency slice wavelet transform, and the non-stationary characteristics is highlighted. According to the difference of time-frequency distribution of discharge signal, narrow band interference and white noise, using the advantages of frequency slice wavelet transform, such as free time-frequency plane segmentation, to select the time-frequency domain and extract band signals. For the PD signal,which contains random narrow band interference, the region is divided by the time frequency division, for the convenience of study, the reconstructed signals are transformed to frequency domain by Fast Fourier Transform, and corresponding to the narrowband interference steep peaks truncate, inverse transformation to the time domain, eliminate narrowband interference.(2) White noise is distributed in a wider frequency range in frequency domain and overlapped with partial discharge signals in the frequency spectrum, when the interference is serious, it is difficult to achieve the effective separation of the two in the time domain and frequency domain. Choosing the thinning slice area with discharge signal characteristic frequency to reconstitute the signal based on the time-frequency distribution of the PD signal,meanwhile, combined with 3σ criterion to do Further processing of reconstructed signals, to achieve the adaptive separation and extraction of PD signals.(3) Aiming at the problem of the edge effect of the Fast Fourier Transform in the suppression of narrow band interference, the weak correlation between partial discharge signal and periodic narrowband interference is further used, which is the characteristics that the concentrated energy of narrowband interference and partial discharge signal energy is scattered, the total least squares-estimation of signal parameters via rotational invariance techniques is introduced into the suppression of narrowband noise of partial discharge signal. By dividing the narrowband interference subspace and PD signal subspace to estimate the narrowband interference parameters.This method can improve the extraction accuracy of the periodic narrowband interference parameters. The results manifest the method can effectively identify narrowband noise parameters and deal with various forms of partial discharge signal.