Study On Suppressing White Noise Of Partial Discharge Signals By Complex Wavelet Transform And Evaluation Parameter

Partial Discharge (PD) in Gas Insulated Switchgear (GIS) is one of the important factors for the fault; therefore the on-line monitoring PD in GIS has been researched at home and abroad in order to guarantee the safe operation of GIS. At the same time, PD pattern recognition is still an advanced problem to be studied, and it is difficult to accurately obtain the PD signals from the intense electromagnetic interferences.According to the idea of changing the phase-frequency characteristic and holding the amplitude-frequency characteristic of real wavelet, the dbN (N=2, 3,…) complex wavelets are presented. It has the same characteristic to the dbN real wavelet.Normalized Correlation Coefficient (NCC) can only describe the whole similarity of the waveforms, can not evaluate the oscillating signals accurately. So a new parameter which is used to describe the oscillating signals─Variational Trend Parameter (VTP) is put forward. And its conception is given and its mathematical and physical meanings are formulated. The necessity and rationality of setting up VTP is verified by signals'power spectrum. It is further proved that the de-noising effect can be better synthetically evaluated by VTP, NCC and SNR.The combined information of serial WTRIn, which is used to suppress white noise, is constructed based on the characteristic of PD signals and white noise. The study proves that the effective de-noising combined information of different kinds of PD signals are not only related to the signal characteristic, but also related to the change of noise amplitude. When the noise intensity is increasing, index n is raising. But different kinds of signals have different raising rates.The complex wavelet coefficients of each scale have different characteristics, so an algorithm of complex multilevel threshold by which each scale's complex wavelet coefficients can select their own complex thresholds separately is presented. It is indicated that the complex multilevel threshold can suppress white noise more effectively than complex whole threshold.According to the characteristic of the complex wavelet coefficients of PD signal and white noise and the basic theory of block threshold, self-adaptive threshold and complex threshold, the complex self-adaptive block threshold is proposed. Then the de-noising theory of this threshold is analyzed and its calculating process is given. Through the emulation, it is proved that the complex self-adaptive block threshold can suppress white noise effectively with small distortion and high SNR.Through the research on the real threshold de-noising the simulation signals and the complex threshold de-noising the real PD signals, it is shown that the complex threshold can suppress white noise more effectively.