Study On Denoising Methods And Feature Extraction For On-line Transformer Partial Discharge Monitoring

Partial discharge is an important factor resulting in transformer insulation degradation,transformer partial discharge on-line monitoring can reflect real-time transformer insulation states, which has important significance to preventing power transformer accidents and power system security.In the process of partial discharge monitoring, partial discharge denoising technology and discharge characteristics extraction are the keys to the study of partial discharge. This paper Studies the discharge mechanism and discharge process of partial discharge, and establishes a kind of single air gap partial discharge simulation model, and illustrates the physical process of partial discharge's generating by theoretical analysis and simulation. The paper also describes mathematical models of partial discharge and noise interference in theoretical analysis.The white-noise and periodic narrow-band interference is often accompanied with on-site detection of partial discharge. A new method that interfuses wavelet package transform with generalized morphological filters is used for denoising in order to extract the partial discharge signal from mixed noise.The results show that the method of wavelet package interfusing with generalized morphological filters for denoising can eliminate the mixed noise effectively and the characteristics of original signals can be well retained, and the distortion rate of filtered partial discharge signal is low.Aiming at four typical partial discharge pulses in transformer, the paper puts forward two new discharge feature extraction methods. Firstly, the application of wavelet analysis technology, combines fractal theory, and the fractal dimension of signals can be calculated and extract fractal characteristics in each frequency-band. Secondly, the pattern spectrums of partial discharge signals are extracted by multi-scale open operation of mathematical morphology. Results show that each type of discharge possesses its own pattern characteristic. This conclusion can be applied in the identification of the type of discharge.