Study On Digital Method For Transformer PD Denoising Based On Wavelet

On-line partial discharge (PD) monitoring of power transformer is very important to find the fault in transformers and has become a hotspot in research. Because of the strong interference and weak PD signal, it is difficult to obtain the PD pulse onsite for a PD monitoring system. The de-noise technique is a key and difficult issue in PD on-line monitoring. Aiming at this technical problem, the main contents of this paper is presented as follows:1) Reasons of white noise generation are analyzed. How colored noises generate in hardware and digital de-noise process are discussed. Based on the wavelet decomposition of PD signal and white noise, a new method of transformer PD de-noising using the best wavelet for every layer is proposed. It can reduce the PD signal magnitude errors and distortion of the PD pulse waveform. The optimum wavelet is selected by the criticism of the lowest energy on every scale of wavelet coefficient. Different thresholds on every scale are chosen. It is demonstrated that the proposed de-noising method can extract the small partial discharge pulse while the partial discharge pulse waveform is in low distortion.2) The 2^nd order IIR lattice filter is applied to remove the narrow-frequency-band periodical interferences in PD monitoring system. Because the IIR filter may influence the signal waveform, a new de-noising method, integrated IIR filter with the wavelet de-noising method, is presented. It is used to suppress the overlapped narrow-frequency-band noises. De-noising results indicate that the IIR filter order can be reduced effectively and distortion of de-noised PD pulse can be lower.3) Polarity estimation and identification of oscillatory pulse is presented. According to the analysis of signal obtained synchronously from two channels, external pulse-shaped noise can be removed by the polarity comparison. This de-noise approach for removing pulse-shaped interference is flexible to be used and not limited by sensor output signal waveform. The simulation and test results show the method is an effective tool to de-noise external interference.The theoretical analysis, experimental research, and the on-site measured PD signal de-noising examples, are used to qualify our proposed methods.