Study On The Method Of Calibration And Wavelet Denoising For On-Line Monitoring Partial Discharge In Transformer

On-line PD monitoring of power transformer is Presently the research hotspot in the area of high voltage insulation measurement and diagnosis. Due to the weakness of PD signals and bigness of interferences, the extraction of Pd pulses faces many difficulties. The developed techniques, with some achievement, have not met the demand of engineering application yet.Aiming at the present problems, this thesis studies the PD signal extraction of Pulse Current Method (PCM) in two directions: quantitative measurement and interference suppression.Firstly, theoretic and simulation research work is carried out on resonant nonmagnetic core Rogowski coil current sensor, which is usually used in PD monitoring of transformer. The equivalent circuit of sensor was designed and its transfer function was obtained through theoretically analysis. Using software Saber, the effects of coil number, parallel connection capacitance and loading resistance on the frequency responses of PD signals and the sensitivity was studied; Taking the locale application into account, the safety and reliability of the current sensor under the large power current and lightning pulse current was ensured.The directional coupling measurement, which employed resonant nonmagnetic core current sensor to induce signal from HV bushing, was first presented. The impendence characteristics of substation network are analyzed to build the equivalent circuit, which is used in PD quantitative measurement of transformers.Interference rejection and recognition is necessary to acquire the real PD pulse. Based on the analysis of wavelet decomposition characteristics of PD signals and white noise, the wavelet shrinkage denoising is analyzed in order to decrease the distortion of PD pulse caused by noise removal. Optimal wavelet selection for PD feature extraction is studied. As can be seen from the studied results, the PD S/N ratio is increased and the distortion of PD pulse caused by noise removal is decreased with the level-dependent thresholding. The method is effective for the extraction of PD signals.