Research On Principle And Implementation Method Of Interference Suppression In Partial Discharge On-line Monitoring In Excitation Transformers Of Power Plants

Excitation system of generator provides excitation current for generator, which is the key link to ensure stable operation of generator. Excitation transformer is the device which provides three-phase AC excitation power for large generator excitation system. Its insulation is the combined effect of AC and pulse voltage, so electrical aging process is very complicated. On-line PD (partial discharge) monitoring has been widely used to assess the insulation condition of electrical equipment, On-line PD monitoring of excitation transformer is an important means to ensure safe and stable operation of excitation transformer. Since the PD signal inside the excitation transformer is very weak and it is disturbed by field electromagnetic interference. How to extract partial discharge signal effectively from field interferences is still a problem for on-line PD monitoring of excitation transformer.This paper studies the principles and methods of interference suppression in partial discharge on-line monitoring.Firstly, a basic method of on-line PD monitoring for excitation transformer is introduced, and the basic principle of pulse current sensor for excitation transformer partial discharge is analyzed. Then the wave's characters of PD signals inside excitation transformer are analyzed by discharge correction experiments, and the wave's characters of external discharge pulse interference for excitation transformer is analyzed based on field measured signals. Finally, PD signals of excitation transformer PD monitoring are presented and the mathematical model of external discharge pulse interference signals is putted forward.Secondly, according to suppression of narrow-band interference in Partial discharge signal, a new method for suppressing narrow-band interference is proposed by using the sensitivity of Duffing chaos oscillator to periodic signal. Duffing oscillator model is presented, which inhibit mixed narrow-band interference for multiple frequency. Comprehensive discrimination method of chaos oscillator motion state is presented by the combination of Lyapunov exponent and phase diagram.Maximum-variance Algorithm of system output signal is proposed to ensure narrow-band interference frequency. The analysis of simulation experiment indicates that the method can monitor effectively frequency and amplitude of narrow-band interference among the PD signals because of the immunity of Duffing oscillator model to white noise and PD signals, at the same time, the method can ensure the distortion of PD signals is smaller. When the frequency of narrow-band interference and the center frequency of PD signal are identical, inhibitory effect of narrow-band interference in chaos oscillator system will not be affected.Thirdly, around the removal of partial discharge signals in white noise interference problem, a new de-noising method based on wavelet transform is proposed——block thresholding spatial combined de-noising method. First this method use wavelet thresholding de-noising method to preprocess the sub-band signal of wavelet decomposition of noise, in order to remove part of the pre-noise, and reduce the number of iterations of the later spatial correlation method significantly. Then spatial correlation method for data processing is executed, and gets the size comparison of correlation coefficients modulus and the wavelet transform modulus at different scales directly. And then set wavelet coefficients which have a smaller correlation coefficient modulus to zero. Finally, reconstruct the results and return it to the original signal. In order to select the optimal parameters of de-noising, the impact of decomposition level, noise intensity and iterations of spatial correlation method on de-noising effect is discussed by the analysis of simulation experiment.And then, around the removal of external discharge pulse interference for partial discharge signals, a new pulse extraction method for partial discharge signals is introduced. Multistage energy parameters of wavelet packet coefficients, which are used as characteristic quantity, are extracted based on wavelet packet decomposition. The discharge mode is identified by using binary tree support vector machine. The partial discharge signal and external discharge pulse interference signals are distinguished effectively.According to PD signals of field monitoring for excitation transformer, single pulse waveform extracted from the sample sequence have been indentified and de-noised. The results show that the binary tree support vector machine can identify external discharge pulse interference, and remove it from the sample sequence of PD signal.Finally, the analysis results of field monitoring signal indicate that layered de-noising integration method adopting the above interference suppression method can be used in PD monitoring of excitation transformer effectively, and greatly improve the monitoring level of system.