Research On Noise Suppression In Partial Discharge On-Line Monitoring

The insulation defects in power transformer contact with partial discharge (PD) closely, The PD on-line monitoring can be used to judge the insulation situation of power transformer, and it is valuable to prevent accident faults in power transformer and insure stable performance of power system. One of the most important problems in on-line PD monitoring is how to detect the weak stochastic PD pulses from the strong background noises. There are many methods for the interference suppression, but the real success for the monitoring system is not much, and the result is not satisfactory. With the development of modern digital signal processing technology, the measure of interference suppression in partial discharge monitoring on-line is moving towards the direction of software.In this paper, we investigate the suppression methods of narrowband interference and white noise interference. The frequency-domain filtering technology has been applied for restraining narrowband interference. One second-order IIR notch filter has been designed, according to the simulation, such notch filter could successfully eliminate the narrowband interference. When it comes to white noise interruption, wavelet with characteristics of time-frequency analysis was used to handle the signal and wavelet threshold de-noising method has been applied because of the research which focused upon many different de-noising methods. In order to achieve better noise elimination result, one kind of two-variables threshold function has been put forward which was proved successfully.Improved two-variable threshold function could overcome the shortcomings that the hard-threshold method is not continuous and soft threshold method biased. This method, which changes its parameter k could enable the threshold function to transform in soft and hard threshold function, is used to adjust the threshold of the wavelet coefficients with the original constant deviation between the wavelet coefficients; Parameter m is used to adjust smooth transition curve of the order sub-region for the local fine-tuning. Of simulation data were de-noising, in combination with SNR and NCC, the threshold function is de-noising for improving the effect of a comprehensive evaluation results show that: by adjusting the two variable parameters, the new threshold function is available with the best signal to noise ratio and minimum root mean square error of the de-noising effect.