Noise Suppression and Pulse Extraction Method Based on Fluctuation Characteristics of Partial Discharge Signals

The detection and analysis of partial discharge signal is an important basis and means to evaluate the operation status of high-voltage equipment. In practical engineering, PD signals often have a lot of noise and interference, which is very unfavorable to PD waveform extraction and feature analysis. In this paper, a signal denoising and pulse extraction method is proposed, which uses the fluctuation characteristics of actual signals to eliminate the noise interference superimposed in the fluctuation process. By setting a few thresholds, the start and end coordinates of each pulse can be accurately and effectively extracted. In order to reduce the inconvenience and error caused by manually setting the threshold, this paper uses OTSU and other methods to adaptively calculate the required threshold. In addition, in order to verify the practicability of the above methods, a group of discharge signal pulses are denoised and extracted in this paper. The results show that this method can quickly and accurately separate the PD signal pulse from the noise interference, while maintaining the high-frequency characteristics of the signal to the maximum extent.