局部放电宽带检测系统分类性能的改善方法

Methods to Improve Classification Performance of Wide-Band Detection System for Partial Discharge Pulse Signals

用脉冲波形的宽带检测及分类技术解决常规局部放电检测系统难以处理的多局部放电源和抑制随机干扰脉冲信号问题.针对在局部放电检测中的某些特殊工况,给出了提高检测系统分类性能的方法——基于波形峰值-时间序列的阈值判别法、基于小波软阈值的脉冲波形去噪法以及基于脉冲波形二维特征平面的极限坐标分类法.局部放电试验结果表明,提出的方法在直流下可提高宽带检测系统的分类性能.

Wide-band detection and classification techniques based on pulse waveshapes were developed to detect and classify muhi-PD （partial discharge） sources and suppress random interference signals, with which it is difficult for the conventional PD measurement systems to deal. Approaches, including threshold discrimination based on peak-time sequence of pulses, wavelet soft threshold denoising for pulse waveshapes and marginal coordinate classification based on 2D characteristic plane of pulse sequences, were presented to improve the classification performances of the detection and classification system for some special cases. PD experiments under DC voltage show that these approaches effectively improve the classification performances of the PD detection system.