一种GIS在线监测中UHF局部放电信号新型去噪方法研究

摘要

超高频法是目前广泛使用的局部放电在线监测手段之一,虽然能够有效避免现场大量低频噪声的干扰,但是仪器热噪声、无线电通讯等因素仍然会对超高频信号产生干扰.小波阈值法是一种有效的去噪方法,核心在于阈值的有效选取,从信号能量的角度出发,利用奇异谱分析对小波分解后各尺度小波系数噪声成分的标准差进行估计,进一步通过Donoho广义阈值法求取各尺度阈值.通过模拟局部放电实验获取含噪信号,并利用所提方法去噪,从噪声抑制比的角度考虑,与传统的Donoho阈值法对比,具有更好的去噪效果.%Although UHF,one of the widely used PD online monitoring means,can effectively avoid interference from large amounts of on-spot low-frequency noise,such factors as instrument thermal noise and radio communication will cause interference to UHF signals.Wavelet threshold as an effective de-noising means has effective selection of the threshold value as its core.From the perspective of signal energy,we use singular spectrum analysis to estimate the standard deviation of the noise components in the coefficients of various decomposed wavelets.Then,corresponding threshold values are obtained by means of the universal threshold approach proposed by Donoho.Signals with associated noise are acquired through simulation of PD experiment,and de-noising is completed in the proposed way.From the viewpoint of noise suppression ratio,the proposed algorithm can produce better de-noising effect than traditional Donoho threshold value method.