介损在线监测去噪方法的有效性  

Validity of Noise Reduction Method in Dielectric Loss On-line Monitoring

作　　者：徐志钮[1] 

机构地区：[1]华北电力大学新能源电力系统国家重点实验室,保定071003

出　　处：《高电压技术》2015年第3期909-917,共9页High Voltage Engineering

基　　金：国家重点基础研究发展计划(973计划)(2009CB724508)~~

摘　　要：为了减少白噪声给介损在线监测带来的误差,比较常规离散Fourier算法与加窗插值算法的准确性,分析了信号点数及其使用方式对计算所得介损误差的影响,分析了小波分解层数、阈值获得准则和阈值模式对介损计算准确性的影响,获得了不同类型的小波方法的去噪效果。结果表明:受噪声干扰明显时应采用常规离散Fourier算法。随信号点数的增加误差呈幂规律减少,采样时应该选择更高的采样频率。叠加平均法可有效提高介损测量的准确性;计算时仅用长信号计算一次即可,而不应该采用长信号的分段或滑动窗计算计算。含有白噪声干扰时小波方法虽可对时域信号去噪但该方法对提高介损测量的准确性无效。同时用实验室实测含噪信号的计算验证了分析结果。To decrease the error caused by white noise in dielectric loss on-line monitoring, the conventional DFT algorithm with the windowing and interpolation DFT algorithm was compared. Moreover, the influences of signal length and its usage mode, the wavelet decomposition level, threshold selection rule and threshold mode on the accuracy of dielectric loss were investigated. The noisy voltage and current signals were denoised by different wavelet methods and superposition averaging method. The results reveal that if the signal has significant noise, the conventional DFT algorithm should be selected to calculate the dielectric loss. The error decreases with the signal length, and it obey a power law. The superposition averaging method can improve the accuracy, the method makes full use of all signals at a time of calculation rather than using segmentation calculation or sliding window calculation. Although the wavelet method can suppress the noise in the noisy signals, the accuracy of the dielectric loss calculation cannot be improved. The above conclusions are validated by the calculation results of the experimental data.

关 键 词：电容型设备 介损 在线监测 白噪声 小波变换 叠加平均 FOURIER变换 

分 类 号：TM855[电气工程—高电压与绝缘技术]