史坦无偏估计自适应奇异值分解在局放信号白噪声抑制中的应用  

Application of Adaptive Singular Value Decomposition Based on Stein Unbiased Risk Estimation in Partial Discharge Signal White Noise Suppression

作　　者：谢敏[1] 周凯[1] 何珉 陈泽龙[1] 黄永禄 赵威 XIE Min;ZHOU Kai;HE Min;CHEN Zelong;HUANG Yonglu;ZHAO Wei(School of Electrical Engineering and Information,Sichuan University,Chengdu610065,Sichuan Province,China;Electric Power Research Institute of State Grid Chongqing Electric Power Company,Yubei District,Chongqing401123,China;Kunming Power Supply Bureau of Yunnan Electric Power Grid Co.,Ltd.,Kunming650200,Yunnan Province,China)

机构地区：[1]四川大学电气信息学院,四川省成都市610065 [2]国网重庆市电力公司电力科学研究院,重庆市渝北区401123 [3]云南电网有限责任公司昆明供电局,云南省昆明市650200

出　　处：《电网技术》2018年第12期4153-4159,共7页Power System Technology

基　　金：中国博士后科学基金项目(2015T80976)~~

摘　　要：局部放电(简称局放)噪声抑制是电力电缆局放检测中的重要步骤之一。为有效保留局放信号的细节,提出了一种基于史坦无偏估计的完全无监督奇异值分解局放去噪方法,为准确快速地获取最优阈值,引入差分进化算法进行迭代搜索。对实验室实测和现场实测局放信号进行去噪处理,并将去噪结果与标准软阈值去噪方法和基于能量最大化的小波去噪方法进行对比。结果表明:相比于标准软阈值去噪方法和基于能量最大化的小波去噪方法,文中提出的去噪方法的去噪结果更好,即使在强噪声背景下也能有效恢复原始局放信号的细节,具有良好的应用价值。Partial discharge(PD)noise suppression is one of the main tasks in PD detection of power cable.To preserve more features of PD signals,a totally unsupervised de-noising method based on singular value decomposition(SVD)and Stein unbiased risk estimate(SURE)was proposed.To obtain quickly and accurately the best threshold of singular values, differential evolution algorithm was adopted.The proposed method was examined on laboratory-obtained and field-detected PD signals under different signal-to-noise ratio(SNR), and its results were compared with existing discrete wavelet transform(DWT)de-noising methods,standardsoft thresholding method(STM)and energy based wavelet selection (EBWS).Results show that the proposed method has better performance than STM and EBWS,and the features of original PD signals can be effectively restored even under high noise level,showing good application value in practical PD detection.

关 键 词：电力电缆 局部放电 噪声抑制 史坦无偏估计 差分进化算法 

分 类 号：TM86[电气工程—高电压与绝缘技术]