中文摘要：

为了提高局部放电在线监测的灵敏度,根据独立成分分析（ICA）基本模型的定义,提出将背景噪声信号与局放脉冲信号看成两个相互独立的信号源,不同观测点获取的观测信号为两者的线性混合,从而可以实现两者的分离的方法。利用ActiveX自动化技术实现了可用于局放脉冲信号提取的ICA软件模块;基于双通道超宽带检测技术,对加噪后的直流下油纸绝缘针板缺陷模型的局放脉冲序列提取进行了试验研究并且利用峭度准则与试验电压极性相结合的方法消除了局放脉冲提取的ICA含混性问题。实验结果表明：该方法能够提取被背景噪声完全淹没的脉冲信号,可以恢复出单个局放脉冲波形、脉冲波形之间的相对幅值关系以及脉冲极值所对应的时间点等重要局部放电信号信息。

英文摘要：

To improve the PD on-line monitoring sensitivity, this paper deals with a new way to extract PD pulse signal from strong background noise with using the Independent Component Analysis （ICA） method which has become an important signal processing and data analysis technique, the typical application is blind source separation in a wide range of signals, such as acoustical, biomedical and astrophysical ones. As the observation signals of PD at different detection points are linear mixture of the background noise and PD pulse signal which are generated by two independent sources, the PD pulse signal can be extracted from the noise according to the basic definition of ICA model. In this paper, the ICA software for processing observation data to extract PD pulse signals is realized with ActiveX automation technology belonging to mixed-language programming techniques. Simulation tests based on two channels by UWB detections technology are done with noised PD pulse sequence in oil-paper insulation of needle-plate model under DC voltage. And the ambiguities elimination method for the basic ICA model to extract PD pulse under DC voltage is also presented, which is made with combination of the polarity of test voltage and the kurtosis criteria. The results show that the PD pulse signal can be exactly extracted from the strong background noise with keeping PD pulse amplitude of the relative relationship, the maximum amplitude corresponding to the time point, and some other important information of the PD pulse sequence.