基于独立分量分析算法的局部放电信号去噪方法研究

Noise Elimination of PD Signals by Independent Component Analysis

针对局部放电监测信号中多种干扰噪声的消除难题,提出结合经验模态分解的独立分量分析算法,进行变压器局部放电信号的去噪。针对独立分量分析进行信号分离时需要多元信号的要求,首先提出利用经验模态分解算法构造参考信号,然后通过独立分量分析算法进行信号分离。利用该方法进行变压器局部放电信号的去噪,较好地恢复出窄带干扰及白噪声下的局部放电脉冲信号的波形、波形之间相对幅值关系以及脉冲极值所对应的时间点等局放信息,在仿真及实测数据的处理中都取得了较好的效果,验证了该方法的有效性和优越性。

Since the requirement of Independent Component Analysis （ICA） was multi-channels signals, the Empirical Mode Decomposition （EMD） was proposed to construct the multi-channels signals and a new method of eliminating noises in partial discharge signals of power transformer was presented. Using this method in online monitoring of transformer, the partial discharge signals wave and the information of the relative range between the signals and the time point of the max-value can be drawn from narrowband interference and white noise very well. The artificial data and the on-site test results showed that the performance of this method was good.