改进经验模态分解阈值算法抑制强窄带干扰

Suppressing of PD Signal's Strong Narrow-band Interference Based on Empirical Mode Decomposition

为抑制电力设备局部放电在线监测中出现的较强窄带干扰,提出改进经验模态分解(EMD)阈值算法,改进了频域阈值算法并解决了其阈值和干扰频带难以确定的问题。首先在频域处理时设置一个较大阈值来降低窄带干扰幅值,接着EMD以得到含有特征频率的固有模态函数(IMF),然后对IMF分量进行阈值处理,利用窄带干扰和局部放电信号在IMF分量上的特性差异来抑制窄带干扰。

To suppress strong narrow-hand interference in on-line partial discharge （PD） monitoring system of electric power apparatus, this paper investigates the threshold algorithm for Empirical Mode Decomposition （EMD）, which improves the threshold value method of frequency domain and effectively resolves the problem that threshold value and interference frequency are difficult to determine in frequency domain analysis. First, the mid-signals can be produced by reducing the amplitude of narrow-band interference in frequency region, which is decomposed with EMD, then intrinsic mode functions （IMF） which contain specific frequency can be obtained, then for every IMF, the effective threshold method for suppressing the periodic narrow-band interference is used. Furthermore, the interference is suppressed according to the characteristic difference of narrow-band interference and partial discharge signal in every IMF.