基于ICA的变压器局部放电超声信号直达波分离研究

Research on Direct Wave Separation of PD Ultrasonic Signal Based on ICA

针对变压器局部放电超声波检测法存在的定位精度问题,分析了其主要原因在于时延选取的误差,而时延选取误差的主要原因是由于变压器局放超声信号传播规律极其复杂,通过超声传感器接收到的信号是直达波和非直达波以及各种噪声的混合叠加,只有正确分离提取出直达波的时延,才能提高定位精度。为此,本文提出了一种基于独立分量分析(ICA)的变压器局放超声信号直达波分离方法,通过单独分析直达波信号、非直达波信号以及噪声信号各自的主分量特征,运用ICA从混合信号中分离提取出直达波信号并选取时延。仿真和实验结果证明了该方法的可行性。

For Positioning accuracy problems existed in the partial discharge（PD） of transformer ultrasonic detection method,the main reason mostly due to the time delay selected error is analyzed,while the main reason for time delay selected error is due to the transformer PD ultrasonic signal propagation which is extremely complex,and the signals received by the ultrasonic sensor are the direct wave,non-direct wave and a mixture of all kinds of noise superimposed.So in order to improve positioning accuracy the correct separation of the direct wave＇s time delay is important.This paper proposes a method of transformer PD direct wave ultrasonic signal separation based on independent component analysis（ICA）,by a separate analysis of the direct wave signal,non-direct wave signal and noise characteristics of each principal component,with the use of ICA to separate from the mixed signal,extract the direct wave signal and select the time delay.Simulation and experimental results demonstrate the feasibility of the method.