基于变分模态分解和奇异值分解的局部放电信号去噪方法

Partial discharge signal denoising method based on variational mode decomposition and singular value decomposition

为有效抑制局部放电信号中的复杂噪声干扰,提出一种基于变分模态分解和奇异值分解的去噪方法。首先通过泄露能量确定VMD算法中的模态分解个数,对局部放电信号进行分解,得到数个有限带宽的固有模态分量;然后基于峭度指标选择包含有用信息的分量进行信号重构,将周期性窄带干扰去除;最后采用奇异值分解去噪方法抑制信号中剩余的高斯白噪声。运用该方法对含噪局放信号进行去噪处理,并通过时变峰度法对信号初至时刻进行拾取。与传统的小波方法和经验模态分解算法进行对比,该方法能有效抑制局放信号的复杂噪声干扰,去噪后信号波形畸变较小,信号拾取精度较高。

In order to effectively suppress the complex noise interference in partial discharge signals,a denoising method based on variational mode decomposition and singular value decomposition was proposed.Firstly,the number of modal decomposition in VMD algorithm was determined by leakage energy,and partial discharge signals were decomposed to obtain several natural modal components of finite bandwidth.Then,the components containing useful information were selected based on kurtosis index to reconstruct the signal and remove the periodic narrowband interference.Finally,singular value decomposition(SVD)was used to suppress the residual white Gaussian noise.The noise signal was denoised by the proposed method,and the first arrival time of the signal was picked up by the time-varying kurtosis method.Compared with the traditional wavelet method and empirical mode decomposition algorithm,the proposed method could effectively suppress the complex noise interference of local signal,and the signal waveform distortion after denoising was small and the signal pickup accuracy was high.