结合奇异值分解与最小描述长度准则的变压器极化电流数据去噪方法

Denoising method of transformer polarization current data based on singular value decomposition and minimum description length

现场测量变压器极化电流受噪声干扰造成测量精度偏低,为消除噪声干扰,本文提出一种奇异值分解(SVD)结合最小描述长度准则(MDL)的信号去噪算法。利用测量数据构建Hankel矩阵并进行奇异值分解,将信号分解为有用分量与无用分量的线性叠加,再利用MDL确定信号与噪声的界限,提取有用分量重构信号。对变压器极化电流的仿真和实测数据表明,利用MDL能有效区分有用分量与噪声,去噪数据趋势完整,噪声得到有效去除。与小波硬、软阈值去噪结果对比,信噪比最大可提高12.61dB,方均根误差最大可减小47%。

In order to eliminate the noise interference,a signal denoising algorithm is proposed based on singular value decomposition(SVD)and minimum description length(MDL).The measured data are used to construct Hankel matrix and perform singular value decomposition.The signal is decomposed into a linear superposition of useful components and useless components.The MDL is used to determine the boundary between signal and noise,and the useful components are extracted to reconstruct the signal.The results of simulation and measured data show that MDL can effectively distinguish the useful component from the noise,the trend of denoising data is complete,and the noise is effectively removed.Compared with the results of wavelet hard and soft threshold denoising,the signal to noise ratio(SNR)can be increased by 12.61dB,and the root mean square error(RMSE)can be reduced by 47%.