A Method for Extracting Weak Signals from Power Grid Based on Two-dimensional Singular Value Transformation

Power system fault warning, equipment status monitoring, and data modeling all need to be based on correct data. To improve the accuracy of weak signal detection in the power grid, a non ideal data filtering method based on two-dimensional singular value transformation is proposed. Firstly, a high-dimensional sample matrix is generated based on the oversampling sequence of the power grid signal, and it is subjected to Hankel matrix singular value decomposition. Secondly, based on the mutation property of the relative rate of change of singular values, p-order low rank approximation is performed on the original sample matrix to obtain the filtered Hankel matrix. Finally, the filtered weak signal information is reconstructed by concatenating the first row and last column of the matrix. The experiment shows that the proposed method has a significant filtering effect on weak signals in the power grid under strong noise background. The L2 norm ratio of the extracted signal to the ideal signal under strong random noise background is over 95%, which can be widely used in various scenarios of power system data collection, analysis, modeling and reconstruction.