A tensor decomposition‐based approach for the analysis and visualization of ambient power system oscillations

Abstract

In recent years, more and more data analysis tools have been developed to analyze and visualize ambient power system oscillations that are recorded from modern measurement systems. In this paper, a tensor decompositionbased approach is proposed for characterizing and visualizing the dynamic information contained in the ambient data. The approach exploits the correlation of the data through several covariance matrices that are used to form a third-order tensor. From this tensor, the autocovariance signal corresponding to each modal response and its modal amplitude vector (pseudo-mode shape) can be extracted by the parallel factor decomposition. Subsequently, a matrix pencil algorithm is applied to each autocovariance signal to estimate its natural frequency and damping ratio. The proposed methodology was applied to ambient data obtained from a transient stability model of the New England test system/New York power system, and the effect of various method parameters on the quality of the modal identification was analyzed. The results indicated that the oscillatory frequencies, damping ratios, and pseudo-mode shapes can be accurately estimated using the proposed method. Therefore, it can be concluded that the proposed approach is promising in terms of monitoring the ambient oscillations using measured data.