Using Spectral Graph Wavelets to Analyze Large Power System Oscillation Modes

Abstract

This paper presents a novel method for modal analysis to extract the spatial-temporal characteristics of oscillations in large electrical networks. A vector-fitted approximation of the Spectral Graph Wavelet Transformation (SGWT) and the inverse SGWT are derived to identify intra-network oscillations within a system response. This method scales linearly with the number of branches and leverages sparse solution techniques to develop a fast, low-memory estimation of modal frequency, shape, and damping. A case study on synthetic networks (2k-80k buses) with full dynamic modeling demonstrates consistent sub-second performance of modal estimation. Compared to existing methods, the SGWT approach can estimate modes with fewer channels and a shorter time-domain window. This presents a fast, general method for identifying true multiscale network behavior and localized oscillation sources, marking a novel application of graph-based signal processing.