Discovering low-rank representations of large-scale power-grid models using Koopman theory

Abstract

The description of coherent features in modern power grids is fundamental in understanding the underlying transient phenomena. While the system dynamics is large-scale and governed by strong nonlinear behavior, an efficient sparse representation can be formulated in a suitable coordinate system. One such representation is given by the Dynamic Mode Decomposition (DMD). In this contribution, we use DMD to obtain low-dimensional reconstructions of power system models from data obtained via a direct numerical simulation or a physical experiment. Notably, we show that DMD can describe the underlying oscillatory swing dynamics captured in data or project the large-scale solution manifold on a system having fewer degrees of freedom.