Multi-Way Partitioning of Power Networks via Koopman Mode Analysis

Abstract

This paper deals with the partitioning problem of power networks by applying the so-called Koopman Mode Analysis (KMA) to sampled voltage-angle data. KMA is based on the so-called Koopman operator defined for arbitrary nonlinear dynamical system and provides an infinite dimensional but linear description of the evolving nonlinear dynamics. By computing a set of dynamically relevant modes using the KMA, a new partitioning method for power networks is introduced, and multiple partitions are derived for a benchmark system. The obtained result is compared with two conventional methods previously applied to the partitioning problem; spectral graph theory and slow coherency.