A real-time distributed Prony-based algorithm for modal estimation of power system oscillations

In this paper, we present a set of distributed algorithms to estimate the oscillatory electro-mechanical eigenvalues of large power system networks using real-time measurements of phase angles and frequencies. We consider the underlying network to exhibit a strong coherency structure resulting in a two-time-scale behavior of the states. Our goal is to develop a distributed strategy by which only the slow eigenvalues, or the inter-area oscillation modes, can be estimated. Assuming a reliable bound for the slow frequencies, we first pass the actual measurements from the full-order network through a bandpass filter, and use the filter outputs as the effective inputs to our distributed estimation routine. We integrate the centralized Prony method for modal extraction with two commonly-used distributed optimization algorithms, namely, distributed subgradient method (DSM), and alternating direction method of multipliers (ADMM), and show how the filtered outputs can be used in each case to asymptotically estimate the slow eigenvalues of interest. We illustrate our results using a IEEE 39-bus power system, and show the robustness of our methods in face of communication failures.