Blind estimation of states and topology (BEST) in power systems
Idan Gera, Yair Yakoby, Tirza Routtenberg · 2017
In this paper, we consider the problem of state estimation and topology identification in power systems. We assume the DC model of real power measurements with unknown voltage phases and an unknown admittance matrix. We show that this problem is equivalent to the blind source separation (BSS) problem, where the mixing matrix is a weighted Laplacian matrix. We propose two new Blind Estimation of States and Topology (BEST) methods for this problem. The first method, Cov-BEST, is based on utilizing the states' second-order statistics and the positive-definiteness of the reduced Laplacian matrix. The second method, Generalized Laplacian Separation (GLS)-BEST, is obtained by applying any general BSS method, followed by an approach that resolves the inherent BSS ambiguities by utilizing the Laplacian matrix properties. In contrast to existing methods, the proposed methods achieve full recovery of the topology matrix and are not limited to matrix eigenvectors estimation. The performance of the proposed methods is evaluated for a general network with an arbitrary number of buses and for the IEEE-14 bus system, and compared with the oracle state estimator.