Blind topology identification for power systems

Abstract

In this paper, the blind topology identification problem for power systems only using power injection data at each bus is considered. As metering becomes widespread in the smart grid, a natural question arising is how much information about the underlying infrastructure can be inferred from such data. The identifiability of the grid topology is studied, and an efficient learning algorithm to estimate the grid Laplacian matrix (i.e., the graph equivalent of the grid admittance matrix) is proposed. Finally, the performance of our algorithm for the IEEE-14 bus system is demonstrated, and the consistency of the recovered graph with the true graph associated with the underlying power grid is shown in simulations.