Identification of power distribution network topology via voltage correlation analysis

We consider the problem of reconstructing the topology of a portion of the power distribution network, given a dataset of voltage measurements. By using an approximate model for the grid voltage magnitudes, we show that these signals exhibit some specific correlation properties, that can be described via a sparse Markov random field. By specializing the tools available for the identification of graphical models, we propose an algorithm for the reconstruction of the grid topology. Via simulations, we show how the algorithm performs well also when an exact nonlinear model of the grid voltages is adopted, when realistic power demand profiles are considered, and when the voltage measurements are affected by measurement noise.