Phase Identification in Unobservable Distribution Systems

Abstract

Accurate network and phase connectivity models are crucial to distribution system planning, operations, and situational awareness. In this regard, this paper proposes a novel phase identification approach for unbalanced distribution networks where phases are insufficiently labeled and measurements are limited (i.e., system is unobservable). The proposed approach utilizes the nodal voltage magnitude measurements for clustering the graph representing the phase connectivity information. The proposed approach consists of an optimization formulation that infers the connectivity of the entire multi-phase distribution network in a single shot. Simulation results on IEEE 37, IEEE 123 and EPRI Ckt5 bus test system illustrate the ability of the proposed approach to accurately identify the phase labels under varying availability of data, phase labels errors, and unknown phase connectivity information. The accuracy of the proposed approach is above 90% under these different scenarios and offers gain over 70% over existing state of the art spectral clustering based approaches for phase identification.