Application of Constrained K-Means Algorithm for Phase Identification

This paper presents the application and results of the constrained k-means algorithm to the phase identification problem in Horizon Power's Carnarvon network. Voltage measurements from Advanced Metering Infrastructure meters are used in combination with principal component analysis in the application of the constrained k-means clustering algorithm. Constraints ensure that each phase of a three-phase meter is assigned to a distinct network phase in the clustering. The algorithm is applied to the whole of Carnarvon, however, a more detailed analysis is carried out on one section of the network that forms part of the Carnarvon Distributed Energy Resources Trials.