Voltage-based Machine Learning Algorithm for Distribution of End-users Consumption Among the Phases

Distribution networks are poorly observable, which is especially evident in different analyses of low voltage (LV) networks, where observability is decreased by the reduced number of smart meters and the lack of network data. Smart meters are in most cases used only for measuring consumption data, while other important information, such as the phase connection of end-users, is not adequately monitored. This aggravates the problem of phase identification for energy utilities, which consequently complicates the numerous calculations required for the smooth operation of the distribution network. In this paper, a comparison of voltage and consumption measurements-based phase identification is presented. Furthermore, a machine learning model based on the voltage measurements is extended to correctly identify the phases of end-users which are three-phase connected to an LV network. The model is tested on a simple 18-node network and the IEEE benchmark network with over 100 nodes and more than 50 end-users. Even though the results show a possibility of using both methods in simpler cases, the voltage measurement-based method is more robust and leads to smaller error in the phase detection problem but also can be extended and used in the case of three-phase connected end-users.