Distribution Networks Topology Modeling Based on Data of Smart Meters

This paper proposes a methodology for radial medium-voltage distribution networks topology modeling with the voltage-power sensitivity based on smart meters data. Having different characteristic, partial least squares regression (PLS) and deep neural network (DNN) are respectively used to obtain the sensitivity. The relationship between the sensitivity and the topological position is analyzed, based on which, the topology is identified. The impendence of each line is estimated accordingly by the voltage value and the sensitivity obtained by DNN. The methods presented in this paper do not necessitate any prior topological information or additional data apart from voltage and power readings collected by smart meters. Their efficacy is initially evaluated using the IEEE-33 test case, followed by assessment across various states of the IEEE-69, demonstrating remarkable efficiency and accuracy.