Parameter Estimation in Three-Phase Power Distribution Networks Using Smart Meter Data

Accurate estimates of network parameters are essential for advanced control and monitoring in power distribution systems. The existing methods for parameter estimation either assume a simple single-phase network model or require widespread installation of micro-phasor measurement units (micro-PMUs), which are cost prohibitive. In this paper, we propose a parameter estimation approach, which considers three-phase series impedance and only leverages readily available smart meter measurements. We first build a physical model based on the linearized three-phase power flow manifold, which links the network parameters with the smart meter measurements. The parameter estimation problem is then formulated as a maximum likelihood estimation (MLE) problem. We prove that the correct network parameters yield the highest likelihood value. A stochastic gradient descent (SGD) method with early stopping is then adopted to solve the MLE problem. Comprehensive numerical tests show that the proposed algorithm improves the accuracy of the network parameters.