Exact Modeling of Non-Gaussian Measurement Uncertainty in Distribution System State Estimation

Abstract

In power systems, state estimation is a widely investigated method to collate field measurements and power flow equations to derive the most-likely state of the observed networks. In the literature, it is commonly assumed that all measurements are characterized by zero-mean Gaussian noise. However, it has been shown that this assumption might be unacceptable, e.g., in the case of the so-called pseudo-measurements. In this paper, a state estimator is presented that can model (pseudo-)measurement uncertainty with any continuous distribution, without approximations. This is possible by reformulating state estimation as a maximum-likelihood estimation-based constrained optimization problem, in a more generic fashion than conventional implementations. To realistically describe distribution networks, three-phase unbalanced power flow equations are used. Trade-offs between accuracy and computational effort of different uncertainty modeling methods are presented using the IEEE European Low Voltage Test Feeder.