MASc Seminar: State Estimation in Power Distribution Systems and its Application to Electricity Theft Detection

Candidate:  Come Agathange Carquex

Title:  State Estimation in Power Distribution Systems and its Application to Electricity Theft Detection

Supervisor:  Rosenberg, Catherine

Abstract:  State estimation in power distribution systems is a key component for increased reliability and optimal system performance. Well understood in transmission systems, state estimation is now an area of active research in distribution networks. While several snapshot-based approaches have been used to solve this problem, few solutions have been proposed in a dynamic framework.

In this thesis, a Past-Aware State Estimation (PASE) method is proposed for distribution systems that takes previous estimates into account to improve the accuracy of the current one, using an Ensemble Kalman Filter. Fewer phasor measurements units (PMU) are needed to achieve the same estimation error target than snapshot-based methods. Contrary to current methods, the proposed solution does not embed power flow equations into the estimator. A theoretical formulation is presented to compute a priori the advantages of the proposed method vis-a-vis the state-of-the-art. The proposed approach is validated considering the 33-bus distribution system and using power consumption traces from real households.

State estimation is then applied to the problem of electricity theft detection. A method taking into account measurements across several time-steps is presented. The proposed method achieves better theft detection rates and lower false positive levels than currents state of the art methods. The limitations of voltage measurements for theft detection are also underlined. The proposed method is validated using the 33-bus distribution system.