Power System Dynamic State Estimation Using a Layered Local-Global Method

Fast state estimation and system identification are essential to the reliability of the increasingly complex power system. Current estimation, on slower, steady-state, time scales does not model dynamics; however, with the deployment of phasor measurement units, faster estimation is now possible. We introduce a robust state estimator that learns the dynamic state of the power system on fast time scales. This estimator uses a modified Extended Kalman Filter based on an implicit trapezoidal discretization for increased stability when estimating the states of the differential algebraic system that describes the power system. Further, it combines a local and a global estimator into a hybrid architecture that effectively handles unknown changes in the system dynamics, such as faults. The overall goal of this work is to develop new state estimation techniques that will support a future fast and flexible estimation layered architecture that integrates state estimation, change point detection, and classification of disturbances.