State Estimation and Diagnosis for Electric Power Systems Under Adverse Conditions

Drexel University, Philadelphia PA

Investigators

Abstract

This project addresses power system state estimation and diagnosis as part of a control strategy for power systems that will maintain continuity of service in a hostile environment. The goal is to determine a precise knowledge of the system state, i.e., to establish 'situational awareness.' The 'state' to be estimated is composed of both a continuous system state and a discrete event state. A power system involves discrete events and highly nonlinear continuous dynamics. It is of large scale and uncertain. When a power system operates in a stressed state, i.e., near operating limits (bifurcation points), control system properties like controllability and observability deteriorate. At bifurcation points the system typically fails to be linearly controllable and/or observable. Conventional estimation methods fail, resulting in poor decisions based on inaccurate information Observability of nonlinear dynamics involves aspects that have no counterpart in linear systems. There are entirely new paradigms for observer construction that enable state determination even when the system's linearization is not observable. One consequence of this is that problems of parameter identification and fault detection may be cast in an (nonlinear) observer design setting. These constructs are quite complicated, and are certain to become more so when discrete events are included. The project has three goals: 1) adapt advances in nonlinear system observability/observer theory to power systems, using a nonlinear DAE hybrid system formulation, 2) develop and implement symbolic/numeric computational tools to evaluate observability and to design observers, 3) validate observability tests and new observer designs in a laboratory experiment.