CAREER: Open-Loop Discrete-Event Control in Electric Power Systems

Indiana University, Bloomington IN

Investigators

Abstract

9983653 Rovnyak The goal of this research is to develop and validate pattern recognition methods such as Decision Trees (DTs) and Neural Networks (NNs) for electric power system wide-area open-loop discrete-event control. The proposed controls will provide Independent Service Operators (ISOs) and present-day utilities with tools necessary to achieve the cost reduction goals of deregulation without sacrificing reliability. The proposed research will investigate the selection of contingencies to simulate for the training data, algorithms for converting simulation data into input- output pairs that make up the training sets, and measures of performance used to evaluate the classifier operation. The PI and graduate students will collaborate with researchers at Cornell University and the Bonneville Power Administration and will test the proposed methods against other schemes reported in the literature for out-of-step relaying, generator tripping, active load modulation, HVDC fast power changes, and reactive power switching. Students in EE 588, Advanced Topics in Power Systems, will train and test NNs on data from the research. Undergraduate students in EE 479, Automatic Control Systems Laboratory, will design feedback controls for the simple power system model that is analyzed in this proposal. The PI will engage all students in active learning through frequent and extensive exercises that lead students through the discovery process of problem solving instead of just rewarding the memorization of techniques. The PI will distribute detailed, annotated solutions to all homework problems on the due date. These teaching methods have been found by the PI to be successful with a broad spectrum of students, including women, minorities and people with disabilities. ***