Prony analysis and Robust RLS for electromechanical oscillations: an evaluation technique

摘要

This paper investigates a combination of methods to extract modal information from synchrophasor measurements. Ringdown measurements have been preprocessed with Empirical Mode Decomposition (EMD), analyzed with Prony's method and the results postprocessed with clustering techniques. Ambient measurements before and after a disturbance have been analyzed with Robust Recursive Least Squares, with the same clustering technique applied for continuous evaluation of the estimated modes. Together, ambient and ringdown analysis provide accurate, near real-time awareness of electromechanical oscillations. Postprocessing logic is an issue for parametric methods, as higher model orders often are chosen to improve the estimation process. This paper demonstrates that close inspection of true and trivial modes can reveal important properties of the estimation, and that they can be used to evaluate if the choice of parametric model is justified for the analyzed signal.