A variable dimension Gauss-Newton method for ill-conditioned parameter estimation with application to a synchronous generator.

摘要

This thesis presents a variable dimension Gauss-Newton (VDGN) parameter estimation algorithm that can be used for fault detection, and diagnosis of a synchronous generator. The algorithm is derived as an extension of the variable dimension Newton Raphson algorithm proposed to solve nonlinear systems of equations. We study the conditioning of the parameter estimation problem for a linearized small-signal model of the synchronous generator using local sensitivity analysis. The conditioning analysis is performed on simulated data and experimental data for the FC5HP synchronous generator located at the Four Corners Generating Station of the Arizona Public Service Company (APS), rated at 483 MVA. Results demonstrated that local sensitivity analysis is an effective tool to diagnose ill-conditioning. The developed VDGN algorithm is shown to be a robust method for ill-conditioned parameter estimation and its performance is compared with the subset selection method. Results using experimental and real data for the synchronous machine parameter estimation problem showed that the VDGN algorithm computes better parameter estimates than the subset selection method and requires less prior information to deal with the ill-conditioning.