Closed-loop non-parametric model identification of synchronous generator using NARX polynomials

Abstract

SUMMARYSystem identification can be carried out by perturbing the system input(s) and processing the recorded input (s) and output(s) of the system. In synchronous generators, if the governor is not in action, one of the inputs (the mechanical torque) is not available for measurement, and the experiments cannot be carried out. In the published literature, the input perturbation is mostly carried out through the excitation system, and the mechanical torque is assumed to be constant during the experiment. This would affect the accuracy of the results. In this paper, various multivariable closed-loop identification methods (direct and indirect and linear and nonlinear) are used to obtain an accurate and comprehensive model for the synchronous generator. To simplify model structure, the orthogonal least squares with D-optimality method is used to remove unnecessary terms. A comparison of the performance of various techniques shows that closed-loop nonlinear identification using nonlinear auto regressive with exogenous input polynomials is very effective, and a simple nonlinear model for the synchronous generator can be identified successfully using multivariable closed-loop input and output data.