Identification of Nonlinear Systems Using Parameter Estimation Techniques

摘要

The identification of nonlinear systems using parameter estimation methods based on input/output difference equation models is considered. A general representation of a wide class of nonlinear systems is derived by considering an observability condition. The Hammuerstein, Wiener, bilinear and other well known models are shown to be special cases of the nonlinear model. The effects of internal noise are investigated and the estimation of the coefficients using linear estimation algorithms is shown to yield biased estimates. A modified extended least squares algorithm is presented and structure detection methods and model validity checks are briefly discussed.