On the Potential of Constraints in Nonlinear System Identification

Abstract

Local model networks (LMNs) offer a versatile structure for the identification of nonlinear static and dynamic systems. In this contribution the integration of prior knowledge in LMNs using equality constraints is presented. For that purpose, an equality constrained Generalised Total Least Squares (GTLS) algorithm for the local model parameter estimation of the LMN is presented. On the one hand GTLS parameter estimation helps to avoid bias in the estimates when some or all observations are subject to noise and on the other hand equality constraints allow to mathematically enforce desired system properties.