Non-linear robust identification: Application to a thermal process

Abstract

In this article, a methodology to obtain the Feasible Parameter Set (FPS) and a nominal model in a non-linear robust identification problem is presented. Several norms are taken into account simultaneously to define the FPS which improves the model quality but, as counterpart, it increases the optimization problem complexity. To determine the FFiS a multimodal optimization problem with an infinite number of minima, which constitute the FPS, is presented and a special evolutionary algorithm (ε-GA) is used to characterize it. Finally, an application to a thermal process identification, where ∥ · ∥∞ and ∥ · ∥1 norms have been considered simultaneously, is presented to illustrate the technique. © Springer-Verlag Berlin Heidelberg 2007.