RBF-ARX Modeling and Predictive Control Strategy Applied to a Liquid Level System

摘要

The main objective of this paper is to show in the first place that the RBF-ARX modeling technique can be used to model a dynamic nonlinear SISO liquid level system with higher precision and then to demonstrate that when the model obtained is taken as predictor of a model predictive controller (MPC) one may obtain an enhanced control performance. The RBF-ARX model is in fact a locally expanded Taylor ARX model with Gaussian Radial Basis Function (RBF) network-style coefficients depending of the working point; it can be estimated off line to avoid any online uncertainty. It is built to globally describe the behavior of nonlinear dynamic system and exhibit an easy and advantageous means of obtaining a local linearization of any working point. The RBF-ARX model based MPC (RBF-ARX-MPC) is a predictive control strategy based on RBF-ARX model. It doesn''t require online but offline parameters optimization in which the nonlinear parameters estimation depends on the Levenberg-Marquardt Method (LMM) and the linear one on the Least-Square Method using Singular Value Decomposition (SVD).