Nonlinear System Modeling and Robust Predictive Control Based on RBF-ARX Model

Hui Peng,Zi-Jiang Yang,Weihua Gui,Min Wu,Hideo Shioya,Kazushi Nakano
DOI: https://doi.org/10.1016/j.engappai.2006.05.012
2007-01-01
IFAC Proceedings Volumes
Abstract:The multi-input multi-output (MIMO) RBF-ARX model is presented for describing the dynamics of the multivariable nonlinear systems whose working-points vary with time and which may be linearized around working-point. The MIMO RBF-ARX model has a basic linear ARX model-type structure, and its regression matrices’ elements are composed of Gaussian radial basis function (RBF) neural networks that are dependent on current system's working-point state. An offline estimation approach to the MIMO RBF-ARX model is given, and based on the estimated MIMO RBF-ARX model a predictive control strategy is designed to control such nonlinear system. Case study on a simulator of thermal power plant is also given to demonstrate the effectiveness of the modelling and control approaches proposed in this paper.
What problem does this paper attempt to address?