Effective algorithms of optimizing predictive control with neural and fuzzy models of nonlinear processes

Aim of project
The goal of the research are numerically effective algorithms of optimizing predictive control using neural and fuzzy models of nonlinear processes. Model Predictive Control (MPC) technology is widely applied in industrial applications, because it enables to take directly into account constraints of input and output signals. Moreover, due to direct use of a process dynamical model, MPC algorithms may be naturally applied to multi-input multi-output systems, to processes with difficult dynamics. There is now a tendency to integrate MPC feedback control with on-line set-point optimization. Then, the use of nonlinear models both to dynamic feedback control as weel as to set-point optimization is needed. Due to certain advantages, the use of neural and fuzzy nonlinear models in the considered algorithms is proposed. There are three key points here: numerical effectiveness, robust stability and fault-tolerance. These problems are central for the proposed research.

Expected results
The proposed research should result in new structures and algorithms of multi-purpose predictive control and appropriate design methods, documented in technical reports and described in publications. Particularly, feedback control with set-point optomization and reconfigurable control will be considered. Results of the analysis, both simulative and theoretical, will be delivered and presented on scientific conferences and published in journals.