Strategies with guarantee of stability for the integration of model predictive control and real time optimization

Author(s)

Luz Adriana Alvarez Toro

• Document type: Doctoral Thesis
• Press: São Paulo.
• Institution: Universidade de São Paulo (USP). Escola Politécnica (EP/BC)
• Defense date: 2012-03-23

Examining board members

Darci Odloak; Claudio Garcia; Luís Cláudio Oliveira Lopes; Oscar Alberto Zanabria Sotomayor; Antônio Carlos Zanin

• Advisor: Darci Odloak

Abstract

The aim of this Thesis is the development of predictive controllers (MPC) with guarantee of stability and that are part of a control structure where Real Time Optimization (RTO) is present and produces optimizing targets for the predictive controller. The approaches of two and three-layer are considered. Three different strategies are presented: the first strategy is developed for integrating systems; it consists on an infinite horizon MPC algorithm in two extended versions. This controller is designed to be implemented in the two-layer structure. Simulation results in a linear system with integrating and stable modes show the ability of the MPC controller to follow input targets from the RTO layer even when there are targets for integrating systems. As a second strategy, it is developed an algorithm that guarantees nominal stability of the MPC controller when it interacts with the intermediary layer of the three-layer structure. In order to produce a robust structure, an extension to uncertain systems is also developed. This approach is tested with both a linear and a nonlinear system. For the nonlinear system, which is an industrial process, the full structure that includes the RTO with the robust algorithm is simulated. Results show that the structure is capable to follow target changes when disturbances affect the process. Finally, the last strategy proposed in this Thesis consists on the inclusion of a convex function in the MPC controller to follow the RTO targets. The gradient of this convex function is considered in a quadratic term in the objective function of the MPC controller. This controller is also simulated with both a linear system and a nonlinear system. For the MPC with gradient, an extension to the case of uncertain systems is developed in order to provide a more robust controller. For the strategies developed in this Thesis, theorems that guarantee the recursive feasibility and the convergence of the closed-loop system are provided. (AU)