No abstract

No abstract

Summary This note presents a stochastic formulation of the model predictive control for tracking (MPCT), based on the results of the work of Lorenzen et al. The proposed controller ensures constraints satisfaction in probability, and maintains the main features of the MPCT, that are feasibility for any changing setpoints and enlarged domain of attraction, even larger than the one delivered by Lorenzen et al, thanks to the use of artificial references and relaxed terminal constraints. The asymptotic stability (in probability) of the minimal robust positively invariant set centered on the desired setpoint is guaranteed. Simulations on a DC‐DC converter show the benefits and the properties of the proposal.

No abstract

The success of Model Predictive Control (MPC) strategies in industrial and academic environments in the last decades has been remarkable. However, there are many open questions in the area, especially when the simplifying hypothesis of perfect model is dropped. The explicit consideration of uncertainties lead to important progresses in the area of robust control, but it still exhibits a few drawbacks: high computational load and over conservative behavior are issues that may have hindered the application of robust strategies in practice. The approach of Stochastic Model Predictive Control (SMPC) aims at the reduction of conservativeness due to the incorporation of statistical information about noise. Since processes in chemical industry are always subject to disturbances, resulting from model-plant mismatch or from unmeasured disturbances, this technique is an interesting alternative for future control algorithms. The main objective of this thesis is the development of SMPC algorithms that take into account some of the specificities of such processes, which have not been adequately handled in the literature so far. The most important contribution is the inclusion of integral action in the controller through a velocity description of the model. Besides, hard input constraintsassociated with safety or physical limitsand probabilistic state constraintsusually derived from product specification-are also included in the formulation. Two approaches were followed in this work, the first is more direct and the second provides closed-loop stability guarantee at the price of increased conservativeness. Another interesting feature that is developed in this thesis is the zone control of systems subject to disturbances. This form of control is often present in industrial arrays due to the lack of degrees of freedom, and the proposed approach is the first to merge zone control and SMPC. Different simulations of all proposed controllers and comparison to literature benchmarks are provided to show the application potential of the developed techniques.