Reduction of the QP-MPC cascade structure to a single layer MPC

• It is proposed a method to reduce the QP-MPC structure to an equivalent single layer.
• There is no need to specify the inputs and/or outputs optimizing targets.
• The proposed method simplifies the design of the MPC as only one single QP is solved.
• The one layer MPC has a smaller number of tuning parameters than the QP-MPC.

Some commercial MPC packages are implemented in two layers, the QP static layer and the MPC dynamic layer. In the absence of an upper Real Time Optimization layer, the static layer solves a simplified economic optimization problem, which defines optimum feasible targets for the dynamic layer. Since the LP/QP static layer and the MPC dynamic layer are usually executed within the same sampling period, it is not trivial to guarantee that the interaction between the two layers will not disrupt the stability of the whole structure. In this paper, it is proposed an approach to reduce the two-layer structure of some commercial MPCs to a single dynamic layer where the control cost function is extended to include the economic objective. In the proposed approach the convergence and stability of the closed-loop system can be obtained if the economic term of the cost function is properly weighted. A simulation example of a simple industrial system shows the efficiency of the proposed strategy.