A method to guarantee the feasibility of dynamic controller in double-layered model predictive control

The double-layered model predictive control (M-PC), including a steady-state target calculation (SSTC) layer (upper layer) and a dynamic control layer (lower layer), is broadly used in the process industries. However, because of the inconsistence between the admissible set in upper layer and the domain of attraction in the lower layer, the steady-state targets given in the upper layer can not be reached. In this paper, the following measures are adopted: (i) calculate a series of maximal invariant sets and domains of attraction centering at the steady-state targets by relaxing soft constraints step by step, (ii) choose a suitable domain of attraction which contains the current state and the corresponding soft constraints adopted in the lower layer. Through this measures, the feasibility of the dynamic control optimization problem is guaranteed during the whole tracking process. The effectiveness of the approach is illustrated through an example.