Influential mechanism of switch between manipulated variables and feed-forward variables on feasible region of model predictive control

Model predictive control is widely used in process control because of its decoupling ability and strong robustness. In the actual industrial process, considering operation requirements, a part of manipulated variables are released by controller and transformed into feed-forward variables at the same time. The transformation between manipulated variables and feed-forward variables results in structure changing among different systems, including square system, fat system and thin system. For the thin system, in which the dimension of manipulated variables is less than that of the controlled variables, the control result is usually less effective than the others. Besides, the partial original manipulated variables, released by controller, transformed into feed-forward variables, also affect the control performance. Feasible region can reflect the effect of the control system directly. On the basis of the multi-step prediction formula derivation for state space model, the method of spatial geometry mapping is used to express the transformation between manipulated variables and feed-forward variables. In addition, it is also used to analyze the influence on the change of feasible region caused by the transformation. Finally, aiming at the situation for releasing partial manipulated variables, prejudgment method is proposed to ensure the stability.