双层预测控制中保证动态控制可行的稳态目标计算策略

A Strategy of Steady-state Target Calculation to Guarantee Feasibility of Dynamic Control in Double-layered Model Predictive Control

在双层结构模型预测控制(Model predictive control,MPC)中,稳态目标计算(Steady-state targets calculation,SSTC)层(上层)为动态控制(Dynamic control,DC)层(下层)提供操作变量、被控变量设定值和变量约束.但是,上层可行域和下层吸引域间存在的不一致性可能使得上层给出的设定值无法实现.本文为下层事先选取若干组放松的软约束,并对每一组软约束都离线计算出相应的吸引域,其中最大的一个吸引域包含稳态目标计算的可行域.在控制过程中,根据当前状态所属吸引域在线地决定在DC层采用的软约束组.采用上述方法后,对所有处于最大吸引域的初始状态,在跟踪稳态目标的过程中,下层优化问题都是可行的.仿真算例证明了该方法的有效性.

In the double-layered model predictive control （MPC）, the steady-state target calculation （SSTC） layer （the upper layer） can provide the setpoints of manipulated variables and controlled variables, and the constraints of variables for the dynamic control （DC） layer （the lower layer）. However, the setpoints given by the upper layer may not be achievable because of the inconsistence between the feasible region of the upper layer and the region of attraction of the lower layer. In this paper, several groups of softened constraints are chosen in advance for the lower layer, and different regions of attraction can be calculated off line for different groups, where the largest region of attraction should contain the feasible region of the SSTC layer. In on-line control, the soft constraints group used in the dynamic control layer is chosen according to the region of attraction where the current state lies. By this method, for every initial state in the maximal region of attraction, the dynamic control optimization problem is always feasible in tracking the steady-state targets. The effectiveness of the method is illustrated through an example.