On-line squaring of non-square hard constraints of input variable by coordinate alternating in model predictive control

Commonly there exit many square amplitude constraints in process control. These constraints are high and low limits of input variable, input variable variation and output variable. For the requirement of process or control, there may be a few non-square hard constraints of input variable in addition. These non-square constraints are composed of the amplitude limits of linear function of input variables, and they cannot be solved by MPC directly. To ensure the implementation of MPC, a method to transform non-square constraint into square constraint on-line by coordinate alternating is proposed. By using the previous step input value and the high limit of input variable variation, some appropriate treatments of these non-square constraints are made. Simulation results of two system control problems show the effectiveness of the proposed method.