多变量系统控制输入受限的约束预测控制

Constrained Predictive Control for Multivariable Systems Subject to Input Constraints

针对多变量系统控制输入变量受约束的情况,利用预测控制器在设计过程中能显式地处理约束问题的特点和变量间的相互耦合关联作用,在计算控制律时,首先假设控制输入变量不受约束,计算出控制量,然后把不满足约束条件的控制量设定为边界值,并重新计算那些在约束范围内的控制量,使得原不满足约束条件的控制输入变量的作用得到补偿,该过程可反复进行,直到使得所有的控制量都满足约束条件,从而得到了一种新的约束预测控制算法。该算法物理意义明确,计算量小。仿真研究表明其控制效果优于硬约束限幅控制。

A new algorithm for solving the problem of input constraints in multivariable system is proposed. The algorithm takes advantage of both the feature of explicitly dealing with constraints in predictive control and the coupling effect among variables in multivariable system. In every process of optimizing control input variables, firstly, under the assumption of input unconstraint, the control input variables were calculated, some of which are beyond the constraint bounds, the others are not. Secondly, make the control input variables that don't satisfy constraints equal to the bound value, and then recalculate the rest control input variables. If the recalculated control input variables can't satisfy constraints yet, do with them again in the same way until all control input variables satisfy constraints. Like this, some control input variables which are not beyond the bound value can compensate the effects of the controlled input variables which are beyond the bounds, and thus the optimized constrained control is obtained. The new algorithm offers great flexibility to obtain a desirable output response in the presence of active control input constraints. Due to significant features of low computation, simple structure, high efficiency and good performance, the proposed algorithm appears to be a promising and practical approach to physical control systems subject to input constraints. The simulation results verify the effectiveness of the proposed algorithm. 