模型预测控制器的广义最小方差性能评价

Performance assessment of a model predictive controller based on generalized minimum variance

基于滚动时域最小方差性能评价方法和广义最小方差控制原理,提出了可处理约束问题的滚动时域广义最小方差性能评价方法。该性能评价方法既可考虑过程变量的硬约束,同时相对于滚动时域最小方差性能评价方法,又可考虑操纵变量的软约束。基于Wood-Berry精馏塔模型的仿真实例证明了该性能评价方法较滚动时域最小方差性能评价方法具有更高的实际意义,并验证了这种性能评价方法的有效性。

The performance assessment of model predictive controllers has been attracting wide attention.The benchmark design and constrained processing are essential issues.A moving horizon generalized minimum variance performance assessment approach which can process constraints is derived in this paper,based on a moving horizon minimum variance performance assessment approach and the generalized minimum variance control principle.The performance assessment approach can account for hard constraints on process variables,and,unlike the moving horizon minimum variance performance assessment approach,it can also account for soft constraints on manipulated variables.Simulations based on the Wood-Berry distillation column model show that the performance assessment approach derived in this paper is of greater practical utility than the moving horizon minimum variance approach for a model predictive controller,and the validity of the approach is demonstrated.