Abstract: Rolling optimization is the core of predictive control. A powerful optimization algorithm——tabu search algorithm was discussed, and was applied to rolling optimization. The objective function of rolling optimization is taken as taboo subject, a value range of control is divided into several super rectangular spaces, two-major-criteria is used to judge whether each point is tabooed, and the corresponding special pardon rule and termination principle were proposed, which makes the rolling optimization transformed into unconstrained optimization problems. Finally, its effectiveness was confirmed through simulating triangular wave and random signals.

Key words: Model predictive control, Tabu search algorithm, Rolling optimization

摘要： 滚动优化是预测控制的核心，探讨了一种寻优能力强的算法——禁忌搜索算法，并将其用于滚动优化中。它将滚动优化中的目标函数值作为禁忌对象，将控制量取值范围进行邻域划分为几个超矩形空间，使用两重准则来判断各点是否被禁忌，并提出相应的特赦准则和终止原则，使滚动优化转化为一般无约束的优化问题。最后通过对三角波和随机信号的跟踪仿真验证了其有效性。

关键词: 预测控制, 禁忌搜索算法, 滚动优化