基于在线优化的鲁棒模型预测控制  

Robust Model Predictive Control Based on Online Optimization

作　　者：杨世忠[1,2] 任庆昌[1] 

机构地区：[1]西安建筑科技大学土木工程学院,陕西西安710055 [2]青岛理工大学自动化工程学院,山东青岛266520

出　　处：《信息与控制》2013年第6期742-749,共8页Information and Control

基　　金：住房和城乡建设部科学技术项目(2012-K1-35);陕西省教育厅自然科学专项基金资助项目(11JK0906)

摘　　要：基于线性矩阵不等式的鲁棒模型预测算法分为在线和离线两个部分,为了提高系统响应速度和控制精度,将离线算法求得目标函数的上界作为已知量,重新优化得到一系列较大渐近稳定的不变椭圆集.在线时,根据当前测量得到的状态变量,可以确定该变量位于椭圆集序列中的2个椭圆集之间.通过增加一个相邻的小的椭圆集,用3个椭圆集优化对状态变量进行精确定位,并得到系统控制量.给出了在线优化的理论证明.通过与传统算法的仿真比较表明,该算法比传统算法控制效果更好.以变风量中央空调中风机为例进行控制,实验结果验证了所提出算法的有效性.The robust model prediction algorithm based on linear matrix inequality breaks into two parts- online and offline. In order to improve the response speed and control accuracy of system, the upper bound of objective function obtained by offline algorithm is set as a known parameter and reoptimization is carried out to get a sequence of asymptotically stable invariant ellipsoid sets. The state variables can be determined-lying between two ellipsoid sets within the ellipsoid set sequence, on the basis of online measured state variables. The state variables can be accurately located and the system control variable can be obtained by adding an adjacent small elliptical set and optimizing with three ellipsoid sets. The online optimization is theoretically proved. Simulation results indicate that the control effect of the proposed algorithm is better than conventional algorithms. Taking fan control system in variable air volume （VAV） air-conditioning system as an example, the results verify the effectiveness of the proposed algorithm.

关 键 词：模型预测控制 线性矩阵不等式 鲁棒 在线优化 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]