多变量动态解耦预测PI在精馏塔中的应用  

Application of Multivariable Dynamic Decoupling Predictive PI Control Algorithm in Distillation Column

作　　者：杨自 任正云[1] 余雯 YANG Zi;REN Zheng-yun;YU Wen(College of Information Science and Technology,Donghua University,Shanghai 201620,China;Sichuan Tobacco Redrying Co.,Ltd.,Chengdu 610041,China)

机构地区：[1]东华大学信息科学与技术学院,上海201620 [2]四川烟叶复烤有限责任公司,四川成都610041

出　　处：《控制工程》2022年第9期1688-1694,共7页Control Engineering of China

基　　金：国家自然科学基金资助项目(61673199)。

摘　　要：精馏塔是一个多输入多输出的高阶系统,其内在机理复杂,具有强耦合、强干扰、大滞后等特性。精馏塔精馏工艺对控制器要求较高,常规的PID控制过程中存在震荡和超调现象,会导致产品品质不佳,而一般的智能控制鲁棒性较差,参数不易整定,无法满足实际需求。针对该问题,提出了多变量动态解耦预测PI控制算法。该算法使用对角动态解耦器解除变量间的耦合特性,对由解耦器和过程对象组成的广义过程对象进行模型降阶,并对每一个相互独立的降阶模型运用预测PI算法进行控制研究。仿真结果表明,多变量动态解耦预测PI控制算法具有良好的稳定性和鲁棒性,且抗干扰能力强、响应速度快、参数易整定,具有一定的实际应用价值。The distillation column is a high-order system with multiple inputs and multiple outputs. Its internal mechanism is complex, and it has the characteristics of strong coupling, strong interference, and large lag. The distillation process of the distillation column has high requirements on the controller. There are oscillations and overshoots in the process of the conventional PID control, which leads to poor product quality. The general intelligent control has poor robustness, and the parameters are difficult to adjust, which can not meet the actual needs. To solve these problems, a multivariable dynamic decoupling predictive PI control algorithm is proposed.In this algorithm, the diagonal dynamic decoupler is used to decouple the variables, and the model order of the generalized process object composed of the decoupler and the process object is reduced. The predictive PI algorithm is applied to the control of each independent reduced-order model. The simulation results show that the proposed algorithm has good stability and robustness, strong anti-interference ability, fast response speed,and easy parameter tuning, which has certain practical application value.

关 键 词：精馏塔 预测PI控制 动态解耦方法 模型降阶 鲁棒性 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]