【作者】 吴斌；

【导师】 靳其兵；

【作者基本信息】 北京化工大学 ， 控制理论与控制工程， 2008， 硕士

【摘要】 传统的系统辨识及其PID控制器的设计方法,虽然在很大程度上能满足工业系统的控制要求,但对一些具有MIMO、强藕合性、不确定性、非线性、信息不完全性和大纯滞后性等特征的工业控制系统,这些经典方法无法得到满意的效果,迫切需要对多变量系统的辨识及其PID参数整定进行研究,以求整定出合理的、最优的PID参数,以获得比传统的PID参数整定法更好的控制效果,更能适应复杂多变的工业生产过程的需要。论文主要分成两部分来展开论述:多变量系统的辨识部分和PID整定方法部分,其中对多变量系统的辨识方法未涉及太深,只是做了初步研究,重点在多变量系统的PID整定方法上,做了大量工作,取得了一定的成果。在进行多变量PID参数整定方法的研究之前,首先介绍了两种多变量系统的辨识方法:最小二乘法和递阶随机梯度法。在论述了算法基本原理和实现步骤的基础上,通过仿真实验证明了算法的有效性。在多变量PID参数整定部分,首先对PID控制的基本原理与特点进行了概述,然后说明了PID参数整定的分类和传统方法,针对多变量系统的特点,概述了多变量PID的设计方法。而后对这些方法进行了具体研究,主要有预测PID参数整定、内模PID参数整定和鲁棒PID参数整定,分别在文中的第四章、第五章和第六章展开论述:第四章在研究预测控制算法基本原理的基础上,将预测控制算法与PID参数整定相结合,形成预测PID整定算法,考虑到对预测PID参数的优化,提出了改进型预测PID整定算法,并将单变量预测PID控制推广,设计出符合多变量特点的预测PID控制器。比起传统的PID参数整定法（如Z-N整定法）,预测PID整定法对模型失配、大滞后等传统整定法难以控制好的情况均有满意的控制效果。第五章介绍了内模控制器的基本原理,在此基本上,将内模控制算法与PID参数整定相结合,形成IMC-PID整定算法,并将单变量IMC-PID控制推广,设计出符合多变量特点的IMC-PID控制器,主IMC-PID控制器用于保证输出的动态响应,副IMC-PID控制器用于及时克服回路间的耦合干扰,实现控制器解耦。并通过仿真实例证明了IMC-PID参数整定算法的有效性和鲁棒性。第六章基于鲁棒控制的相关理论,介绍了两种多变量鲁棒PID控制器参数整定的方法,一种为多指标相容下多变量PID参数整定,该算法保证了闭环系统的稳定性,并具有一定衰减度,且符合鲁棒性能指标H_∞约束;另一种为基于结构Lyapunov矩阵的多变量PID参数整定,保证了闭环系统的稳定性,且符合鲁棒性能指标H₂/H_∞约束。最后对化工生产中的实际模型,进行了仿真,以证明算法的有效性。在论文的最后,对所做的工作做了总结与前景展望。更多还原

【Abstract】 The traditional identification and PID control are suitable for industry produce in many cases, but for model error, the system with big delay time and the multivariable system with coupling which often appear in industrial processes, the classical methods can not work well. Therefore, the reasonable methods for multivariable system should be studied, to get better control effect and adaptability for complex industrial processes.The paper begins to research in two sections: identification of the multivariable system and the methods of tuning PID. The important point is how to tune PID parameters for multivariable system, many tests have been done and get some valuable conclusions.Before the research for tuning PID, the two methods of identification are introduced: least squares algorithm and stochastic hierarchical gradient algorithm. After discussing the theory of the algorithms, the validity and veracity are checked by simulation experiments. In the section of tuning PID, the summary about the basic principle and characteristic of PID controller firstly are introduced. Secondly, the paper explains the sort and traditional method of PID parameters tuning. Meantime, the methods for multivariable system are summarized, then the detailed studies are discussed: predictive-PID parameters tuning method, IMC-PID parameters tuning method and robust-PID parameters tuning method. These methods are studied in three chapters.In the forth chapter, with the basic principle of prediction algorithm, the idea of GPC with PID control is provided. Considering to optimize PID parameters, an amelioration Predictive-PID parameters tuning method brings forward. Otherwise, GPC-PID is applied in multivariable system. Comparing with the traditional PID parameters tuning methods （for example, the Z-N tuning method）, Predictive-PID tuning method not only has better dynamic response characteristic, but also has satisfying control effect for the instances which have difficulties in controlling for traditional tuning methods, for example, model error and big delay time.In fifth chapter, firstly, the basic principle of IMC algorithm is introduced, then IMC-PID parameters tuning method that combines the idea of IMC with PID control is offered. Otherwise, IMC-PID is applied in multivariable system to form four PID controllers, two primary PID is used to satisfy the dynamic response of system and two assistant PID is used to eliminate disturb between two recycles. The validity and robustness are checked by simulation experiments.In the sixth chapter, under the basic principle about robust algorithm, two PID parameters tuning methods for multivariable system are introduced .One is multivariable PID controller’s design with multi-objective consistency, the other is the multivariable H₂/H_∞PID controller’s design based on structureLyapunov matrix. Then, the validity and robustness are checked for real model in chemical process.At last, the paper concludes by summarizing some key contents of what have been done and future work.更多还原