【作者】 李林欢；

【导师】 苏宏业； 褚健；

【作者基本信息】 浙江大学 ， 控制理论与控制工程， 2003， 硕士

【摘要】 预测控制是一种基于模型的先进控制技术，亦称模型预测控制(model predictive control)，它是二十世纪七十年代中后期，在欧美工业领域内出现的一类新型的计算机控制算法。1978年，法国学者Richalet等首次系统的提出了模型预测启发式控制的概念及其产生的背景、机理与应用效果。之后，Mehra、Culter和Clark等人进一步发展了模型预测控制理论。八十年代，模型预测控制技术以其独特的优点和特性引起了西方学者的广泛重视，众多学者从不同的理论角度，运用各种数学手段对其进行了深入的研究和分析，使得模型预测控制逐渐发展成为先进控制技术中相对独立的一个分支。模型预测控制是一种十分有效的优化控制策略，它弥补了现代控制理论解决复杂对象控制问题时所无法避免的不足之处。近年来随着智能控制、PID控制、小波理论、鲁棒控制、混杂系统控制等理论研究领域的不断进步，预测控制理论得到了快速的完善和发展。在变结构控制、时滞系统控制、混杂系统控制、输出反馈控制等各方面均已有引入预测控制算法策略的混合控制理论研究和应用报道。 本文正是根据模型预测控制理论的研究现状，以及实际工业过程控制应用中所面临的对模型预测控制所提出的新要求，主要针对当前模型预测控制理论所面临的若干亟待解决的问题进行了深入的研究和探讨，并最终给出了相应的研究成果。本文的主要研究工作概括如下， 1、针对实际工业过程控制中常见的多输入/多输出系统，研究了一类参数未知的多变量系统广义预测PID控制参数整定问题。基于广义预测控制(GPC)思想和γGPC与PID的相互关系，提出了一种新的离散多变量系统PID参数整定方法。同其他方法相比，该预测PID控制器不仅具有常规PID所具有的结构简单的特点，而且在性能上兼有广义预测控制这类先进控制算法的优点。此外，在满足通常匹配条件的情况下，该控制策略适用于非最小相位系统和有时滞的系统。 2、一般模型预测控制是建立在被控对象模型基础上，而实际工业过程中对象一般均存在非线性特性，往往难以得到系统的精确内部模型。此外，若对非线性系统直接采用模型预测控制器的设计方法，将会导致繁复的在线非线性优化过程，既满足不了系统实时控制的要求，也可能无法保证满意的控制效果。因此，本文提出了一种新的基于T-S模糊模型的非线性系统预测PID控制策略。在此基础上，还进一步探讨了该系统的稳定性问题，并给出了相应的分析结果。 11 摘 要 3、在讨论并总结当前有关预测PID控制研究现状的基础上，提出了一种新 的基于输出误差预测的非线性系统模糊预测PID控制器设计方法。该方 法采用TS模糊模型来描述复杂的非线性系统，同时在局部线性化基础 上结合广义预测控制的思想和有限脉冲响应滤波器的定义，提出了一类 新型的模糊预测PID控制器的实现方法，该方法不仅有别于一般的预测 PID控制器的设计，而且解决了一般预测PID控制器设计当中对系统模 型阶次的限制问题，从而将其推广到了一类更广泛的系统。 一、针对实际工业过程中对象本质为非线性的问题，提出了一种新型的基于 双线性模型的预测PID控制算法。该算法综合了模型预测控制（MPC） 的输出预测功能和常规PID控制的优点。同时，由于该算法采用了双线 性模型，从而不仅避兔了非线性预测控制中涉及到的繁复在线寻优问 题，而且所得到的控制器是解析解而非数值解的形式。同常规的PID控 制器相比，该控制算法具有更好的控制品质。 5、针对一个pH中和过程的系统，利用基于TS模糊模型的预测PID控制 方法对被控对象进行了综合设计。这种方法不仅在性能上能够达到满意 的控制效果，而且在参数的调节上，对于已熟悉PID参数调节的操作人 员来说，亦便于学习和掌握。 最后是全文的总结以及展望。更多还原

【Abstract】 Predictive control is a kind of advanced control technologies that is based on the model of plants, so it is also called model predictive control. It is a novel computer control algorithm first appeared in industrial area in Europe and America in 1970s. French researcher-Richalet et al first proposed the conception, background, mechanism and performance of the model predictive heuristic control. Then the theory of model predictive control was further developed by Mehra, Culter and Clark et al. In 980s, model predictive control received extensive attention of the western researchers for its distinct merits and features and it has been deeply studied from different aspects with many mathematic methods, In the recent years, with the development of many kinds of the control theory, such as intelligent control, PID control, wavelet theory, robust control, and hybrid systems control etc., the theory of model predictive control has also obtained rapid development. Moreover, the theory researches and corresponding application reports of model predictive control also appeared in the field of the variable structure control, time-delay system control, hybrid system control, output feedback control and so on.Based on the study status of model predictive control and the new requirements from the practiceb industrial processes for the theory of model predictive control, some desiderated problems are studied and discussed in this article, also the corresponding results are given. The main contents are as follows:1. For a class of multivariable systems, the tuning of PID controller parameters is discussed in this paper. Based on the idea of GPC and the relation between y GPC and PID controllers, a new tuning algorithm of PID controller parameters for discrete multivariable systems is presented. Compared with the conventional PID control, the proposed predictive PID control not only has simple control structure as conventional PID, but also maintains performance of the advanced control algorithm such as generalized predictive control. Furthermore, when the matching conditions are satisfied, the proposed method can also be used for nonminimum-phase systems and time-delay systems.IV Abstract2. Considering the model predictive control algorithm based on the model of the plant, the accuracy of the model is closely connected with the control performance. As the actual plant is nonlinear, it is hard to obtain the accurate internal model. Furthermore, if the nonlinear model is considered for control design directly, it will reduce to nonlinear optimization, which can not satisfy the requirement of on-line control. In addition, the current research results on predictive PID control mainly focus on the linear systems and the results about the nonlinear predictive PID control. In this paper, a novel predictive PID control based on T-S fuzzy model is proposed, which can deal with a class of large-scale nonlinear systems. What’s more, the stability of the closed-loop systems is also studied and some analytical conclusions are given.3. Based on the discussion and summarization for the merits and shortcomings of current predictive PID control theory, a new T-S fuzzy model predictive PID control is given based on the output-error predicted. A proper T-S fuzzy model is adopted to describe the complex nonlinear system. Based on the method of local linearization, the idea of generalized predictive control (GPC) and the definition of Finite Impulse Response are utilized to design the novel control strategy. The new method not only is different from the common predictive PID control, but also solves the problem that the system’s order is restricted in the reported papers. So the presented method is fit for more common systems and PID tuning knobs on the industrial controller also can be used to adjust the performance of the closed-loop system. Furthermore, the compute burden is greatly reduced.4. Considering the fact that a class of la更多还原