【作者】 张智焕；

【导师】 王树青；

【作者基本信息】 浙江大学 ， 控制科学与工程， 2002， 博士

【摘要】 实际系统大都是非线性的，因此研究非线性系统的预测函数控制无论在理论上还是在实际应用中都具有重要意义，也是当前控制理论界的热点研究问题之一。本文从实际出发，对时变线性和非线性系统的预测控制的理论和应用进行了深入研究。提出了几种新的预测函数控制方法，本文主要内容有以下几个方面： 1．在简要介绍模型预测控制基本原理之后，从理论及应用方面概述了模型预测控制的发展历史及现状，并给出了本文的研究内容和成果。 2．针对pH过程的严重非线性特性，提出了多模型预测函数控制方法，并设计出相应的预测函数控制器，同时用模糊识别调度控制器来解决多模型预测函数控制切换的关键问题。 3．针对多模型控制切换中的抖动现象，提出了一种隶属度加权的多模型预测函数控制方法，该方法解决了多模型切换过程中存在的扰动问题，并以连续搅拌槽反应器(CSTR)为对象进行了仿真，研究结果表明能实现平滑控制。 4．提出了基于Laguerre模型的自适应双值预测函数控制方法，该方法综合了Laguerre模型的非参数化和参数化的优点以及自适应控制在线修改模型参数和预测函数控制快速跟踪能力的优点，克服了常规自适应预测函数控制必须事先知道系统时延和阶次信息的缺陷。 5．针对非线性系统和不确定对象，进行了多模型预测控制的切换算法比较；讨论了本文第三章的切换算法、第四章的加权算法及其它方法的优缺点；仿真结果表明：对非线性系统，基于多模型的加权算法比切换算法具有较好的控制品质；对大范围变化的不确定性对象，给出了新的切换控制策略，其控制品质明显优于常规的单模型控制。 6．针对多输入多输出双线性系统，研究了基于非线性过程精确反馈解耦线性化的预测函数控制方法。这是一种分层的控制策略，首先通过一个非线性状态反馈，使得闭环系统是输入输出解耦和线性的；然后设计一组单输入单输出预测函数控制器。下层为上层预测函数控制提供一组单输入单输出模型，而上层预测函数控制以其固有的鲁棒性来补偿参数变化和11 浙江大学博士学位论文解耦线性化的误差，以纸机加压网前箱工程为例进行了仿真。 7．研究了多模型预测函数控制在连续生化过程和流化床热风温度控制中的应用。 在总结全文的基础上，提出了复杂系统预测控制有待深入研究的若干问题。更多还原

【Abstract】 Predictive control of nonlinear systems is one of the most important problems in control theory. From the standpoints of theory and practical applications, this dissertation is devoted to predictive control strategies for time-varied linear system and complex nonlinear systems. Several new predictive control methods are presented. The main research works are as follows:1. A survey of model predictive control on theory and applications is introduced. The main research results are given.2. Multiple model predictive control strategy is presented based on different equilibrium point state of a nonlinear pH process, the local linear model is recognized according to maximum membership function rule and the corresponding controller is chosen, and the key problem of multiple model predictive control switching is solved.3. Multiple Models Predictive Function Controller (MMPFC) based on membership function weighting is proposed for disturbance in multiple models switching. Simulation results in a continuous stirred tank reactor (CSTR) show that controlled dynamic process is more smoothly than the multiple models switching control strategy.4. A new strategy of adaptive predictive function control based Laguerre model is suggested. It combines advantages of Laguerre non-parameter and parameter model, adaptive control on-line modifying parameter of model with predictive function control fast tracking. The drawback of knowing in prior time-delay and order in general adaptive predictive function control is overcame.5. Multiple models switching control algorithm is compared for a class of nonlinear systems, the sudden changes of model parameters and uncertain linear systems. The advantages and disadvantages of switching control algorithm in chapter 3, weighted-sum algorithm of the linear model by using model validity functions hi chapter 4 are discussed. Simulation result of aCSTR show that the proposed switching control algorithm is indeed superior to the conventional counterpart.6. The control strategy based on global feedback linearization for nonlinear systems is developed for MIMO bilinear systems. Using the linearized input/output decoupling model, the Predictive Function Controller (PFC) is designed. The simulation results of the paper machine pressured headbox model show that the performance of the derived control strategy is good.7. The application of multiple models predictive function control in continuous fermentation processes and drying fluid-bed is studied.The dissertation concludes with a summary and perspectives of future research of nonlinear predictive control in complex system.更多还原