【作者】 贺国艳；

【导师】 杨马英；

【作者基本信息】 浙江工业大学 ， 控制理论与控制工程， 2002， 硕士

【摘要】 模型预测控制作为一种新型计算机控制算法，在工业控制界取得许多成功应用，现已成为一种重要的先进控制策略。这主要得益于它的三个基本特征：模型预测，滚动优化和反馈校正。 与其他控制算法相比，预测控制有其自身特点： a．对模型的精度要求不高，建模方便，过程描述可由简单实验获得。 b．采用滚动优化策略，而非全局一次优化，能及时弥补由于模型失配、干扰等因素引起的不确定性，鲁棒性较好。 c．易将算法推广到有约束、大迟延等实际过程，能有效处理多变量、有约束问题。 本文主要对其中的两种算法：动态矩阵控制(DMC)和预测函数控制(PFC)进行研究。对动态矩阵控制，主要应用多变量DMC对复杂系统进行仿真；对预测函数控制，主要进行了算法改进。论文主要内容如下： (1)对预测控制的发展概况、研究现状及研究动向作了概述，同时介绍了与本论文相关的预测控制方案。 (2)对单变量预测函数控制的基本算法进行讨论，针对一阶加纯滞后对象进行PFC算法推导，并仿真验证PFC算法所具有的跟踪快速、抗干扰能力强、控制效果良好的特点。 (3)将极点配置思想与PFC算法结合起来，提出了极点配置的PFC算法。通过在预测函数优化性能指标中引入加权多项式来进行极点配置，适当选择加权多项式，将极点配置在给定位置，获得所期望的闭环响应特性。 (4)结合解耦思想研究多变量系统神经网络解耦PFC算法，通过引入神经网络补偿环节，对多变量系统进行解耦，并在此基础上，对解耦后各子系统进行单变量预测函数控制，以确定各个控制量。仿真表明，该算法有较好的跟踪 摘 要特性，对解决多变量系统的优化和控制具有一定的适用性。 归）结合分层协调的思想，研究多变量分层协调预测函数控制算法。仿真表明，该算法是一种有效的处理多变量系统的方法。 …）采用多变量动态矩阵控制结合Simulink对典型多变量非线性系统一CSTR模型进行仿真。仿真结果表明该算法对多变量系统具有较好的跟踪和控制作用。表明该算法对复杂系统控制的有效性。 最后，对全文的工作进行了总结，分析了存在的问题，并对今后的发展方向进行了展望。更多还原

【Abstract】 As a computer control algorithm, Model Predictive Control (MFC) has many successful applications in industry control process and has become an important advanced process control strategy. It has three basic characteristics: model prediction, receding horizontal optimization and feedback correction.Compared with other control algorithms, MFC has its own features:a. Not require precise model prediction, model can be obtained by simple experiment.b. By using receding horizontal optimization strategy, the uncertainty is compensated for model mismatch or disturbance existed.c. The algorithm can be easily extended to constrained and large delay processes and can deal with constrained multivariable problems efficiently.This thesis focuses on two algorithms: Dynamic Matrix Control (DMC) and Predictive Function Control (PFC). For DMC, its application is studied. A complex process is simulated based on multivariable DMC. For PFC, the algorithm is improved and control strategy is studied. The thesis includes following contents:(1) A survey of the development and status of model predictive control is given. An introduction of related control strategies is also given.(2) Based on the detailed description of the principles and algorithm of the PFC, it can be verified that the PFC has the following advantages: simple algorithm, less on line calculation, high control precision and disturbance attenuation by the theory analysis and simulation.(3) PFC algorithm based on pole-placement is given. Poles are assigned to given position by selecting suitable weighted polynomial and adding it in optimal performance index. The expected closed loop response is obtained.(4) Neural net decoupled PFC algorithm is discussed. Multivariable process is decoupled with neural net compensation. The decoupled subsystem is applied tomsingle variable PFC. The simulation result shows the algorithm has better tracking performance. It is suitable for resolving optimization and control of multivariable process in some extent.(5) Combined with coordination control algorithm, multivariable and two-layer coordination PFC is studied. The result of simulation shows the algorithm is an efficient way to deal with the multivariable system.(6) A typical multivariable nonlinear system ~ CSTR model is simulated by applying multivariable DMC with Simulink. The simulation result shows the algorithm has a better tracking and control ability for multivariable process.Finally, based on the summarization of the research results in this thesis, further research areas about MFC are discussed.更多还原