【作者】 刘富春；

【导师】 钱积新； 赵均；

【作者基本信息】 浙江大学 ， 控制科学与工程， 2003， 博士

【摘要】 预测控制技术自70年代产生以来，以其优良的控制性能受到过程控制界的青睐。以预测控制技术为核心的多变量有约束控制软件包，更以其在应用中所显现的巨大经济效益而备受世人瞩目。相关文献仅仅介绍软件包的功能与应用，却很少涉及其技术细节。为了加快发展我国的工业控制技术，替代进口，开发出我们自己的先进控制软件，势在必行。 本文面向工程应用对我们自主开发的多变量模型预测控制工程软件FrontAPC的体系结构，算法，技术以及在过程控制中的应用进行了深入研究。从模型预测控制工程软件的实际需要出发对模型预测控制算法的模型预测、稳态优化算法，动态优化算法进行了研究。针对实际的两个工业过程，通过对其特性的分析，给出了基于模型预测控制的控制方案，并进行了应用和仿真研究。 本文的主要贡献包括： 1．介绍了我们自主开发的模型预测控制工程软件—FrontAPC的体系结构以及其具有的功能和特点。并给出了这些功能的实现形式。 2．在FrontAPC稳态优化部分，作者给出了一种多优先级稳态优化算法。 该算法充分考虑实际工程需要，根据工艺工程师定义的优先级依次满足优化目标，从而避免了将优先级不同的优化目标放到一个目标函数中所导致的多目标问题，并充分利用系统的自由度，保证了解的唯一性。 3．在FrontAPC动态优化部分，采用区间优化动态控制算法。该算法中，作者采用改进的预测模型来减小模型时域与消除截断误差，形式上采用区间控制防止操纵变量的频繁动作。并在算法中实现了漏斗技术与block技术，利用性能比来使得控制器调整简单直观。 4．作者给出了一种多变量有约束快速DMC算法，该算法与常规多变量有约束DMC算法比较，极大减小了计算负担，而性能上损失很少，可以在现有的DCS系统上进行二次开发，对于一些资金缺乏的工厂在现有DCS系统基础上实现先进控制很有实际意义。 5．介绍了模型预测控制在PTA装置中的应用。针对扬子石化PTA装置中摘要 的加氢反应浆料浓度控制系统，作者进行了系统分析，方案设计，测 试，辨识，以及预测控制器投运。作者还参与了溶剂脱水塔控制与优 化的，方案设计，测试，辨识，以及控制器投运，两项目己通过江苏 省科技厅鉴定，取得年经济效益1913.2万元。更多还原

【Abstract】 Model predictive Control (MPC) technology has gained considerable attention for its good performance since the generation of the MPC in the late 1970s. The great economic benefits by using the MPC packages make the people’s focus attention upon them. In the published papers, only introduction of the functions and applications of the packages can be found, little technology is introduced in detail. In order to advance the technology of Industrial control in our country, the only way is to develop our own commercial industrial control package.In the dissertation, we investigate the Framework, key algorithm and technology of FrontAPC, the model predictive control package developed by ourselves, and its application in the field of process control. According to the implementation and practical requirement, we research the local steady-state optimization and dynamic optimization algorithm in MPC package.This dissertation is composed of follows:1. The Framework, functions and technology in FrontAFC, the MPC package developed by ourselves, is presented.2. In the part of steady optimization in FrontAPC, a multi-priority algorithm is given. This algorithm satisfies the optimized objectives sequentially according to the priorities defined by operator, so avoids the problem of multi-objectives, which caused by putting all optimized objectives in a objective functions. Also this algorithm can use up the degrees of freedom of system, and acquire unique solution.3. In the Part of dynamic optimization in FrontAPC, author give a range dynamic optimization algorithm and its transformation. It adopts developed predictive model to shorten the predictive time and eliminate the truncation error, adopts range control to avoid the frequent change of inputs, adopts block technology to reduce the dimension of QP, uses performance ratio to make the controller’s tuning sample and intuitionistic.4. A multivariable fast constrained DMC algorithm is presented, which adopts improved model to remove truncation error, and computes one control move to optimize one point of future horizon for each output. Compared to traditional DMC algorithm, this algorithm reduces calculational burden greatly, and loses less performance. It has practical significance for some factories that are short of money to achieve advanced control.5. The application of Model Predictive Control to PTA equipment is presented. For the control of slurry density system and advanced control and optimization of solvent dehydration tower, author does a lot of work in system analysis, the design of control scheme, the identification of models, and the implement of controller. The applications to PTA equipment have acquired big economic benefit.更多还原