基于支持向量机建模的非线性预测控制研究
发布时间：2018-08-24 16:26
【摘要】：针对化学工业中经常出现的纯延迟对象以及非最小相位对象,传统的控制方法如PID控制,LQ最优控制的控制效果都不够令人满意,在面对多变量系统也无能为力。预测控制算法的出现正好解决了这个难题,在面对操作变量和被控变量维数很高,需要满足物理约束,时滞系统等问题中,预测控制具有十分明显的优势。但是随着过程工业日益大型化、连续化,现代工业生产过程日益复杂,此时,线性预测控制方法就不能很好的完成控制任务,而非线性预测控制又存在在线计算量大的问题。本文主要针对非线性预测控制,利用支持向量机进行建模,提出了一种在每一个采样点进行局部线性化的方法,把在线计算最优解问题转化为求解一个二次规划问题,通过这样的方式减少了非线性预测控制在线计算量大的问题,并保证了控制的精确度和稳定性。 本文的研究工作主要为以下几方面： 1.预测控制和支持向量机的内容研究 本文对预测控制的基本原理、发展过程和主要特点进行了简明的介绍,同时详细的阐述了支持向量机的基础理论,包括支持向量机的基本原理以及针对分类和回归两种不同用途的介绍和对比。 2.基于SVM-Wiener模型的非线性预测控制研究 论文详细的阐述了SVM-Wiener模型的建模过程,包括动态线性部分和静态部分非线性部分的详细分析和研究。并给出了非线性预测控制器的设计流程,重点介绍和分析了非线性预测控制局部线性化的计算方法。 3.基于SVM的Hammerstein-Wiener模型非线性预测控制研究 论文简单介绍了Hammerstein-Wiener模型的结构特点,详细推导了基于SVM的Hammerstein-Wiener模型的建模过程。根据模型的结构和特点设计了非线性预测控制器。 4.基于连续搅拌釜式聚合反应器的仿真 在论文的最后,对文章介绍的非线性预测控制方法进行了基于连续搅拌釜式聚合反应器的仿真,并通过大量的数据分析和仿真图对比,证明了所提出方法的控制精确度、稳定性以及高效性。 本文中有各类结构图、仿真图、流程图等共29张,进行数据对比以及定义参数等表共8张,引用参考文献52篇。
[Abstract]:The traditional control methods such as PID control and LQ optimal control are not satisfactory for the pure delay object and non-minimum phase object which often appear in the chemical industry and can not be used in the face of multivariable system. The emergence of predictive control algorithm solves this problem. In the face of the high dimension of operational variables and controlled variables, it needs to meet the physical constraints, time-delay systems and other problems, predictive control has a very obvious advantage. However, with the process industry becoming larger and more continuous, the modern industrial production process is becoming more and more complex. At this time, the linear predictive control method can not complete the control task well, and the nonlinear predictive control has the problem of large amount of on-line calculation. In this paper, for nonlinear predictive control, support vector machine (SVM) is used to model, and a method of local linearization at every sampling point is proposed. The problem of online optimal solution is transformed into a quadratic programming problem. In this way, the problem of large online computation of nonlinear predictive control is reduced, and the accuracy and stability of the control are guaranteed. The main work of this paper is as follows: 1. Research on the content of Predictive Control and support Vector Machine in this paper, the basic principle, development process and main characteristics of predictive control are briefly introduced. At the same time, the basic theory of support vector machine is expounded in detail. Including the basic principles of support vector machines and the classification and regression for two different uses of the introduction and comparison. 2. Research on nonlinear Predictive Control based on SVM-Wiener Model the modeling process of SVM-Wiener model is described in detail, including the detailed analysis and research of dynamic linear part and static part nonlinear part. The design flow of nonlinear predictive controller is given, and the calculation method of local linearization of nonlinear predictive control is introduced and analyzed emphatically. Research on nonlinear Predictive Control of Hammerstein-Wiener Model based on SVM in this paper, the structural characteristics of Hammerstein-Wiener model are briefly introduced, and the modeling process of Hammerstein-Wiener model based on SVM is deduced in detail. According to the structure and characteristics of the model, a nonlinear predictive controller is designed. At the end of the paper, the nonlinear predictive control method introduced in this paper is simulated based on the continuous agitator polymerization reactor. The control accuracy, stability and efficiency of the proposed method are proved by a large number of data analysis and simulation diagram comparison. In this paper, there are 29 structural diagrams, simulation diagrams and flowcharts, 8 tables for data comparison and definition of parameters, and 52 references.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TB114.2
本文编号：2201369
[Abstract]:The traditional control methods such as PID control and LQ optimal control are not satisfactory for the pure delay object and non-minimum phase object which often appear in the chemical industry and can not be used in the face of multivariable system. The emergence of predictive control algorithm solves this problem. In the face of the high dimension of operational variables and controlled variables, it needs to meet the physical constraints, time-delay systems and other problems, predictive control has a very obvious advantage. However, with the process industry becoming larger and more continuous, the modern industrial production process is becoming more and more complex. At this time, the linear predictive control method can not complete the control task well, and the nonlinear predictive control has the problem of large amount of on-line calculation. In this paper, for nonlinear predictive control, support vector machine (SVM) is used to model, and a method of local linearization at every sampling point is proposed. The problem of online optimal solution is transformed into a quadratic programming problem. In this way, the problem of large online computation of nonlinear predictive control is reduced, and the accuracy and stability of the control are guaranteed. The main work of this paper is as follows: 1. Research on the content of Predictive Control and support Vector Machine in this paper, the basic principle, development process and main characteristics of predictive control are briefly introduced. At the same time, the basic theory of support vector machine is expounded in detail. Including the basic principles of support vector machines and the classification and regression for two different uses of the introduction and comparison. 2. Research on nonlinear Predictive Control based on SVM-Wiener Model the modeling process of SVM-Wiener model is described in detail, including the detailed analysis and research of dynamic linear part and static part nonlinear part. The design flow of nonlinear predictive controller is given, and the calculation method of local linearization of nonlinear predictive control is introduced and analyzed emphatically. Research on nonlinear Predictive Control of Hammerstein-Wiener Model based on SVM in this paper, the structural characteristics of Hammerstein-Wiener model are briefly introduced, and the modeling process of Hammerstein-Wiener model based on SVM is deduced in detail. According to the structure and characteristics of the model, a nonlinear predictive controller is designed. At the end of the paper, the nonlinear predictive control method introduced in this paper is simulated based on the continuous agitator polymerization reactor. The control accuracy, stability and efficiency of the proposed method are proved by a large number of data analysis and simulation diagram comparison. In this paper, there are 29 structural diagrams, simulation diagrams and flowcharts, 8 tables for data comparison and definition of parameters, and 52 references.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TB114.2
【参考文献】
相关期刊论文 前10条
1 李丽娜,侯朝桢;基于支持向量机(SVM)的工业过程辨识[J];北京理工大学学报;2003年05期
2 朱国强,刘士荣,俞金寿;支持向量机及其在函数逼近中的应用[J];华东理工大学学报;2002年05期
3 李超峰;卢建刚;孙优贤;;基于SVM逆系统的非线性系统广义预测控制[J];计算机工程与应用;2011年02期
4 张智焕,谢磊,张泉灵,王树青;基于非线性(Wiener)模型自适应预测函数控制[J];计算技术与自动化;2003年02期
5 王远景;萧蕴诗;;应用于火电厂主蒸汽温度预测控制的PSVM算法研究[J];计算技术与自动化;2006年04期
6 黄宴委,吴惕华,赵静一;回归型支持向量机的系统辨识及仿真[J];计算机仿真;2004年05期
7 朱小萍;夏陆岳;孙小方;蔡亦军;周猛飞;潘海天;;基于SVM增量学习的氯甲烷含水量预测控制[J];科技通报;2012年01期
8 戴文战;娄海川;杨爱萍;;非线性系统神经网络预测控制研究进展[J];控制理论与应用;2009年05期
9 王定成,方廷健,高理富,马永军;支持向量机回归在线建模及应用[J];控制与决策;2003年01期
10 张浩然,韩正之,李昌刚;基于支持向量机的非线性系统辨识[J];系统仿真学报;2003年01期
，本文编号：2201369
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2201369.html
