支持向量机α阶逆系统解耦控制方法  

αth-order inverse decoupling control method based on support vector machines

作　　者：宋夫华[1] 李平[2] 

机构地区：[1]中国计量学院机电工程学院,浙江杭州310018 [2]浙江大学工业控制技术研究所工业控制技术国家重点实验室,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2007年第2期226-229,共4页Journal of Zhejiang University：Engineering Science

基　　金：国家"973"重点基础研究发展规划资助项目(2002C13312200);高等学校博士学科点专项科研基金资助项目

摘　　要：为了解决采用传统逆系统方法难以建立逆模型的困难,提出了基于支持向量机(SVM)的α阶逆系统解耦控制方法.在被控对象精确模型未知的情况下,根据SVM的非线性辨识原理,利用具有高斯核函数的SVM离线建立被控对象的非线性逆模型,根据逆系统方法原理,将得到的SVM逆模型串联在原系统之前,将复杂的非线性多变量系统解耦成多个相对独立的单输入输出伪线性子系统.通过对一个典型的两输入两输出强耦合非线性系统进行仿真,结果表明,在没有系统模型先验知识的情况下,利用该方法能够建立准确的非线性系统的逆模型.基于SVM的α阶逆系统解耦控制方法适用于较一般的离散非线性系统,且结构简单,易于工程实现.An ath-order inverse decoupling control method based on support vector machines （SVM） was presented to resolve the difficulty of inverse modeling with the traditional inverse system methods. On the premise of not knowing the accurate mathematical model of the controlled plant, according to the nonlinear identification theory of SVM, the nonlinear offline inverse model of the plant was built by SVM with Gauss kernel function. Based on the theory of inverse system method, the SVM inverse model was cascaded before the original system to decouple a complex nonlinear multivariable system into several relatively independent single input single output （SISO） pseudo-linear sub-systems. Simulation of a two-input two- output strong coupling nonlinear system demonstrated that even without prior information about the plant, the presented method could accurately model the inverse dynamic system. The method can be applied to general nonlinear multivariable inputs and multivariable outputs （MIMO） discrete systems, and can be easily implemented for engineering.

关 键 词：支持向量机 非线性多变量系统 逆系统方法 解耦 

分 类 号：TP271.8[自动化与计算机技术—检测技术与自动化装置]