一种基于改进聚类算法的模糊模型辨识  

A Fuzzy Identification Based on Improved Clustering Algorithm

作　　者：武俊峰[1] 艾岭[1] 

机构地区：[1]哈尔滨理工大学自动化学院,黑龙江哈尔滨150080

出　　处：《哈尔滨理工大学学报》2010年第3期1-5,共5页Journal of Harbin University of Science and Technology

基　　金：教育部科学技术研究重点项目(206041);教育部高校博士点专项研究基金项目(20060214004);哈尔滨市科技攻关计划项目(20051AA1CG037)

摘　　要：传统建模方法对于建立精确的存在多变量、强耦合、大时滞以及不确定性的非线性系统模型无能为力,从而难于精确表达复杂系统及实施整体优化控制.针对传统模糊C-均值聚类算法对初始值敏感及无法确定最优规则数的缺陷,提出了一种基于改进聚类算法的模糊辨识方法.它通过减法聚类和有效性函数确定初始聚类中心,然后采用一种全局模糊C-均值聚类算法寻找出最终聚类中心,并利用最近临域法确定合适的区域半径,最后通过递推最小二乘法建立系统的T-S模糊模型,对电阻炉温度系统进行仿真,说明本文所述方法的有效性.The traditional modeling methods are unable to identify a nonlinear system which has multivariables randomicity,strong coupling and long delay.Thus,it is difficult to express complex system and implement the whole optimal control.A kind of method of fuzzy identification based on improving clustering algorithm is proposed in connection with the traditional Fuzzy C-Means clustering algorithm＇s defects which were sensitive to the initial value and unable to determine the optimum rule numbers.The initial clustering centers are decided by the subtractive clustering and the validity function,and the final clustering centers are found by a global Fuzzy C-Means clustering algorithm,and then a closest adjacent area method is used to determine the suitable area radius,and finally the T-S model is built with weighted recursive least-square algorithm.In this paper,the simulation of a temerapure system of resistance furnace can illustrate that the method is accurate and effective.

关 键 词：模糊C-均值聚类 T-S模型 模糊辨识 温度控制系统 电阻炉 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]