复杂系统的遗传-模糊建模方法  

A Novel Genetic-fuzzy Modeling Method for Complex Systems

作　　者：黄自元[1] 甄兰兰 费敏锐[1] 沈昱明[3] 

机构地区：[1]上海大学机自学院,上海200072 [2]公安部上海消防研究所,上海200438 [3]上海理工大学光电学院,上海200093

出　　处：《系统仿真学报》2003年第9期1285-1287,共3页Journal of System Simulation

基　　金：上海市教育发展基金会曙光计划项目(99SG37);上海市纳米科技专项(0214nm090)

摘　　要：针对复杂系统的模糊建模问题,提出了一种遗传-模糊建模新方法。首先,利用竞争学习算法对输入空间进行自适应聚类,基于聚类结果提取模糊模型的规则前件隶属函数参数,采用局部最小二乘法求得规则后件参数,从而初步建立起系统的T-S模糊模型。然后,对规则前、后件参数进行编码,借助于实值编码遗传算法优化模糊系统。最后,数字仿真结果验证了算法的可行性和有效性。In this paper, we present a novel genetic-fuzzy modeling method for complex systems. Firstly, the input space is clustered by competitive learning algorithm. With the result, we get the membership function of the antecedent fuzzy sets. The consequent parameters of each individual rule are obtained as a local least squares estimate. Thus, the coarse T-S fuzzy model is taken. Secondly, the parameters of rules are encoded and the fuzzy systems are optimized by real-coded genetic algorithm. At last, digital simulation is performed to demonstrate the validity of our approach.

关 键 词：竞争学习 T-S模糊模型 实值编码遗传算法 最小二乘法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]