一种新的TS模型辨识算法-《集美大学学报（自然科学版）》

文章信息/Info

Author(s):: LIN Mei-jiao¹; CHEN Shui-li²; (1.College of Mathematics and Computer ScienceFuzhou UniversityFuzhou 350108China2.School of ScienceJimei University Xiamen 361021China)

Keywords:: TakagiSugeno model identification ; Mountain C-Regression method; MCR; modified GK algorithm; Adaptive Particle Swarm Optimization; APSO

摘要:: 提出一种新的TS模型辨识算法.该算法思想：首先采用MCR算法（Mountain CRegression method）自动确定聚类数目和初始聚类中心，然后采用改进的GK（GustafonKessl）聚类算法得到最优的划分矩阵，再根据最优划分矩阵计算系统前件参数的最优值，最后用自适应粒子群优化算法（Adaptive Particle Swarm Optimization,APSO）对后件参数进行优化.此辨识算法能够用较少的规则数描述给定的未知系统，并且容易实现.仿真实验表明该算法能够实现非线性系统的辨识，并且可获得相对高的精度

Abstract:: In this paper，a novel TS model identification algorithm is proposed.The identification algorithm is on the base of the following ideas ：Firstly，the Mountain C-Regression method (MCR) is used to automatically identify the number of clusters and initial cluster center.Secondly，the modified Gustafson-Kessl (GK) algorithm is used to obtain an optimal input-output space fuzzy partition matrix which provids the values of premise parameters.Finally，Adaptive Particle Swarm Optimization (APSO) algorithm is adopted to precisely adjust consequent parameters.It can express a given unknown system with a small number of fuzzy rules and is easy to implement.The simulation results show the proposed algorithm realizes the identification of the nonlinear system with relative high accuracy.