粒子群优化的广义T-S模糊模型参数学习方法

Parameters Learning Approach for Generalized Takagi-Sugeno Fuzzy Model Using Particle Swarm Optimization

提出了一种基于粒子群优化的广义T-S模糊模型参数学习方法。该方法用离散二进制微粒位置表示模型的结构参数,用普通微粒位置表示模型规则中模糊集隶属函数的参数;这两种微粒位置联合体构成一个模型完整的前件参数集。每一学习循环分两步,前一步用粒子群进化迭代调整所有前件参数,后一步用正交最小二乘法估计后件参数。该方法不需任何先验知识,运算量小,能产生紧凑的模糊模型。非线性动态系统模糊建模的数字仿真说明了该方法的有效性。

A parameters learning approach for generalized takagi-sugeno （T-S） fuzzy model is proposed in this paper on the base of analysis of generalized T-S Fuzzy model. The structural parameters of the approach are denoted by the position of discrete binary particles and the parameters of membership function in the approach are denoted by the position of ordinary particles. The combination of positions of the two kind of particles composes complete premise parameters set of a model. A learning cycle consists of two phases： In＇st, all reasoning parameters are adjusted by evolutionary iteration of particle swarm; second, all consequent parameters are estimated through orthogonal least square error algorithm. The method requests scarcely any previous information about objects, take less calculating time, and is able to obtain compact fuzzy model. The simulation result shows the validity of the approach.