模糊竞争神经网络在线辨识及在CSTR中的应用  

On-line Identification of Fuzzy Competitive Neural Networks and Its Application to CSTR Systems

作　　者：江善和[1] 潘康生[1] 余春日[1] 

机构地区：[1]安庆师范学院物理与电气工程学院,安徽安庆246011

出　　处：《计算机仿真》2005年第4期190-193,共4页Computer Simulation

摘　　要：基于改进的T-S模糊模型构造了一种自适应模糊竞争神经网络模型(FCNN),给出了网络的连接结构和学习算法。它依据模糊竞争学习算法确定系统的模糊空间和模糊规则数,得出每个样本对每条规则的适用程度,并利用卡尔曼滤波算法在线辨识FCNN的后件参数。将其应用于化工过程连续搅拌反应器(CSTR)的建模中,仿真结果表明,FCNN具有结构简洁、收敛速度快、辨识精度高等特点,可当作复杂系统建模的一种有效手段。Based on modified T-S fuzzy model, this paper introduces an adaptive Fuzzy Competitive Neural Network model.Also,the neural link structure and the le arning algorithm are given.It determines the fuzzy space structure of system and the number of fuzzy rules based on fuzzy competitive learning algorithm and obt ains the fitness degree of every rule contrast to every sample.Further, the para meters of FCNN are on-line identified by means of Kalman filtering algorithm.Thi s approach has been applied to the model of continuous stirred tank reactor(CSTR ). The simulation results demonstrate the proposed FCNN has simple model str ucture, fast convergence and high precision of identification.It is turn ed out an effective approach for the modeling of a complicated system.

关 键 词：模型 模糊竞争神经网络 模糊竞争学习 模糊辨识 连续搅拌反应器 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]