基于自组织竞争优化结构的混沌系统建模

Chaotic System Modeling Based on Fuzzy Neural Network Model and Self-Organization Competition Network

针对混沌系统建模引入模糊神经网络模型时隶属函数不具有自适应性,系统模糊规则数的确定有一定的人为主观性的问题,本文对模型参数进行遗传退火算法优化,然后利用自组织竞争网络优化模型结构,使模型具有最佳结构即最简单的模糊规则数;再对有最佳结构的模型进行参数优化,得到具有最佳结构和参数的建模模型。以一维的Logistic系统、二维的Henon系统和Mackey-Glass混沌时间序列为例进行仿真分析,结果表明模型能够拟合原混沌系统,精度良好而且结构最简。

A fuzzy neural network model is proposed to identify chaotic system. Weight matrices and bias vectors of the model are membership function parameters. The algorithms of GAAnnealing strategy optimize the model parameters. Then self-organization competition network is used to optimize the model structure ; the new model parameters are optimized again, and the fuzzy neural network model having the simplest structure and best parameters is obtained. Simulations for model chaotic systems of Logistic system and Henon system and Mackey-Glass system show that the models can approach original systems and have advantages of simple fuzzy rules and good precision.