基于TS模糊模型的广义模糊神经网络的系统辨识研究.pdf

th Proceedings of the 25 Chinese Control Conference 7-11 August, 2006, Harbin, Heilongjiang 基于T-S模糊模型的广义模糊神经网络的系统辨识 蔚瑞华，徐立鸿 同济大学电子与信息工程学院上海，200092 E-mail: laurrywei@ 摘 要: 广义模糊神经网络（GFNN，Generalized Fuzzy Neural Network ）集成了模糊逻辑系统和神经网络的优 点，其具有的参数自学习和结构自调整算法给非线性系统建模提供了强大的工具。本文针对GFNN （Generalized Fuzzy Neural Network ）不易于传统控制器设计相结合的缺点，提出了基于T-S模糊模型的广义模 糊神经网络（TS-GFNN，Takagi-Sugeno Gneralized Fuzzy Neural Network ），改进了原有GFNN 的结构，并针对 新的结构提出了基于聚类的离线结构和参数辨识方法。此外，为了更好发挥GFNN在线辨识的特点，以更好地 满足系统实时控制的要求，本文还提出了相应的在线结构自组织和参数自调整学习算法。仿真表明，本文提出 的TS-GFNN对于非线性系统或者线性时变系统可以以更精简的结构（也即更少的规则）进行辨识，且逼近精度 较高。 关键词: 广义模糊神经网络GFNN, T-S 模糊模型，非线性系统辨识，在线辨识 Identification Based on TS-GFNN （Takagi-Sugeno Generalized Fuzzy Neural Network ） Ruihua Wei, Lihong Xu Department of Control Science Engineering, Tongji University, Shanghai, 200092 China E-mail: laurrywei@ Abstract: GFNN (Generalized Fuzzy Neural Network), which integrates the advantages of Neural Network into that of the Fuzzy Logic System, is a powerful method in the modeling of the nonlinear system. However, it is difficult for the GFNN to be used as a model in the traditional way of designing the controller, because GFNN is intrinsically nonlinear. A new design of GFNN based on T-S (Takagi-Sugeno) model and its corresponding off-line architecture and parameter identification algorithm is presented in the paper. In addition, to better use the on-line self-adjusting advantages of GFNN, the on-line architecture-self-organizing and parameter-self-learning algorithm is also presented. The on-line