Fuzzy Neural Network Based on Improved T-S Model and Its Application

详细信息    查看全文

• 作者：Zhiwei Huang ; Jianzhong Zhou ; Chaoshun Li ; Fengpan Li ; Yongchuan Zhang
• 关键词：Fuzzy neural network ; T ; S model ; Adaptive learning rate ; Momentum factor ; Function approximation ; System identification
• 刊名：Lecture Notes in Computer Science
• 出版年：2009
• 出版时间：2009
• 年：2009
• 卷：5552
• 期：1
• 页码：155-164
• 全文大小：217.6 KB

文摘

Traditional fuzzy neural networks have many problems like long duration of training learning algorithm, slow convergence and possibility of getting into local dinky value in the network. In view of all these above, this paper proposes an improved training algorithm in fuzzy neural network which takes vantage of the excellent learning and expression ability of fuzzy neural networks based on T-S (Takagi-Sugeno) model. Improved T-S model introduces adaptive learning rate η and momentum factor γ into the learning algorithm which helps stabilize the network and makes acceleration weights of three adjusting parameters, including network membership function’s central value C _ij and width σ _ij and connection weight of output layer y _kj , decrease along the reductive direction. Results of a simulation test with a two-dimensional nonlinear function show that improved network has high convergence rate and good generalization capability. Finally, T-S fuzzy model of hydro-turbine governing system based on the fuzzy neural network is presented, of which the results has indicated that the model has/its higher precision and good generalization ability.