论文部分内容阅读
针对模糊神经网络结构设计问题及模糊集在语言描述上存在的不足,提出一种基于扩展的卡尔曼滤波(EKF)的自组织T-S模糊Elman网络,并推导了网络训练算法.分别采用递归最小二乘法和EKF对线性参数和非线性参数进行更新;基于模糊规则生成准则和误差下降率修剪策略实现了模糊规则的增删减.最后通过系统辨识和污水处理建模实验,表明了该算法在保证网络精度和泛化能力的同时,可以有效地简化网络结构.
In view of the structural design of fuzzy neural network and the shortcomings of fuzzy sets in language description, a self-organizing TS fuzzy Elman network based on Extended Kalman Filter (EKF) is proposed and the network training algorithm is derived. Multiplication and EKF to update the linear parameters and non-linear parameters.According to fuzzy rule generation rules and error reduction rate pruning strategy, the addition and deletion of fuzzy rules are realized.Finally, the system identification and sewage treatment modeling experiments show that the algorithm guarantees Network accuracy and generalization capabilities at the same time, can effectively simplify the network structure.