引用本文:姚兰,肖建,王嵩,蒋玉莲.自组织区间二型模糊神经网络及其自适应学习算法[J].控制理论与应用,2013,30(6):785~791.[点击复制]
YAO Lan,XIAO Jian,WANG Song,JIANG Yu-lian.Interval type-2 fuzzy neural networks with self-organizing structure and adaptive learning algorithm[J].Control Theory and Technology,2013,30(6):785~791.[点击复制]
自组织区间二型模糊神经网络及其自适应学习算法
Interval type-2 fuzzy neural networks with self-organizing structure and adaptive learning algorithm
摘要点击 3116  全文点击 1899  投稿时间：2012-08-02  修订日期：2013-01-14
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2013.20849
  2013,30(6):785-791
中文关键词  自组织  区间二型模糊神经网络  梯度下降法  自适应学习算法
英文关键词  self-organizing  interval type-2 fuzzy neural networks  gradient-descent algorithm  adaptive learning algorithm
基金项目  国家自然科学基金资助项目(51177137); 国家自然科学基金重点资助项目(61134001).
作者单位E-mail
姚兰* 西南交通大学 电气工程学院
成都信息工程学院 控制工程学院 		dancyyao@163.com 
肖建 西南交通大学 电气工程学院  
王嵩 西南交通大学 电气工程学院  
蒋玉莲 西南交通大学 交通运输与物流学院  
中文摘要
      针对复杂不确定非线性系统的辨识问题, 提出一种基于聚类的自组织区间二型模糊神经网络学习算法. 首先采用具有两个不同加权参数的FCM算法对输入数据进行划分来获取规则前件的不确定均值, 同时结合聚类有效性标准确定模糊规则数目, 从而自动完成神经网络的结构辨识和规则前件参数辨识; 随后给出了基于梯度下降法和Lyapunov函数稳定收敛定理的规则后件权向量学习速率的自适应学习算法. 通过非线性系统辨识实例, 验证了该算法与其他方法相比具有更快的收敛速度和更高的逼近精度; 并且利用该算法建立了某市电力短期负荷预测模型, 结果表明该模型具有较高的预测精度, 泛化性能更佳.
英文摘要
      For the identification problems of nonlinear system with complex uncertainties, an interval type-2 fuzzy neural network with self organizing structure and learning algorithm is proposed. Firstly, the fuzzy c–means algorithm with two different weighting parameters are used to partition the input data to obtain the uncertainty means of rule antecedent; meanwhile, according to the cluster validity criterion, the number of fuzzy rules is determined. Thus, the structure and parameters of rule antecedent identification are automatically completed. Then, based on the gradient descent method and Lyapunov function stability convergence theorem, the adaptive learning algorithm for weight vectors of rules consequent is presented. Finally, the experimental results of nonlinear system identification indicate that the proposed algorithm has faster convergence rate and higher approximation precision than other algorithms. In addition, based on the power load data of a city, a short-term load-forecasting model is developed by the algorithm, which has higher prediction precision and better generalization.