【作者】 田建艳； 武增懿； 韩肖清；

【Author】 TIAN Jian-yan~1,WU Zeng-yi~1,HAN Xiao-qing~2 (1.College of Information Engineering,Taiyuan University of Technology,Taiyuan 030024,China; 2.College of Electrical and Power Engineering,Taiyuan University of Technology,Taiyuan 030024,China)

【机构】 太原理工大学信息工程学院； 太原理工大学电气与动力工程学院；

【摘要】 针对最近邻聚类学习算法不能很好地反映学习样本的分布规律,提出采用负梯度下降法在线修正网络的中心、宽度和权值的最近邻聚类学习算法,使径向基函数神经网络不仅可以自动确定网络结构,而且可以自动适应学习样本的分布规律,从而有效地提高了径向基函数神经网络的学习精度。将这种改进算法应用于电力系统断路器故障诊断,仿真结果表明了该算法的有效性。更多还原

【Abstract】 In view of the problem that the nearest neighbor-clustering algorithm can not reflect the distribution of study samples,an improved learning algorithm is proposed that the center,width and weight are adjusted on-line using negative gradient descent.RBF neural network can not only determine the network structure but also adapt to the distribution of learning samples automatically.Thus the learning accuracy of RBF neural network is improved effectively.A large number of simulation results show that the improved algorithm is effective in fault diagnosis of circuit breakers.更多还原