径向基函数神经网络在电力变压器故障诊断中的应用  

Application of Radical Basis Function Neural Network in Diagnosis of Power Transformers Fault

作　　者：麻闽政[1] 

机构地区：[1]广东省电力设计研究院,广东广州510663

出　　处：《广东电力》2012年第1期80-83,共4页Guangdong Electric Power

摘　　要：基于油中溶解气体分析法,采用径向基函数(radical basis function,RBF)神经网络模型对电力变压器进行故障诊断。为了提高诊断模型的辨识精度,分两步对RBF神经网络的模型参数进行辨识:首先采用减聚类算法确定RBF神经网络隐含层基函数的中心点,然后采用量子粒子群优化(quantum-behaved particle swarm opti-mization,QPSO)算法求解基函数的宽度以及隐含层与输出层的连接权重。仿真实验结果表明,该方法的故障诊断正确率较高,达90.67%。Radical basis function （RBF） neural network is adopted to diagnose on faults of power transformer on the basis of DGA method. Model parameters of RBF neural network are identified by two steps so as to improve the identification accu- racy： subtractive clustering method is adopted first to determine the center of basis function of hidden layer of RBF neural network, whereafter quantum-behaved particle swarm optimization （QPSO） algorithm is used to solve function width and connection weight between hidden layer and output layer. The simulation experiment shows that the method is of high accu- racy in fault diagnosis and the accuracy is as high as 90.67%.

关 键 词：电力变压器 故障诊断 油中溶解气体法 径向基函数神经网络 量子粒子群优化算法 

分 类 号：TM726[电气工程—电力系统及自动化] TM853