基于改进Fisher Ratio的RBF网络在油浸变压器故障诊断中的应用  

Application of RBF Network with Improved Fisher Ratio to Default Diagnosis of Oil-immersed Transformer

作　　者：孙晓云[1] 刘东辉[1] 梁永春[1] 李爱华[1] 

机构地区：[1]河北科技大学,河北石家庄050054

出　　处：《高压电器》2009年第2期73-76,共4页High Voltage Apparatus

基　　金：河北省自然科学基金项目(F2007000636)资助

摘　　要：针对Fisher Ratio不能使网络结构得到最大优化这一缺点,提出应用改进Fisher Ratio方法选择RBF网络隐层节点数及径向基函数中心,并采用改进的Fisher比算法完成基于DGA的油浸变压器故障诊断。结果表明,该方法能极大地简化网络结构,提高了分类能力和收敛精度。与改良电协研法相比,具有较好的故障识别效果。An improved Fisher Ratio method is presented for the default diagnosis of oil-immersed transformer based on the dissolved gas analysis. Owing to the disadvantages of Fisher Ratio method in RBF network structure optimization, this paper adopts the improved Fisher Ratio method to seleet the RBF centers and to optimize the network structure. The results show that the neural network structure is simplified significantly, the convergence precision and classification capacity are improved. Compared with the improved three Ratio ns method, this method has better faults identification.

关 键 词：改进Fisher Ratio方法 油浸变压器 故障诊断 结构优化 

分 类 号：TM411[电气工程—电器]