摘要
根据对变压器常见故障原因及危害程度的分析，构造了电力变压器故障诊断决策树．其判定级别为自上而下，而决策树的每一叶都对应着一种具体的故障模式，并在决策树的不同分支中选用不同的神经网络单元模块作为基本分类器，建立组合神经网络模型，实现了对故障的多分辨识别．该方法克服了以往单神经网络模型在结构复杂性和学习难于收敛方面的不足，大大提高了故障分析的准确度．
A decision tree method is presented using the synthetic analysis of the cause and injury extent of transformer malfunction. The identification order is directed downward such that each leaf serves as a malfunction scenario. While each branch corresponds to different artificial neural network (ANN), a multi resolution identification of transformer malfunction can thus be constructed. Convergence of ANN and accuracy of diagnosis are improved.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
1999年第6期11-16,共6页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金
关键词
电力变压器
神经网络
故障诊断
决策树
power transformer
dissolved gas analysis
artificial neural network
fault diagnosis
decision tree
