基于神经网络的变压器全局故障诊断系统

Global Fault Diagnosis Method of Traction Transformer Based on Neural Network

在传统分析方法的基础上,利用神经网络强有力的关系处理能力,研究提出变压器全局故障诊断方法。采用ART-2和BP两种神经网络进行数据分类,得到能较准确反映牵引变压器故障信息。采集来的数据聚类融合,形成故障诊断策略,给出变压器全局故障诊断模型。试验结果表明:该方法能够更好地分析变压器各类故障产生的原因,明确故障特征类型,避免用单一特征数据集诊断变压器故障带来的局限性,可以提高故障诊断的准确率。

Based on traditional method, global fault diagnosis method of transformer was introduced by using powerful re-lation processing capability of neural net-work. Through classify data by ART-2 and BP neural network to get the fault diagnosis information of transformer. The collected data classification and dealing with them in whole repository, forming fault diagnosis tactics in detail, composing complete control system, Global fault diagnosis model was established. The ex-periment result indicates that this method can better analyze the causes for different faults of ttransformer, ascertain the type of fault characteristics, avoid the localization to transformer fault diagnosis by single characteristics data set, and thus improve the precision of fault diagnosis.