基于变压器故障分类的DGA特征提取

Feature Selection of DGA Data Based on Transformer Fault Classification

为提高诊断效果,利用DGA诊断电力变压器故障时,可从模式识别的角度出发,针对具体的分类模式,提取出能区别不同类别模式的“选择性”信息。选择和测试放电与过热故障、电路过热与磁路过热故障的气体特征的结果表明,根据不同的分类模式提取气体特征对提高故障识别效果有益。

From the standpoint of pattern recognition, this paper proposes that the selective information of feature of dissolved gas for distinguishing various fault patterns should be picked up based on fault classification. The features reflecting thermal or discharge fault and the ones reflecting thermal fault in electric circuit or magnetic circuit are selected respectively, and based on these features, the accuracy of diagnosis is tested. The results show that appropriate feature selection for dissolved gas in oil is helpful to fault pattern recognition for power transformer.