摘要
针对变压器故障诊断油中溶解气体的故障样本较少以及特征气体与变压器故障之间建立的非线性分类关系存在着一定的不确定性问题,提出了一种基于SVM和D-S证据理论多证据体相融合的故障诊断新方法。首先,将SVM的标准输出直接拟合Sigmoid函数得到SVM输出结果,然后,将其进行信息融合。在Visual Studio2008环境结合LIBSVM软件包对DGA数据进行试验,结果表明,该方法在变压器故障诊断中取得了较高的正判率,同时诊断输出结果包含更多信息量。
Aiming at the lack of enough oil dissolved gas samples in transformer fault diagnosis and the uncertainty of nonlinear classification relationship established between the characteristic gas and transformer failures, a new fault diagnosis method based on evidence of bulk integration of SVM and D - S evidence theory is supported. First, the standard output of the SVM directly fits with the Sigmoid function, from which the SVM output of posterior probability is acquired, and then more SVM output is used to information fusion. With the help of LIBSVM software package the DGA data is tested under Visual Studio 2008. The experimental results show that the method that is applied to transformer fault diagnosis obtains higher accuracy and diagnostic result with more information
出处
《黑龙江电力》
CAS
2013年第2期124-126,129,共4页
Heilongjiang Electric Power
关键词
变压器
故障诊断
证据理论
信息融合
transformer
fault diagnosis
evidence theory
information fusion
