基于VPMCD的变压器故障诊断方法

摘要

针对传统的变压器故障诊断方法诊断速度低、诊断精度差等缺点，设计了基于变量预测模型的模式识别方法（VariabIe Predictive ModeI Based CIass Discriminate，VPMCD）来实现变压器故障诊断。首先提取训练样本的特征值，其次针对特征值之间的内在关系选择合适的变量预测模型，并通过最小二乘法估计模型参数，然后将建立的模型运用到测试样本中检验它对变压器故障的分类和诊断能力。实验证明，与径向基神经网络相比，基于变量预测模型的模式识别方法具有诊断速度快，误诊率低等优点。%A method of the vriabIe predictive modeI based cIass discriminate (VPMCD) is designed to achieve the trans-former fauIt diagnosis by using the programming pIatform MatIab 2010b in this paper.FristIy,extract the characteristic vaIues of the training sampIe.Then,seIect the appropriate variabIe predictive modeIs on account of the inner reIationship between characteristic vaIues and estimate the modeI parameters through the Ieast square method. LastIy,appIy these modeIs to the testing sampIe for testing its sorting and diagnostic capabiIities of the transformer fauIt.The experiments reveaIed that com-pared with the radiaI basis function neuraI network.