基于VPMCD的变压器故障诊断方法

Transformer Fault Diagnostic Method Based on VPMCD

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(Variable Predictive Model Based Class 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 vriable predictive model based class discriminate（VPMCD） is designed to achieve the transformer fault diagnosis by using the programming platform Matlab 2010 b in this paper.Fristly,extract the characteristic values of the training sample.Then,select the appropriate variable predictive models on account of the inner relationship between characteristic values and estimate the model parameters through the least square method. Lastly,apply these models to the testing sample for testing its sorting and diagnostic capabilities of the transformer fault.The experiments revealed that compared with the radial basis function neural network.