基于主成分分析和基因表达式程序设计的变压器故障诊断

摘要

A kind of Gene Expression Programming algorithm (GEP) based on Principal Component Analysis (PC A) is proposed and used for transformer fault diagnosis. The application of Principal Component Analysis can reduce the dimension of the feature vectors and eliminate the irrelevance between vectors so as to decrease the computational complexity of the diagnosis classifier and increase the training and testing accuracy. Then the obtained new sample data are trained using GEP algorithm so as to construct transformer fault diagnosis model. 170 groups of the transformer DGA data which can reflect the variety of the faults without redundancy are used to study and get the GEP classifier, while the other 130 instances are diagnosed by the GEP classifier. The experiment shows that the proposed algorithm has obviously higher diagnostic accuracy and speed than solely using GP or GEP.This work is supported by Natural Science Foundation of Hebei Province (No. E2009001392).%提出一种基于主成分分析的基因表达式程序设计算法,并将其应用于变压器故障诊断中.用主成分分析对原数据进行一系列变换,可降低特征向量的维数,并消除向量间的不相关性；从而减小了故障分类器的计算复杂度,提高训练及测试的精度.然后将得到的新样本数据用基因表达式程序设计算法进行训练,构建变压器故障的诊断模型.利用该诊断模型对170组能反映出各种故障而又不冗余的变压器DGA数据进行学习,对另外1 30个实例进行诊断,取得了很好的效果.实验表明,所采用的算法比单独使用遗传规划或基因表达式程序设计具有更高的诊断精度和稳定性.