基于遗传规划和数据归一化的变压器故障诊断

Oil-immersed Transformer Fault Diagnosis Based on PCA and GP

提出了一种基于数据归一化和遗传规划的分类模型,并将其应用于变压器故障诊断中。该模型以变压器的DGA数据为评价指标,先用数据归一化方法对原始数据进行预处理,以消除不同属性之间的量纲的影响;然后将处理后的数据用遗传规划算法进行训练,得到相应的分类器,并构建变压器故障诊断的二叉树模型。利用该诊断模型对121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100个实例进行诊断,取得了很好的效果,实验证明该模型是高效可行的。

A classification model based on data normalization and genetic programming is proposed, and it is ap- plied to fault diagnosis of transformers. The classification model used the DGA data of dissolved gas in the trans- former oil as the evaluation standard. Firstly, the data normalization method is used to preprocess the original data, so as to eliminate the dimensions＇ effects of different attributes. Then Genetic Programming algorithm is used for training the processed data and getting the corresponding classifier. At last the binomial transformer fault diagnosis model is built. In this paper, Section 121 groups of DGA data diagnostic model which can reflect the variety of fault without redundant transformer is studied and another 100 instances is used for the diagnosis. Experiments show that the model is efficient and feasible.