Small Sample Transformer Fault Diagnosis Method Based on DGA and Enhanced SMOTE
The transformer diagnosis methods are affected obviously by sample numbers and data quality.Moreover,the current synthesized sample quality of traditional small sample learning can't meet the requirements for actual applications,causing difficulty in transformer diagnosis with few data by intelligent diagnosis algorithms.To solve these above problems,this paper proposes a transformer diagnosis model based on enhanced synthetic minority oversampling technique(SMOTE)and deep learning.Firstly,this method uses SMOTE to augment the training set.Secondly,based on cosine similarity,an optimum selecting method of the synthesized samples to enhance the quality.Thirdly,a convolutional neural network model is used for classification prediction.The paper uses measured data to conduct analysis and validation,and compare the proposed model with traditional methods.The results indicate that the proposed method has improved the performance of fault diagnosis.
万方数据
维普
