In the field of transformer fault diagnosis

the extreme values and noise caused by the imbalance of a data set have seriously affected the fault recognition ability of the classification algorithm. Therefore

a method of power transformer fault sample equalization based on the support vector machine（SVM） synthetic minority over-sampling technique（SMOTE） algorithm is proposed

and machine learning is combined for fault diagnosis to effectively solve the problem of low overall-precision of transformer fault diagnosis under the unbalanced data set. Firstly

the traditional SMOTE algorithm and SVM SMOTE algorithm from the aspects of principle

characteristics

and application were researched and analyzed in detail Then

the gas concentration dissolved in transformer oil was taken as a sample set

and the transformer fault diagnosis model based on fault sample equalization was constructed. Finally

simulation study was carried out.The result shows that

compared with traditional transformer fault diagnosis

the evaluation indexes have been greatly improved after fault sample equalization with SVM SMOTE

and the αmacro-F1

namely

overall accuracy

has increased by nearly 18.9%. The research results prove that the proposed method can be adopted to effectively solve the problem of high missing report ratio of transformer fault sample under unbalanced data set

and outperform traditional classification algorithms with respect to accuracy.