CD-BSMOTE Based D-S Evidence Fusion Transformer Fault Diagnosis
Aiming at the solving the unbalanced characteristics of the dissolved gas data set in transformer oil on the fault diagnosis results,a transformer fault diagnosis model is proposed based on the fusion of critical removal improved boundary synthesis minority class oversampling algorithm equalized data set and Pearson conflict distance improved D-S evidence.Firstly,the minority class samples are equalized,and the samples in critical position are removed according to K-means clustering results.Secondly,the fault diagnosis model of gradient boost decision tree,random forest,and BP neural network is built to realize the preliminary diagnosis of transformer faults.Then,Pearson conflict distance is ap-plied to improve the D-S evidence fusion model to realize the fusion decision of preliminary diagnosis results.Finally,af-ter analyzing the cases,the precision rate of the diagnosis results reached 92.65%.The results show that the proposed model can effectively eliminate the influence of data imbalance on the diagnostic results and improve the fault diagnosis precision.
NETL
NSTL
万方数据
