基于贝叶斯优化XGBoost算法的变压器故障诊断

Diagnosis on Transformer Fault Based on Bayesian Optimization XGBoost Algorithm

摘要: 为提升对高能放电等小样本故障诊断的敏感度，提出基于贝叶斯优化极端梯度提升算法（BO－XGBoost）的变压器故障诊断模型。分析了贝叶斯优化XGBoost算法的基本原理和基于该算法进行变压器故障诊断的流程，选取259组故障样本，探讨了该模型的具体应用，并将其与XGBoost、支持向量机（SVM）、随机森林（RF）、K邻近法（KNN）等模型进行对比。结果表明，BO－XGBoost模型在变压器故障诊断中的精度为98.08%，比前述模型的诊断精度分别提高了5.77%、27.42%、22.58%、19.5%。

Abstract: In order to improve the sensitivity of small sample fault diagnosis, such as high energy discharge, a transformer fault diagnosis model is proposed based on Bayesian optimization extreme gradient lifting algorithm (BO-XGBoost).The basic principle of Bayesian optimization XGBoost algorithm and the flow of transformer fault diagnosis based on this algorithm are analyzed.Two hundred and fifty-nine groups of fault samples are selected.The specific application of this model is discussed.The model is compared with XGBoost, Support Vector Machine (SVM), Random Forest (RF)and K proximity method (KNN).The results show that the accuracy of BO-XGBoost model in transformer fault diagnosis is 98.08%, which is 5.77%, 27.42%, 22.58% and 19.5% higher than that of the aforementioned model, respectively.