基于XGBOOST算法的变压器故障诊断

Power Transformer Fault Diagnosis Based on XGBoost Algorithm

介绍了目前常用的变压器油中溶解气体分析方法的不足之处,针对油色谱在线监测系统产生的大量数据,尝试运用机器学习算法进行变压器故障诊断判别。由于变压器故障诊断中存在的小样本特点,运用一般的机器学习算法泛化能力差,通过运用XGBoost方法来拟合模型,并与几种常见的机器学习算法性能进行了比较,实验结果证明采用XGBoost提取特征的方法加上简单分类器可以达到很好的效果。

This paper introduces the shortcomings of the commonly used dissolved gas analysis methods in transformer oil. For the large amount of data generated by the oil chromatography online monitoring system, this paper attempts to use the machine learning algorithm to diagnose the transformer fault diagnosis. Due to the small sample characteristics of transformer fault diagnosis, the general machine learning algorithm is poor in generalization ability. This paper uses XGBoost method to fit the model and compares it with several common machine learning algorithm performances. The experimental results prove that XGBoost is adopted. It is proved that the method of extracting features add a simple classifier can achieve good results.