基于特征选择和XGBoost的风电机组故障诊断  

Fault diagnosis for wind turbine based on Random Forest and XGBoost

作　　者：靳志杰 霍志红[1] 许昌[1] 郭宏宇 周华建 Jin Zhijie;Huo Zhihong;Xu Chang;Guo Hongyu;Zhou Huajian(College of Energy and Electrical Engineering,Hohai University,Nanjing 211100,China)

机构地区：[1]河海大学能源与电气学院,江苏南京211100

出　　处：《可再生能源》2021年第3期353-358,共6页Renewable Energy Resources

基　　金：国家自然科学基金项目(U1865101);江苏省青年基金项目(BK20180505)。

摘　　要：随着风电规模的不断增加,风电机组的运行维护成为研究的热点。针对风电机组的故障诊断问题,文章提出了一种基于特征选择和XGBoost算法的故障诊断方法。该方法采用随机森林的袋外估计进行特征选择,降低了特征选择过程的主观性;以XGBoost算法为基础搭建诊断模型,采用网格搜索和交叉验证对算法进行参数优化。以风电场SCADA实测数据对所提方法进行验证,通过准确率、AUC值等指标将文章所提方法与传统机器学习算法的诊断结果进行对比。对比结果表明,文章提出的方法比传统机器学习算法的预测准确率更高,可用于风电机组故障诊断的工程中。This paper presents a fault diagnosis model based on Random Forest and eXtreme Gradient Boosting algorithm to reduce the wind turbine failure rate.The out-of-bag estimation of Random Forest was used to select the feature parameters that are highly correlated with common faults to replace the way of subjective feature-selection based on prior knowledge.We trained the eXtreme Gradient Boosting algorithm to construct the fault diagnosis model and optimize its parameters by using grid search algorithm and cross-validation.The model was verified by the actual operating data of a wind farm and it was compared with the traditional machine learning algorithms through the accuracy,AUC value and other indices.The experimental results show that the accuracy of eXtreme Gradient Boosting algorithm is higher than the traditional machine learning algorithm,so the fault diagnosis model can be applied to the engineering application of wind turbine fault diagnosis.

关 键 词：风电机组 SCADA数据 XGBoost 故障诊断 

分 类 号：TK83[动力工程及工程热物理—流体机械及工程]