基于XGBoost模型的风电机组故障检测  

Fault detection of the wind turbine based on XGBoost model

作　　者：马康原 李忠虎[1] 王金明[1] 杨立清[1] 芦建文 MA Kangyuan;LI Zhonghu;WANG Jinming;YANG Liqing;LU Jianwen(Information Engineering School,Inner Mongolia University of Science and Technology,Baotou 014010,China;Inner Mongolia Xinlian IT Co.Ltd.,Baotou 014010,China)

机构地区：[1]内蒙古科技大学信息工程学院,内蒙古包头014010 [2]内蒙古新联信息产业有限责任公司,内蒙古包头014010

出　　处：《内蒙古科技大学学报》2023年第4期349-353,共5页Journal of Inner Mongolia University of Science and Technology

基　　金：内蒙古自治区科技计划资助项目(2021GG0433).

摘　　要：针对风电装机规模快速增长与风电场运维技术相对滞后之间的矛盾问题,研究利用SCADA数据实现完全基于数据驱动的风电机组故障检测具有重要意义.从数据驱动的角度出发,利用DBSCAN算法和正常功率区间估计完成SCADA数据的筛选,所形成的健康数据与SCADA采集到风电机组故障数据共同构成建模所需的数据库.然后通过MIC算法完成特征参量的选择,并输入模型完成XGBoost模型训练.最后进行模型验证,表明该方法可及时确认故障信息,实现故障的快速检测.In view of the contradiction between the rapid growth of wind power installed scale and the relative lag of wind farm operation and maintenance technology,it is of great significance to study and use SCADA data to realize a completely clata-driven fault detection of wind turbime.From the perspective of data-driven,the DBSCAN algorithm and normal power interval estimation were used to complete the screening of SCADA data.The resuttant health data and the fault data collected by SCADA formed the database required for modeling.Then the characteristic parameters were selected using MIC algorithm,and the model was input to complete the XGBoost model training.Finally,model verification shows that the method can confirm the fault information in time and realize the fault detection.

关 键 词：风电机组 故障检测 机器学习 XGBoost算法 

分 类 号：TM315[电气工程—电机]