摘要
发电机作为整个风电机组的核心部件之一,其能否正常运行将严重影响风电机组的持续发电。利用集成机器学习算法中的梯度提升算法XGBoost对风电机组发电机故障监测预警进行研究。首先,提取数据采集与监控(SCADA)系统数据库中风电机组在并网状态下的正常运行大数据,对缺失、异常数据进行预处理后,结合运维专家经验利用XGBoost算法筛选出关键特征变量;然后经过训练和参数调整,建立最优风电机组发电机故障监测预警模型;通过对照研究发现,XGBoost算法对风电机组发电机进行故障监测预警的效果优于随机森林算法和CatBoost算法;最后利用关键特征变量重要性排序作为风电机组发电机故障诊断与定位的参考。
As one of the core parts of the whole wind turbine,the normal operation of generator seriously affects the continuous generation of wind turbine.In this paper,gradient boosting algorithm XGBoost in the integrated machine learning algorithm is used to study the fault monitoring and early warning of wind turbine generator.Firstly,the big data of normal operation of the wind turbine in the grid connected state in the SCADA system database is extracted.After preprocessing the missing and abnormal data,the key characteristic variables are selected by using experience of operation and maintenance experts and XGBoost algorithm.Secondly,the optimal wind turbine generator fault monitoring and early warning model is established by training and parameters adjustment.The comparative study shows that XGBoost algorithm is better than random forest algorithm and CatBoost algorithm in fault monitoring and early warning of wind turbine generator.Finally,importance ranking of key characteristic variables is used as a reference for fault diagnosis and fault location of wind turbine generator.
作者
苏国梁
汪健冬
付恩强
赵娟娟
黄文广
刘广臣
Su Guoliang;Wang Jiandong;Fu Enqiang;Zhao Juanjuan;Huang Wenguang;Liu Guangchen(GD Power Development Co.,Ltd.Inner Mongolia New Energy,Hohhot 010020,China;School of Mathematics and Statistical Science,Ludong University,Yantai 264000,China;Huafeng Data(Shenzhen)Co.,Ltd.,Shenzhen 518110,China)
出处
《太阳能》
2021年第9期78-84,共7页
Solar Energy
基金
教育部产学合作协同育人项目(201901137017,201801034031,201802257026)
山东省高等学校教学研究与改革面上项目(M2018X066)
国家级大学生创新创业训练计划项目(S202010451021)。