基于XGBoost-LSTM的水轮机轴瓦温度预测  

Turbine bearing temperature prediction based on XGBoost-LSTM

作　　者：谈群 郜振亚 秦拯 苗洪雷[2] TAN Qun;GAO Zhenya;QIN Zheng;MIAO Honglei(College of Computer Science and Electronic Engineering,Hunan University,Changsha 410082,China;HNAC Technology Co.,Ltd.,Changsha 410205,China)

机构地区：[1]湖南大学信息科学与工程学院,湖南长沙410082 [2]华自科技股份有限公司,湖南长沙410205

出　　处：《水利水电快报》2023年第10期65-70,76,共7页Express Water Resources & Hydropower Information

基　　金：长沙市科技计划项目(kh2204007)。

摘　　要：为保障水轮机在工作状态下的安全运行,有必要对轴瓦温度进行预测研究,提出了一种基于XGBoost-LSTM的轴瓦温度预测模型,利用XGBoost进行特征选择,挑选出对轴瓦温度有影响的重要特征;利用LSTM进行时间序列分析,挖掘出特征的未来发展趋势,得到更加准确的预测结果。结果表明:特征选择后模型精度得到了一定程度提升,LSTM模型能够较好地预测出轴瓦温度的变化趋势,预测值与真实值的最大误差小于1℃,研究成果可为水轮机故障预测与健康管理系统的开发提供理论和技术支持。To ensure the safe operation of turbine under working condition,the prediction of bearing temperature of hydropower unit was studied.A temperature prediction model of bearing bush based on XGBoost-LSTM model was proposed.By using the XGBoost for feature selection,important features that have influence on the bearing bush temperature were picked out,and the LSTM was used for time series analysis.The developing trend of the features was extracted to obtain more accurate results.The results showed that the accuracy of the model was improved to some extent after feature selection.The LSTM model can better predict the change trend of bearing bush temperature,and the maximum error between the predicted value and the real value was less than one degree Celsius.The research results of this paper can provide theoretical and technical support for the development of prognostic and health management system of hydraulic turbine.

关 键 词：轴瓦温度 特征工程 XGBoost 时间序列 长短期记忆(LSTM) 

分 类 号：TK730.322[交通运输工程—轮机工程]