基于VMD-XGBoost模型及因果特征选取的汽轮发电机组振动信号预测技术研究  

Research on Vibration Signal Prediction of Steam Turbine Generator Unit Based on VMD-XGBoost Model and Causal Feature Selection

作　　者：陈宇豪 杨为民 郭瑞[1] 姜虓 刘振祥 谭平 CHEN Yu-hao;YANG Wei-min;GUO Rui;JIANG Xiao;LIU Zhen-xiang;TAN Ping(National Engineering Research Center of Power Generation Control and Safety,Southeast University,Nanjing 210096,China;State Grid Fujian Electric Power Research Institute,Fuzhou 350007,China)

机构地区：[1]大型发电装备安全运行与智能测控国家工程研究中心,东南大学,南京210096 [2]国网福建电力科学研究院,福州350007

出　　处：《汽轮机技术》2024年第3期221-224,228,共5页Turbine Technology

基　　金：国家重点研发计划课题(2022YFB4100403);机炉关键部件损伤机理及动态均衡评价体系。

摘　　要：“双碳”目标下,我国能源格局产生深刻变化,对汽轮机发电机安全稳定运行的要求进一步提高,深入挖掘分析海量运行数据有助于机组运行状态的评估及预测。提出构建汽轮发电机组参数因果关系网络探究参数间的因果关系,利用VMD算法分解振动信号并搭建XGBoost预测模型对各信号分量进行预测,叠加各信号分量的预测值以得到振动信号的预测结果。利用国内某1000MW汽轮发电机组运行数据对所提模型进行论证实验,结果表明本文所提模型有较高预测精度。Under the target of"double carbon",the energy pattern of Chinahas changed profoundly,the requirement for the safe and stable operation of steam turbine generator unit has been further improved.In-depth mining and analysis of massive operation data is helpful to the evaluation and prediction of unit operation state.The paper proposes to construct the causal network of steam turbine generator unit parameters to explore the causal relationship between parameters and using VMD algorithm to decompose the vibration signal and build the XGBoost prediction model to predict each signal component,the predicted values of each signal component are superimposed to obtain the prediction results of the vibration signal.The operation data of a 1000MW steam turbine generator unit in China are used to demonstrate the proposed model and the results show that the proposed model has high prediction accuracy.

关 键 词：汽轮发电机组 轴系振动 趋势预测 因果发现 数据驱动 变分模态分解 极端梯度提升 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TM311[自动化与计算机技术—控制科学与工程]