基于机器学习的水电站辅机系统设备故障诊断方法  

Fault diagnosis method of auxiliary system equipment of hydropower station based on machine learning

作　　者：付恩狄 罗勇 莫理 张勇 高菘 FU Endi;LUO Yong;MO li;ZHANG Yong;GAO Song(China Southern Power Grid Peak Frequency Modulation Power Generation Co.,LTD.,Guangzhou 510000,China)

机构地区：[1]南方电网调峰调频发电有限公司,广州510000

出　　处：《自动化与仪器仪表》2023年第10期296-299,共4页Automation & Instrumentation

基　　金：南方电网调峰调频发电公司科技项目资助(STKJXM20190162)。

摘　　要：为了提升水电站设备运行的智能化水平,研究基于机器学习的水电站辅机系统设备故障诊断方法、西部修试公司在调峰调频公司两个中心(数据中心、应用中心)前期研究成果下,开发、利用日臻完善的电力信息化平台获取运行设备状态和工况信息,从大数据本身内在规律分析的角度研究设备状态演变的关联关系和发展趋势,构建天二公司基于机器学习的辅助系统设备状态监测算法及数据分析模型框架,为设备稳定、安全、经济运行提供判断依据,为设备检修、退役决策提供支持。本项目选取了固有模态函数能量熵方法提取水电站辅机系统设备状态特征,该方法利用快速集合经验模态分解方法快速分解水电站辅机系统设备的振动信号,计算分解后振动信号的模态分量能量熵,所获取的模态分量能量熵即水电站辅机系统设备的状态特征,设置所提取的水电站辅机系统设备的状态特征作为模糊支持向量机模型的输入样本,模糊支持向量机模型的输出结果即水电站辅机系统设备的故障诊断结果。实验结果表明,该方法可以有效诊断水电站辅机系统设备故障,故障诊断准确率高于98.5%。In order to improve the intelligent level of hydropower station equipment operation, the fault diagnosis method of hydropower station auxiliary equipment system based on machine learning is studied. Based on the preliminary research results of the two centers (data center and application center) of the Peak shaving and frequency modulation company, the Western Repair and Test Company has developed and utilized the increasingly perfect electric power information platform to obtain the operating equipment status and working condition information, and studied the correlation and development trend of equipment status evolution from the perspective of the internal law analysis of big data itself, The equipment condition monitoring algorithm and data analysis model framework of the auxiliary system based on machine learning of Tian’er Company are constructed to provide judgment basis for stable, safe and economic operation of equipment and support for equipment maintenance and retirement decisions. In this project, the natural mode function energy entropy method is selected to extract the state characteristics of the equipment of the auxiliary machinery system of the hydropower station. This method uses the fast aggregation empirical mode decomposition method to quickly decompose the vibration signal of the auxiliary machinery system equipment of the hydropower station, calculate the energy entropy of the modal component of the decomposed vibration signal, and the obtained energy entropy of the modal component is the state characteristics of the auxiliary machinery system equipment of the hydropower station, Set the extracted state characteristics of the auxiliary system equipment of the hydropower station as the input sample of the fuzzy support vector machine model, and the output result of the fuzzy support vector machine model is the fault diagnosis result of the auxiliary system equipment of the hydropower station. The experimental results show that the method can effectively diagnose the faults of auxiliary equ

关 键 词：机器学习 水电站 辅机系统 设备故障诊断 

分 类 号：TV736[水利工程—水利水电工程]