基于监测数据的蓄能设备故障预测方法研究  

Research on fault prediction of energy storage equipment based on monitoring data

作　　者：曲晓峰 苗东旭 王树新 王振羽 甄宏 刘烁 QU Xiaofeng;MIAO Dongxu;WANG Shuxin;WANG Zhenyu;ZHEN Hong;LIU Shuo(Harbin Electric Group Co.,Ltd.,Harbin 150028,China;Fengman Dam Reconstruction Engineering Bureau,Jilin 132000,China;Harbin Steam Turbine Factory Co.,Ltd.,Harbin 150040,China;School of Computer Science and Technology,Harbin Institute of Technology,Harbin 150001,China)

机构地区：[1]哈尔滨电气集团有限公司,哈尔滨150028 [2]丰满大坝重建工程建设局,吉林132000 [3]哈尔滨汽轮机厂有限责任公司,哈尔滨150040 [4]哈尔滨工业大学计算机科学与技术学院,哈尔滨150001

出　　处：《智能计算机与应用》2019年第5期326-329,333,共5页Intelligent Computer and Applications

基　　金：2017年智能制造新模式应用项目:水力发电设备智能远程运维新模式

摘　　要：蓄能设备是发电厂削峰填谷的重要手段之一。针对发电厂蓄能设备故障种类和原因复杂,提出一种基于运行数据和机器学习方法的蓄能设备故障预测方法。该方法以管理系统监控的历史数据和实时数据为基础,首先对数据集进行特征提取和充放电周期识别,然后建立随机森林分类器,训练分类器模型参数,以实现蓄能设备运行过程中是否存在故障以及故障类别的预测。在实际运行数据中验证了所提方法能够在故障发生的早期有效识别故障类型。Energy storage equipment is one of the key devices for peak cutting in a power station.Aiming at the complex types and causes of energy storage device faults in power station,a fault assessment method based on profile data and machine learning method is proposed.In order to predict the faults and their types in the operation of energy storage equipment,based on profile historical data and real-time data,the feature extraction and charge-discharge cycle identification of profile data sets are carried out first.Then a support vector machine classifier is established and trained to obtain the model parameters.The results on the actual operation data of energy storage device verify that the proposed method can effectively identify the fault type in the early stage of fault occurrence.

关 键 词：蓄能设备 故障预测 机器学习 支持向量机 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]