基于数据挖掘的循环流化床锅炉辅机故障预警  

Early Fault Warning of Auxiliary Equipment for Circulating Fluidized Bed Boilers Based on Data Mining

作　　者：张维[1,2] 刘吉臻 高明明[1] ZHANG Wei;LIU Jizhen;GAO Mingming(State Key Laboratory of Alternate Electric Power System with Renewable Energy Sources,North China Electric Power University,Beijing 102206,China;State Grid Information and Telecommunication Industry Group Co.,Ltd.,Beijing 102211,China)

机构地区：[1]华北电力大学新能源电力系统国家重点实验室,北京102206 [2]国网信息通信产业集团有限公司,北京102211

出　　处：《动力工程学报》2019年第10期826-833,共8页Journal of Chinese Society of Power Engineering

基　　金：国家重点研发计划资助项目(2016YFB0600205);中央高校基本科研业务费专项资金资助项目(2016MS28,2017XS072,2018ZD05)

摘　　要：为保障循环流化床机组辅助设备的安全运行,提出了一种基于模糊推理和滑动平均相似度的辅助设备故障预警方法。以改进模糊关联规则挖掘为基础的模糊推理被用于推断正常运行情况下设备状态参考值,采用测试数据与参考值之间的滑动平均相似度来判断辅机安全性的预警标准,采用层次分析方法来确定相似度函数中变量的权重,并将正常状态下的最低滑动平均相似度作为故障预警阈值,最后以某电厂300 MW循环流化床锅炉引风机为例进行验证。结果表明:当相似度低于故障预警阈值时,系统可发出警告报警和故障变量定位,因此所提方法能有效预测故障。To ensure the safe operation of auxiliary equipment for circulating fluidized bed(CFB) boilers, a method of early fault warning was proposed for the auxiliary equipment based on fuzzy reasoning and moving average similarity. The fuzzy reasoning based on improved fuzzy association rule mining was used to infer the state reference values of the equipment under normal operating conditions, and the moving average similarity between the test data and reference values was used as the evaluation criterion for judging the states of auxiliary equipment, while the weights of the variables in the similarity function were determined by the analytic hierarchy process. The lowest moving average similarity in normal operating state was then taken as the fault warning threshold, which was subsequently verified by the induced draft fan of a 300 MW CFB power plant. Results show that the system can issue an early fault warning and identify the fault location when the similarity is lower than the fault warning threshold, therefore the method proposed is believed to be effective in the prediction of relevant faults.

关 键 词：循环流化床 辅助设备 模糊关联规则挖掘 模糊推理 滑动平均相似度 

分 类 号：TK39[动力工程及工程热物理—热能工程]