基于神经网络的核电厂设备关键指标故障预警系统开发  

Development of the Early Fault Warning System for Key Indicators of Nuclear Power Plant Equipment Based on Neural Network

作　　者：侯建飞 张勋 黄曦 司恒远 HOU Jianfei;ZHANG Xun;HUANG Xi;SI Hengyuan(State Key Laboratory of Nuclear Power Safety Technology and Equipment,China Nuclear Power Engineering Co.,Ltd.,Shenzhen of Guangdong Prov.518100,China;Guangxi Fangchenggang Nuclear Power Co.,Ltd.,Fangchenggang of Guangxi Prov.538000,China)

机构地区：[1]核电安全技术与装备全国重点实验室,中广核工程有限公司,广东深圳518100 [2]广西防城港核电有限公司,广西防城港538000

出　　处：《核科学与工程》2024年第5期1141-1147,共7页Nuclear Science and Engineering

摘　　要：为了避免设备的突然故障影响核电厂的安全性和经济性,基于核电厂的历史运行数据和神经网络算法,开发了一套核电厂设备关键指标预警系统,对设备的关键性能指标进行预测,并采用多种控制图耦合的方式,对预测值与实际测量值之间的偏差进行分析,并对超限的偏差进行预警。系统部署应用于某核电厂后,成功实现了若干起早期故障的预警,以某次凝汽器压力突然上升触发数字化分布式控制系统(DCS)报警事件为例,在DCS报警前及主控室运行人员巡检发现前约4 h,本系统已探知到凝汽器压力的异常上升并发出预警,为运行和检修人员预留足够的处理时间,极大地减少了运行人员的监盘压力,提升了机组运行的经济性。In order to avoid sudden equipment failures affecting the safety and economy of nuclear power plants,a key index early warning system is developed based on the historical operation data and neural network algorithms.Using the system,key performance of the equipment can be predicted and deviation between the predicted value and the actual measured value can be analyzed by using a variety of control charts,and finally early warning for deviations exceeding the limits can be generated.After the system is deployed in a nuclear power plant,it has successfully realized several early warnings of failures.Take a sudden rise in condenser pressure triggering a DCS alarm event as an example,the system has detected the abnormal rise in condenser pressure and issued an early warning about 4 hours before the DCS alarm and before the inspection by the operation personnel in the main control room,which can reserve enough time for the operation and maintenance personnel to handle the problem.The system can greatly reduce the pressure on the operation personnel to monitor the panels and improves the operation safety and economy of the unit.

关 键 词：故障预警 神经网络 控制图 

分 类 号：TL48[核科学技术—核技术及应用]