核电站主蒸汽系统冗余压力传感器异常检测

(1.厦门大学能源学院,福建 厦门 361102; 2.福清核电站,福建 福州 350300)

独立成分分析; 序贯概率比检验; 核电站; 冗余传感器校准; 异常检测

Fault detection for redundant pressure sensors in main steam system in nuclear power plants
JIN Dian1,XIE Shan1*,DING Jun1,WU Yichun1,LI Ning1,CAO Peigen2

(1.College of Energy,Xiamen University,Xiamen 361102,China; 2.Fuqing Nuclear Power Plant,Fuzhou 350300,China)

independent component analysis(ICA); sequential probability ratio test(SPRT); nuclear power plant; redundant sensors calibration; fault detection

DOI: 10.6043/j.issn.0438-0479.201811018

备注

为保障核电站运行安全,电站设置冗余传感器对重要系统的关键参数进行测量,传感器的健康状况将直接影响测量结果.在被测冗余传感器个数较少,为3个左右时,目前核电站常用的传感器校准方法很难区分故障传感器.将独立成分分析和序贯概率比检验(independent component analysis-sequential probability ratio test,ICA-SPRT)相结合的方法用于冗余传感器故障检测,使用核电站主蒸汽系统冗余压力传感器数据进行验证.与简单平均法的结果进行对比可知,该方法在冗余度不高的情况下明显优于简单平均法,能及时准确地检测到传感器信号的漂移.

Redundant sensors are installed for measurement of critical parameter in crucial systems in nuclear power plants(NPPs)to ensure the operation safety,and status of the sensor will directly influence the results of measurements.Now it is difficult for traditional sensor calibration methods(simple average,instrumentation and calibration monitoring program,parity space,etc.)commonly used in NPPs to recognize the faulty redundant sensors when redundancy of systematic parameter is 3 or low.In this article,independent component analysis combined with sequential probability ratio test(ICA-SPRT)is proposed for fault detection in redundant sensors,and is validated by the dataset from redundant pressure sensors in main steam system in NPPs.Compared with results of simple average,the method cannot only outperform traditional ones but also accurately detects the drift of sensors' signal in time and provides more time for fault analysis and maintenance of sensors.