基于神经网络方法的LOCA事故诊断

冷却剂丧失事故(Loss of Coolant Accident,LOCA)是核电厂安全分析中的一类典型事故,不同的破口位置和破口尺寸将直接影响到事故的处置和后果。为判断LOCA事故的破口位置和尺寸,可以借助于神经网络的模式识别功能。针对CPR1000核电系统,利用CATHARE软件建模并仿真不同破口位置和尺寸的LOCA事故,提取事故发生时的6类热工水力参数对BP(Back Propagation)神经网络、Elman神经网络、RBF(Radial Basis Function)神经网络和支持向量机进行训练,再将训练后的神经网络用于破口位置和尺寸的诊断。结果表明,在4种神经网络中,参数优化后的支持向量机对破口位置和尺寸的诊断准确率较高且诊断稳定性较好。在LOCA事故发生时,可以利用支持向量机获取破口的详细信息,辅助操纵员高效地处理事故。

Fault diagnosis of LOCA based on ANN methods

Background: Loss of coolant accident (LOCA) is one of the typical accidents in safety analysis of nuclear power plant and the location and the size of break will affect its treatment and consequences directly. Purpose:This study aims to diagnose the location and the size of break by using artificial neural network (ANN) based pattern recognition approach. Methods: CATHARE program was used to model and simulate different location and size of break in LOCA for the CPR1000 nuclear power system. Six types of thermal-hydraulic parameters were extracted to train four types of ANN methods (back propagation (BP) neural network, Elman neural network, radial basis function (RBF) neural network and support vector machine) and the trained ANNs were utilized to diagnose the location and the size of break. Results: The optimized support vector machine (SVM) is best method in terms of diagnosis accuracy and stability among 4 ANNs. Conclusion: The operators can obtain more detailed information about break by SVM to deal with the accident efficiently, when a LOCA happens.