【作者】 蒋波涛； 赵福宇；

【Author】 JIANG Bo-tao,ZHAO Fu-yu (The department of nuclear science and technology,Xi’an Jiaotong University,Xi’an,Shaanxi 710049,China)

【机构】 西安交通大学核科学与技术学院；

【摘要】 基于三维堆芯的中子动力学模型和热工水力模型,使用堆芯三维仿真程序模拟堆芯弹棒事故,得到三组不同工况下反应堆堆芯功率的变化数据,并把这些数据分为训练和验证支持向量机算法(Support Vector Regression,SVR)的训练集和验证集,来预测弹棒瞬态事故时堆芯功率的动态变化,而且和其他两种预测方法做了比较。预测结果表明,支持向量机算法在预测性能指标上优于其他两种算法,计算速度更快,预测精度更高。除此之外,和传统的预测方法相比,支持向量机算法无需人为假设模型和先验知识,完全从数据内部关系出发,所得结果更加可靠,可用三维堆芯的其他瞬态事故分析中。更多还原

【Abstract】 Based on neutron kinetics model and thermal-hydraulics model of a three-dimensional reactor core,the simulation data of rod ejection transient under three different operation conditions of three-dimensional reactor core were obtained using the simulation program.And,these simulation data were divided into training set and validation set which is used to train and valid the proposed algorithm respectively and used to predict the dynamic power change of reactor core when rod ejection transient happened.Compared with other two prediction algorithms,the prediction results show that the proposed algorithm is better than other two algorithms at several prediction performance indexes,such as having a higher calculation speed,higher prediction accuracy.Also,compared with conventional prediction methods,there does not need artificial assumption model and priori knowledge for the proposed algorithm,but totally is based on the interior relationship of data themselves.Thus,the results obtained by SVR algorithm are more reliable and can be used to analyze other transient accidents of three-dimensional reactor core.更多还原