集成学习模型的堆芯物理关键参数预测  

Prediction of reactor core key physical parameters based on the ensemble learning method

作　　者：马季 郝琛 谢晓芹[2] 生义 MA Ji;HAO Chen;XIE Xiaoqin;SHENG Yi(Fundamental Science on Nuclear Safety and Simulation Technology Laboratory, Harbin Engineering University, Harbin 150001, China;College of Computer Science and Technology, Harbin Engineering University, Harbin 150001, China)

机构地区：[1]哈尔滨工程大学核安全与仿真技术国防重点学科实验室,黑龙江哈尔滨150001 [2]哈尔滨工程大学计算机科学与技术学院,黑龙江哈尔滨150001

出　　处：《哈尔滨工程大学学报》2021年第12期1769-1776,共8页Journal of Harbin Engineering University

基　　金：国家自然科学基金项目(12075067).

摘　　要：为了解决抽样统计的不确定性分析方法需执行大量重复的高保真物理计算而导致计算代价过大问题,本文采用与实际物理计算过程等价的预测模型开展计算,将重复计算时间降低至秒的量级。通过分析物理计算输入数据与输出参数的关系,利用堆叠法构建的集成学习模型建立了高保真反应堆物理计算堆芯关键参数预测模型,实现了堆芯关键参数的预测。为验证预测模型的计算效果,选择2D C5G7及3D C5G7稳态问题及单组件瞬态问题的堆芯有效增殖因子和归一化功率等分别进行了验证。验证结果表明:预测模型的预测结果与实际计算结果符合得很好,预测模型可作为开展进一步堆芯关键参数不确定性分析的工具基础。The uncertainty analysis using statistical sampling method requires several repeated high-fidelity physical calculations,and the introduced computational cost may be unacceptable.In this study,a surrogate model,which is equivalent to an actual physical calculation process,is used to perform calculations to reduce the calculation time into seconds.By capturing the relationship between input data and output parameters of a nuclear reactor physical calculation,a prediction model for the key physical parameter calculation of a high-fidelity reactor core is established based on the ensemble learning method by the stack method,and the prediction of key core parameters is realized.To verify the calculation effect of the prediction model,the reactor core key parameters,such as effective multiplication factor and normalized power of the steady-state problems and single assembly transient problems of the two-dimensional and three-dimensional C5G7 benchmarks,are selected for verification,respectively.The verification results suggest that the predicted results of the proposed model are in good agreement with the actual calculation results and that the model can be used as a basis tool for further uncertainty analyses on reactor core key parameters.

关 键 词：抽样统计 堆芯关键参数 不确定性分析 机器学习 替代模型 C5G7问题 预测精度及效率 不确定性传播工具 

分 类 号：TL32[核科学技术—核技术及应用]