现代应用物理2025，Vol.16Issue(1)：115-124,10.DOI:10.12061/j.issn.2095-6223.202412050
基于深度学习的压水堆瞬态物理场预测方法研究
Prediction Method for Transient Based on Deep Learning Physical Fields in Pressurized Water Reactors
摘要
Abstract
In this paper,a deep convolutional neural network(DCNN)model is constructed by combining the inception module and the residual network(ResNet)structure.The model can improve the accuracy and efficiency of predicting key safety parameters,such as fuel assembly power distribution,temperature distribution,and power level of the reactor,both for single-time-point and multi-time-point under transient conditions of pressurized water reactors(PWRs).Testing and analysis are conducted using a small-scale PWR model.The results show that the core parameters for a single time point can be predicted within 60 ms,with prediction deviations in fuel assembly power distribution and temperature distribution being less than 3％for both single-time-point and multi-time-point.The prediction deviation of total power for multi-time-point is less than 8％.These results demonstrate the feasibility of rapid prediction of PWR transient physical fields using deep learning techniques.关键词
压水堆/深度学习/瞬态工况/快速预测/反应堆物理场参数Key words
pressurized water reactor(PWR)/deep learning/transient conditions/rapid prediction/physical field parameters分类
能源科技引用本文复制引用
刘家旺,刘宙宇,邵世豪,曹良志,吴宏春..基于深度学习的压水堆瞬态物理场预测方法研究[J].现代应用物理,2025,16(1):115-124,10.基金项目
国家自然科学基金资助项目(12375174,U24B2010,12375175) （12375174,U24B2010,12375175）
