基于机器学习的淬冷沸腾最小膜态沸腾温度预测和灵敏度分析研究  

Research on Prediction and Sensitivity Analysis of Minimum Film Boiling Temperature of Quenching Boiling Based on Machine Learning

作　　者：张军权 邓坚[2] 罗彦 卢涛[1] Zhang Junquan;Deng Jian;Luo Yan;Lu Tao(College of Mechanical and Electrical Engineering,Beijing University of Chemical Technology,Beijing,100029,China;Science and Technology on Reactor System Design Technology Laboratory,Nuclear Power Institute of China,Chengdu,610213,China)

机构地区：[1]北京化工大学机电工程学院,北京100029 [2]中国核动力研究设计院核反应堆系统设计技术重点实验室,成都610213

出　　处：《核动力工程》2024年第4期69-76,共8页Nuclear Power Engineering

基　　金：国家自然科学基金(52176052,U2067210)。

摘　　要：淬冷沸腾广泛应用于核反应堆失水事故后燃料棒的冷却过程中,最小膜态沸腾温度(T_(min))的确定对核反应堆的安全运行至关重要。本文基于文献的实验数据,选用了3种典型机器学习模型:随机森林(RF)、人工神经网络(ANN)和极端梯度提升(XGBoost),对淬冷沸腾T_(min)进行预测和影响因素灵敏度分析研究。结果表明,机器学习方式能够有效提高淬冷沸腾T_(min)预测的准确性,其预测性能相较于传统的经验关联式有大幅提升,其中RF模型预测效果最优,决定系数R^(2)为0.9770;通过结合RF模型和Sobol’全局灵敏度方法,得到对淬冷沸腾T_(min)影响最大的参数为冷却剂过冷度,其次为初始壁温,长径比、压力、热物性对其影响较小。本文研究成果将为提高核反应堆的安全性提供理论指导。Quenching boiling is widely used in the cooling process of fuel rods after the loss of coolant accident in nuclear reactor.The determination of the minimum film boiling temperature(T_(min))is very important for the safe operation of nuclear reactors.Based on the experimental data in the literature,this paper selects three typical machine learning models:Random Forest(RF),Artificial Neural Network(ANN)and eXtreme Gradient Boosting(XGBoost)to predict T_(min)during quenching boiling and conduct a sensitivity analysis of influencing factors.The results show that the machine learning method can effectively improve the accuracy of T_(min)prediction compared to the traditional empirical correlation.Among the models,the RF model exhibits the best predictive performance with a coefficient of determination R^(2) of 0.9770.By combining the RF model with the Sobol’global sensitivity method,the study identifies the coolant subcooling as the most influential parameter on T_(min),followed by initial wall temperature,while length-diameter ratio,pressure and thermophysical properties have a smaller impact.The findings of this research will provide theoretical guidance for improving the safety of nuclear reactors.

关 键 词：淬冷沸腾 最小膜态沸腾温度 机器学习 全局灵敏度 

分 类 号：TL334[核科学技术—核技术及应用]