NUCLEAR TECHNIQUES, Volume. 47, Issue 10, 100604(2024)
Prediction method of reactor neutron flux and keff based on the optimized extreme learning machine model
Figures & Tables(24)
Fig. 8. Feature importance evaluation chart for neutron flux prediction
Fig. 9. Comparison of training time for predicting keff in ELM models before and after dimensionality reduction processing (a) Training time before dimensionality reduction processing, (b) Training time after dimensionality reduction processing
Fig. 10. Comparison of neutron flux training time predicted by ELM model before and after dimensionality reduction (a) Training time before dimensionality reduction, (b) Training time after dimensionality reduction
Fig. 11. Comparison chart of MSE in predicting keff by the ELM model before and after dimensionality reduction
Fig. 13. Graph of MSE variation with iteration number during optimization process
Fig. 15. Comparison chart of MSE in predicting neutron flux by the ELM model before and after dimensionality reduction
Fig. 17. Chart of MSE variation with iteration number during optimization process
Table 1. Model specific details
View tableView in ArticleTable 1. Model specific details
总集Aggregate 1 775 100% 训练集Training set 1 242 70% 测试集Test set 533 30% 激活函数Activation function sigm函数sigm function 数据预处理方式Data preprocessing method 归一化函数Normalization function
Table 2. Comparison of time between model training and testing keff before and after dimensionality reduction
View tableView in ArticleTable 2. Comparison of time between model training and testing keff before and after dimensionality reduction
项目Items 训练时间Training time / s 测试时间Test time / s 降维前（L1=60） Before improvement 1.998 424 530 029 297×10-3 2.049 922 943 115 234 40×10-3 降维后（L2=90） After improvement 3.000 020 980 834 961×10-3 1.000 642 776 489 257 8×10-3
Table 3. Comparison of time between model training and neutron flux testing before and after dimensionality reduction
View tableView in ArticleTable 3. Comparison of time between model training and neutron flux testing before and after dimensionality reduction
项目Items 训练时间Training time / s 测试时间Test time / s 降维前（L3=78） Before improvement 2.154 111 862 182 617×10-3 1.964 569 091 796 875×10-3 降维后（L4=100） After improvement 3.003 120 422 363 281 2×10-3 2.677 679 061 889 648 4×10-3
Table 4. Comparison of MSE between model training and testing keff before and after dimensionality reduction
View tableView in ArticleTable 4. Comparison of MSE between model training and testing keff before and after dimensionality reduction
项目Items 训练集MSE Training set MSE 测试集MSE Test set MSE 降维前Before improvement 3.049 596 819 896 787×10-2 3.229 148 618 947 934×10-2 降维后After improvement 7.272 848 783 684 708×10-4 5.106 775 845 028 427×10-4
Table 5. MSE comparison of model training and testing keff before and after whale algorithm optimization
View tableView in ArticleTable 5. MSE comparison of model training and testing keff before and after whale algorithm optimization
项目Items 训练集MSE Training set MSE 测试集MSE Test set MSE 优化前Before optimization 7.272 848 783 684 708×10-4 5.106 775 845 028 427×10-4 优化后After optimizationt 2.343 936 860 088 112×10-4 2.354 153 462 325 817 2×10-4
Table 6. MSE comparison of neutron flux between model training and testing before and after dimensionality reduction
View tableView in ArticleTable 6. MSE comparison of neutron flux between model training and testing before and after dimensionality reduction
项目Items 训练集MSE Training set MSE 测试集MSE Test set MSE 降维前Before improvement 5.760 846 330 435 586×10-2 6.116 552 457 996 486×10-2 降维后After improvement 1.135 978 568 282 202 5×10-2 1.268 986 912 353 550 6×10-2
Table 7. MSE comparison of neutron flux between model training and testing before and after whale algorithm optimization
View tableView in ArticleTable 7. MSE comparison of neutron flux between model training and testing before and after whale algorithm optimization
项目Items 训练集MSE Training set MSE 测试集MSE Test set MSE 优化前Before optimization 1.135 978 568 282 202 5×10-2 1.268 986 912 353 550 6×10-2 优化后After optimization 7.193 538 786 452 611×10-3 7.803 206 858 728 087×10-3
Tools
Get Citation
Copy Citation Text
Jingyu CHEN, Xiyang LIU, Pengcheng ZHAO, Zijing LIU, Wei LI. Prediction method of reactor neutron flux and keff based on the optimized extreme learning machine model[J]. NUCLEAR TECHNIQUES, 2024, 47(10): 100604
Paper Information
Category: NUCLEAR ENERGY SCIENCE AND ENGINEERING
Received: Oct. 27, 2023
Accepted: --
Published Online: Dec. 13, 2024
The Author Email: LIU Zijing (LIUZijing)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20