Development of cross-section calculation module for high-temperature gas-cooled reactor engineering simulator system using an optimal model tree algorithm. (May 2023)
- Record Type:
- Journal Article
- Title:
- Development of cross-section calculation module for high-temperature gas-cooled reactor engineering simulator system using an optimal model tree algorithm. (May 2023)
- Main Title:
- Development of cross-section calculation module for high-temperature gas-cooled reactor engineering simulator system using an optimal model tree algorithm
- Authors:
- Tan, Kai
Liu, Meidan
Wei, Chunlin - Abstract:
- Highlights: A more accurate cross-section calculation model is required for real-time simulation. A machine-learning-based cross-section calculation model is proposed. The proposed method produced accurate cross-section and reactor physics results. It also met the engineering simulator system real-time calculation requirements. Under different operating conditions it conformed to the reactor laws of physics. Abstract: To improve the accuracy of reactor physics module in the high-temperature gas-cooled reactor engineering simulator system (HTGR-ESS), a more accurate cross-section calculation model is required under the premise of real-time calculation in the simulation. Due to the nonlinear relationship between cross-section and reactor state parameters, such as burnup, moderator temperature and fuel temperature, etc., it is difficult to achieve full-range accuracy using the current multiple linear regression method (MLR). In this study, a machine-learning-based cross-section calculation model for the reactor physics model is proposed. We made improvements to the model tree by setting a smoothing function on the boundary and using ridge regression at the leaf node to conduct a continuous cross-section calculation model without overfitting in the subspace over the full range, trained from a scattered database generated by V.S.O.P. In the numerical tests, the proposed method produced far more accurate cross-section and reactor physics calculation results than the current methodHighlights: A more accurate cross-section calculation model is required for real-time simulation. A machine-learning-based cross-section calculation model is proposed. The proposed method produced accurate cross-section and reactor physics results. It also met the engineering simulator system real-time calculation requirements. Under different operating conditions it conformed to the reactor laws of physics. Abstract: To improve the accuracy of reactor physics module in the high-temperature gas-cooled reactor engineering simulator system (HTGR-ESS), a more accurate cross-section calculation model is required under the premise of real-time calculation in the simulation. Due to the nonlinear relationship between cross-section and reactor state parameters, such as burnup, moderator temperature and fuel temperature, etc., it is difficult to achieve full-range accuracy using the current multiple linear regression method (MLR). In this study, a machine-learning-based cross-section calculation model for the reactor physics model is proposed. We made improvements to the model tree by setting a smoothing function on the boundary and using ridge regression at the leaf node to conduct a continuous cross-section calculation model without overfitting in the subspace over the full range, trained from a scattered database generated by V.S.O.P. In the numerical tests, the proposed method produced far more accurate cross-section and reactor physics calculation results than the current method and could also meet the real-time calculation requirement. Meanwhile, the responses of the power and helium outlet temperatures of the 10 MW HTGR (HTR-10) simulator under different operating conditions conform to the reactor laws of physics, further validating the applicability of the proposed method. … (more)
- Is Part Of:
- Annals of nuclear energy. Volume 184(2023)
- Journal:
- Annals of nuclear energy
- Issue:
- Volume 184(2023)
- Issue Display:
- Volume 184, Issue 2023 (2023)
- Year:
- 2023
- Volume:
- 184
- Issue:
- 2023
- Issue Sort Value:
- 2023-0184-2023-0000
- Page Start:
- Page End:
- Publication Date:
- 2023-05
- Subjects:
- High-temperature gas-cooled reactor -- Cross-section -- Engineering simulator system -- Machine learning
Nuclear energy -- Periodicals
Nuclear engineering -- Periodicals
621.4805 - Journal URLs:
- http://www.sciencedirect.com/science/journal/03064549 ↗
http://catalog.hathitrust.org/api/volumes/oclc/2243298.html ↗
http://www.elsevier.com/journals ↗ - DOI:
- 10.1016/j.anucene.2023.109683 ↗
- Languages:
- English
- ISSNs:
- 0306-4549
- Deposit Type:
- Legaldeposit
- View Content:
- Available online (eLD content is only available in our Reading Rooms) ↗
- Physical Locations:
- British Library DSC - 1043.150000
British Library DSC - BLDSS-3PM
British Library HMNTS - ELD Digital store - Ingest File:
- 25665.xml