Scholarlink

This study focuses on the supercritical carbon dioxide (S-CO2) Brayton cycle heat transfer system, which is notably advantageous in small modular nuclear power systems due to its high cycle efficiency and compact structure. Continual improvement of system analysis codes is necessitating to support the engineering design and analysis demands. For enhancement in terms of scope, efficiency, and accuracy, this research developed the FRTAC3 thermal–hydraulic analysis code, which utilizes semi/full implicit numerical schemes and integrates neural network models for isentropic processes, compressors, turbines, and printed circuit heat exchangers. These models have been validated and demonstrated moderate accuracy and generalization capabilities. Furthermore, the paper achieved the coupling of the thermal–hydraulic analysis program with the neural network modules and analyzed the thermal–hydraulic behavior of typical S-CO2 Brayton cycles under electrical load up transient. Through code comparison validation, the results indicate that the coupled program can accurately describe the trend of system parameters with reasonable computational deviations (transient valley deviations less than 4.5%).

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Application of Machine Learning Models in the Analysis of Supercritical Carbon Dioxide Cycle System

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Yang Yu,
  • Donghui Zhang,
  • Yongfa Zhang

摘要

This study focuses on the supercritical carbon dioxide (S-CO2) Brayton cycle heat transfer system, which is notably advantageous in small modular nuclear power systems due to its high cycle efficiency and compact structure. Continual improvement of system analysis codes is necessitating to support the engineering design and analysis demands. For enhancement in terms of scope, efficiency, and accuracy, this research developed the FRTAC3 thermal–hydraulic analysis code, which utilizes semi/full implicit numerical schemes and integrates neural network models for isentropic processes, compressors, turbines, and printed circuit heat exchangers. These models have been validated and demonstrated moderate accuracy and generalization capabilities. Furthermore, the paper achieved the coupling of the thermal–hydraulic analysis program with the neural network modules and analyzed the thermal–hydraulic behavior of typical S-CO2 Brayton cycles under electrical load up transient. Through code comparison validation, the results indicate that the coupled program can accurately describe the trend of system parameters with reasonable computational deviations (transient valley deviations less than 4.5%).