Scholarlink

<p>The aging of Liquid Natural Gas (LNG) in cryogenic conditions is of great significance for the LNG supply chain, and this process must be monitored and predicted reliably. This work developed a soft sensor using flash calculations and the Peng-Robinson equation of state in order to monitor the weathering of LNG during long storage periods. The sensor was able to estimate the methane molar fraction with less than 1% RMSE in near real time. A new hybrid model was also proposed to predict the aging of LNG, combining machine learning and thermodynamics. Four different machine learning techniques were tested, and the best results originated from the gradient boosting model. This model was validated using industry data and compared with a traditional thermodynamic model. The hybrid model was able to offer more accurate predictions, especially for LNG volume in-tank and methane molar percentage.</p>

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

Modeling LNG weathering using a hybrid model

  • 期刊论文

  • 发表日期：

    2025年10月24日

查看全文
  • Gabriela Teixeira Justino,
  • Bruno Deon,
  • Camilla Barros Batista,
  • Gabriela Carvalho Freitas,
  • Breno Alves Machado,
  • Kleyton Pontes Cotta,
  • Angelo Marcelino Cordeiro,
  • Gabriel Augusto Nunes Nóbrega Diniz,
  • Flávio Leite Loução Jr.,
  • Hugo Reiser Vieira Portuita,
  • Elias Cordeiro da Silva Jr.

摘要

The aging of Liquid Natural Gas (LNG) in cryogenic conditions is of great significance for the LNG supply chain, and this process must be monitored and predicted reliably. This work developed a soft sensor using flash calculations and the Peng-Robinson equation of state in order to monitor the weathering of LNG during long storage periods. The sensor was able to estimate the methane molar fraction with less than 1% RMSE in near real time. A new hybrid model was also proposed to predict the aging of LNG, combining machine learning and thermodynamics. Four different machine learning techniques were tested, and the best results originated from the gradient boosting model. This model was validated using industry data and compared with a traditional thermodynamic model. The hybrid model was able to offer more accurate predictions, especially for LNG volume in-tank and methane molar percentage.