Scholarlink

<p>Natural gas from tight formations plays a critical role in global energy, while CO<sub>2</sub> geo-storage offers a dual opportunity for climate mitigation and enhanced methane recovery. This paper proposes a novel generalized structure of group method of data handling (g-GMDH) to predict CO<sub>2</sub> and CH<sub>4</sub> sorption behavior using geological and operational inputs, including total organic carbon (TOC) content, pressure, temperature, and moisture. The g-GMDH demonstrates superior performance compared to other models, such as extreme gradient boosting (XGBoost), random forest (RF), light gradient boosting machine (LGBM), and multilayer perceptron (MLP), and show high coefficient of determination (R<sup>2</sup>) values and lower errors. During the training and testing phases, the models’ performance followed this order: g-GMDH &gt; XGBoost &gt; RF &gt; LGBM &gt; MLP. Using the g-GMDH model to predict CH<sub>4</sub> excess sorption, the training dataset yielded R<sup>2</sup> = 1, mean absolute error (MAE) = 0.01148, and relative root mean square error (RMSE) = 0.00193, while the testing dataset resulted in R<sup>2</sup> = 0.998, MAE = 0.02383, and RMSE = 0.0312. Similarly, for modeling CO<sub>2</sub> excess sorption using the g-GMDH model, the training data showed R<sup>2</sup> = 1, MAE = 0.00283, and RMSE = 0.00126, while the testing data yielded R<sup>2</sup> = 0.969, MAE = 0.0065, and RMSE = 0.0226. These findings highlight the substantial improvements using g-GMDH model to predict gas sorption, offering valuable insights for enhancing methane extraction efficiency and optimizing CO<sub>2</sub> storage. This dual-purpose approach supports sustainable energy production and global climate mitigation goals.</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

Improving Sorption Capacity in Tight Reservoirs Through an Advanced Generalized Group Method of Data Handling (g-GMDH)

  • 期刊论文

  • 发表日期：

    2026年7月5日

查看全文
  • Elieneza Nicodemus Abelly,
  • Feng Yang,
  • Nafouant Mouigni Baraka,
  • Irene Martin Ndunguru,
  • Allou Koffi Franck Kouassi,
  • Elizabeth Frederick Mumbuli

摘要

Natural gas from tight formations plays a critical role in global energy, while CO2 geo-storage offers a dual opportunity for climate mitigation and enhanced methane recovery. This paper proposes a novel generalized structure of group method of data handling (g-GMDH) to predict CO2 and CH4 sorption behavior using geological and operational inputs, including total organic carbon (TOC) content, pressure, temperature, and moisture. The g-GMDH demonstrates superior performance compared to other models, such as extreme gradient boosting (XGBoost), random forest (RF), light gradient boosting machine (LGBM), and multilayer perceptron (MLP), and show high coefficient of determination (R2) values and lower errors. During the training and testing phases, the models’ performance followed this order: g-GMDH > XGBoost > RF > LGBM > MLP. Using the g-GMDH model to predict CH4 excess sorption, the training dataset yielded R2 = 1, mean absolute error (MAE) = 0.01148, and relative root mean square error (RMSE) = 0.00193, while the testing dataset resulted in R2 = 0.998, MAE = 0.02383, and RMSE = 0.0312. Similarly, for modeling CO2 excess sorption using the g-GMDH model, the training data showed R2 = 1, MAE = 0.00283, and RMSE = 0.00126, while the testing data yielded R2 = 0.969, MAE = 0.0065, and RMSE = 0.0226. These findings highlight the substantial improvements using g-GMDH model to predict gas sorption, offering valuable insights for enhancing methane extraction efficiency and optimizing CO2 storage. This dual-purpose approach supports sustainable energy production and global climate mitigation goals.