AI-driven optimization of CO2 efficiency in CCUS projects
Geo-Energy Transition and Carbon Management · 30.06.2026 · 4 (1) · 13-26
DOI: 10.21595/getc.2026.25830 ↗Kurz gefasst
Fachartikel in der Zeitschrift Geo-Energy Transition and Carbon Management, online erschienen am 30.06.2026. Verfasst von Ahmed Wagia-Alla und 2 weiteren Autor:innen. Die Arbeit nennt 21 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Designing effective Water-Alternating-Gas (WAG) injection schemes is central to improving oil recovery while simultaneously enhancing CO 2 storage outcomes in CO 2 -EOR operations. Operational choices such as injection rate, gas-to-water ratio (GWR), and cumulative CO 2 throughput exert strong control on displacement efficiency, recycle rates, and the fraction of CO 2 ultimately retained in the reservoir. Although numerical simulation remains the standard tool for WAG optimization, its high computational cost limits its usefulness for rapid evaluation of multiple operational scenarios, particularly under data uncertainty. To address this limitation, this study introduces a machine-learning-based forecasting workflow using the Temporal Fusion Transformer (TFT) to assess short-term CO 2 -EOR response and guide WAG optimization within a CCUS framework. Monthly injection and production records were digitized for six mature CO 2 -EOR fields in the United States, spanning a wide range of reservoir properties and development strategies. Pre-CO 2 waterflood behavior was characterized using exponential decline functions to separate baseline trends from incremental oil production induced by CO 2 injection. The TFT model was trained on multivariate time-series inputs – including water and CO 2 injection rates-to predict oil and CO 2 production over a 12-month horizon for alternative WAG configurations. Model robustness and predictive skill were improved through systematic hyperparameter tuning. Across most fields, the trained model demonstrated strong performance, achieving coefficients of determination greater than 0.87 for oil forecasts and above 0.91 for CO 2 production. The forecasted results were subsequently used to quantify CO 2 utilization and short-term retention, enabling field-specific operational insights. In the Denver Unit, for instance, increasing the GWR from 0.9 to 1.7 resulted in a 48 % increase in retained CO 2 and a 38 % improvement in utilization efficiency. Conversely, reducing the GWR in the East Vacuum field from 0.7 to 0.2 enhanced near-term sequestration efficiency despite lower injection volumes. These findings highlight the strong sensitivity of CO 2 storage performance to tailored WAG design and demonstrate the potential to reduce recycle losses through targeted operational adjustments. This work represents the first application of a Temporal Fusion Transformer model for multi-field CO 2 EOR forecasting and WAG optimization from a CCUS perspective. Unlike prior machine-learning studies that focus primarily on production prediction, this framework directly links AI-based forecasts to operational CO 2 utilization and retention metrics, enabling rapid scenario screening without full-physics simulation. The proposed efficiency score provides a novel, field-deployable indicator for balancing short-term oil recovery with carbon storage objectives, offering a scalable digital workflow to support CCUS operational decision-making.
Zitieren
APA
Wagia-Alla, A., Alghazal, M., & Alzahrani, T. (2026). AI-driven optimization of CO2 efficiency in CCUS projects. Geo-Energy Transition and Carbon Management, 4(1), 13-26. https://doi.org/10.21595/getc.2026.25830
Vancouver
Wagia-Alla A, Alghazal M, Alzahrani T. AI-driven optimization of CO2 efficiency in CCUS projects. Geo-Energy Transition and Carbon Management. 2026;4(1):13-26. doi:10.21595/getc.2026.25830
BibTeX
@article{WagiaAlla2026,
author = {Wagia-Alla, Ahmed and Alghazal, Mohamed and Alzahrani, Turki},
title = {AI-driven optimization of CO2 efficiency in CCUS projects},
journal = {Geo-Energy Transition and Carbon Management},
year = {2026},
volume = {4},
number = {1},
pages = {13--26},
doi = {10.21595/getc.2026.25830}
}
EUVIMED