Machine learning–enabled optimization of a direct air capture system integrated with enhanced oil recovery

The escalating levels of CO₂ in the atmosphere have heightened global environmental concerns, necessitating the deployment of efficient and scalable carbon extraction strategies. Among emerging methods, direct air capture (DAC) stands out as a viable approach. This research introduces a novel DAC configuration tailored to enhance the efficiency of enhanced oil recovery (EOR). The DAC system was modeled using Aspen Plus V11, employing a hydroxide-to‑carbonate conversion pathway for CO₂ absorption. As part of broader carbon management efforts, Carbon Capture, Utilization, and Storage (CCUS) plays a pivotal role in curbing emissions, particularly through its application in subsurface oil recovery processes. To assess and forecast the impact of DAC-sourced CO₂ on EOR performance in Abadan, a suite of Machine learning techniques was applied. These included XGBoost, Random Forest, Gradient Boosting, Support Vector Regression, Linear Regression, k-Nearest Neighbors, Bagging, and Stacking. Among the models tested, the Decision Tree algorithm demonstrated the highest predictive capability, yielding an R 2 score of 0.87. It effectively estimated a growth in EOR efficiency from 19 % to approximately 21.3 %. • Ensemble models achieved highest accuracy in EOR factor prediction. • SHAP plots revealed top features driving model predictions. • Residual analysis confirmed Random Forest's low bias and variance. • DAC integration increased EOR factor from 19 % to 21.3 %. • Study supports AI-based optimization for carbon-EOR strategies.