فيصل غضبان

This study investigates the optimization of nanoparticle applications for altering the wettability of carbonate reservoirs to improve oil recovery (EOR). The research aims to determine the most effective nanoparticle type, size, concentration, an‎d operating temperature that can shift rock surfaces from an oil-wet to a water-wet state. Experimental data collected from previously published laboratory studies were utilized to build an‎d validate the predictive models. Three nanoparticle materials silicon dioxide (SiO₂), titanium dioxide (TiO₂), an‎d aluminum oxide (Al₂O₃) were examined. To model the relationship between nanoparticles an‎d wettability alteration, five machine learning algorithms were implemented: Decision Tree (DT), Ran‎dom Forest (RF), Extreme Gradient Boosting (XGBoost), Stochastic Gradient Boosting (SGB), an‎d Support Vector Machine (SVM). The results of the comparison revealed that the Ran‎dom Forest (RF) model had the most consistent performance as indicated with high accuracy in training (R² = 0.981), validation (R² = 0.980), an‎d test datasets (R² = 0.932). XGBoost was also doing very well (R²_test = 0.929), followed by SGB (R²_test = 0.912). The Decision Tree (DT) was found to have a high-test accuracy (R² = 0.941) but with considerable overfitting, as suggested by its considerably smaller validation accuracy (R² = 0.903). In the meantime, Support Vector Machine (SVM) model was the worst model, with (R²_test = 0.748), indicating that they have low generalization ability, as opposed to ensemble-based models. In all the models, the use of, SiO₂ nanoparticles was found to be the most effective type, where smaller sizes an‎d moderate concentrations of the particles at moderate temperatures are the best way of increasing wettability in carbonate formations. Overall, the results validate the view that the implementation of artificial intelligence practices an‎d, in particular, ensemble learning models can be a reliable an‎d effective framework to predict an‎d optimize the use of nanoparticles in promoting the wettability of the oil target an‎d thus play a significant role in improving the performance of oil recovery in the carbonate oilfields.

نانوذرات , تغيير ترشوندگي , مخازن كربناته , ازدياد برداشت نفت , يادگيري ماشين

Nanoparticles , Wettability Alteration , Carbonate Reservoirs , Enhanced Oil Recovery , Machine Learning

Faisal Mohammed Ghadhban

Mohammed Taghi Sadeghi