Machine Learning Prediction for Nanoparticles Behavior in Hydrocarbon Reservoirs

Abstract

The use of machine learning to forecast how nanoparticles would migrate through porous material is covered in this research. We employed the random forest, decision tree, artificial neural network, and gradient boosting regression machine learning techniques. Since there are not many experimental datasets available, it is easier to create artificial datasets using verified numerical simulators. Additionally, covered in the paper are data preprocessing, correlations, the importance of features, and hyperparameter adjustment. Moreover, different error metrics and R2-correlation are used to gauge how well the predictive models perform. Finally, examples of the findings are presented. The decision tree model is determined to have the highest accuracy, the best performance, and the lowest root mean squared error.