Prediction by machine learning in nanoparticles-based enhanced oil recovery

Nanotechnology is on the brink of transforming numerous industrial sectors, and the petroleum industry stands as a front-runner in embracing these revolutionary advancements. In recent years, a growing interest has occurred in leveraging nanotechnology within the petroleum industry%, particularly to enhance oil recovery. Extensive research studies on nano-enhanced oil recovery (nano-EOR) have consistently delivered promising outcomes, underscoring its potential to elevate oil production substantially. However, a notable challenge persists within this domain due to the limited data availability concerning nanoparticle transport in porous media. This paper uses machine learning techniques to predict nanoparticle transport in porous media. This study uses the finite difference method to generate simulated datasets from a modified linear adsorption model. These simulated datasets are used to train machine learning models for prediction by considering artificial neural network (ANNs), decision tree (DT), and random forest (RF). We achieve mean squared values for ANN as 0.0478 (training), 0.0496 (testing), 0.0509 (validation), and R-squared values as 0.9798 (training), 0.9780 (testing), 0.9773 (validation), and for DT and RF mean squared values are 0.014683, 0.009807, and R squared values are 0.928775, 0.952425.