Article · Bulletin of the National Engineering Academy of the Republic of Kazakhstan · December 30, 2024
Predicting oil recovery factor using dense and cascade neural networks
Samson Dawit Bekele, Serik Aibagarov, Yerzhan Kenzhebek, Timur Imankulov
Abstract
Polymer flooding is a cost-effective and successful method for enhancing oil recovery and is the focus of this study which utilized the Buckley-Leverett mathematical model to create over 163,000 synthetic samples representing various reservoir conditions using seven input parameters. To mimic real-world conditions and reduce overfitting, artificial noise was added to the dataset. Two neural networks were trained with this data, both achieving an R² (coefficient of determination) greater than 0.906, indicating high predictive accuracy. The study also conducted permutation feature importance analysis to identify crucial parameters in both models. This research, with its extensive and realistically simulated dataset and sophisticated machine learning techniques, provides a robust framework that can enhance decision- making in oil production by pinpointing key parameters and forecasting the oil recovery factor based on specific input values.
More by Samson Dawit Bekele
- Separating Artificial-Viscosity Annealing from the L-BFGS Viscosity in Physics-Informed Neural Networks for Two-Phase FlowBig Data and Cognitive Computing · 2026
- Identifiability and Conditioning of Flux-Parameter Inversion for the Buckley–Leverett EquationMathematics · 2026
- Artificial-viscosity schedules for Buckley–Leverett PINNs: code and reproducibility archiveZenodo · 2026
- Physics-Informed Neural Networks for Coupled Non-Isothermal Two-Phase Flow in Porous Media: Training Pathologies, Remedies, and the Role of Gradient ConflictMachine Learning and Knowledge Extraction · 2026