Related Experiment Video
Updated: Jun 3, 2025
00:07
Accurate Determination of the Equilibrium Surface Tension Values with Area Perturbation Tests
Published on: August 30, 2019
7.4K
Machine learning-based estimation of crude oil-nitrogen interfacial tension
Safia Obaidur Rab1,2, Subhash Chandra3, Abhinav Kumar4,5,6
1Central Labs, King Khalid University, P.O. Box 960, AlQura'a, Abha, Saudi Arabia.
Scientific Reports
|January 6, 2025
Summary
Predicting interfacial tension (IFT) between nitrogen and crude oil using machine learning models is crucial for enhanced oil recovery. Random Forest demonstrated the highest accuracy in predicting crude oil-nitrogen IFT for real reservoir conditions.
Area of Science:
- Petroleum Engineering
- Computational Science
Background:
- Accurate interfacial tension (IFT) prediction between nitrogen and crude oil is vital for optimizing nitrogen-based gas injection in oil reservoirs.
- Existing models often use synthetic oils, limiting their applicability to real crude oil systems.
Purpose of the Study:
- To develop accurate data-driven models for predicting crude oil-nitrogen IFT using real crude oil samples.
- To evaluate the performance of eight machine learning methods for this prediction task.
Main Methods:
- Utilized eight machine learning algorithms: Decision Tree, AdaBoost, Random Forest, K-nearest Neighbors, Ensemble Learning, Support Vector Machine, Convolutional Neural Network, and Multilayer Perceptron Artificial Neural Network.
- Trained and validated models using experimental data from real crude oil samples and assessed performance using statistical indices and graphical approaches.
Main Results:
- All developed models showed suitability for predicting crude oil-nitrogen IFT.
- Sensitivity analysis revealed that pressure, temperature, and crude oil API negatively impact IFT, with pressure being the most influential factor.
- The Random Forest model achieved the highest accuracy, with R-squared of 0.959, MSE of 1.65, and AARE of 6.85% on unseen data.
Conclusions:
- Random Forest is the most accurate and reliable model for predicting crude oil-nitrogen IFT.
- The developed model serves as an accessible tool for optimizing enhanced oil recovery and reservoir investigations.
