Enhanced water saturation estimation in hydrocarbon reservoirs using machine learning.

Area of Science:

• Petrophysics
• Machine Learning
• Reservoir Engineering

Background:

• Accurate water saturation (Sw) estimation is crucial for hydrocarbon reservoir analysis and oil recovery optimization.
• Traditional Sw estimation methods have limitations due to assumptions, core data dependency, and geological complexities.

Purpose of the Study:

• Develop and validate machine learning (ML) models for precise water saturation (Sw) prediction.
• Evaluate the performance of various ML algorithms using a comprehensive well log dataset.

Main Methods:

• Utilized a dataset of 30,660 data points with nine well log parameters.
• Trained and tested five ML algorithms: Linear Regression, Support Vector Machine (SVM), Random Forest, Least Squares Boosting, and Bayesian methods.
• Applied Gaussian outlier removal and rigorous validation techniques for model reliability.

Main Results:

• Support Vector Machine (SVM) demonstrated the highest accuracy, achieving R² values of 0.9952 (test) and 0.9962 (train).
• SVM model yielded low Root Mean Square Error (RMSE) values of 0.002 (test) and 0.001 (train).

Conclusions:

• Machine learning, especially SVM, provides a robust and accurate method for water saturation (Sw) estimation.
• The developed ML models support enhanced reservoir evaluation and optimized oil recovery.