Accurately quantifying porosity in reservoirs is essential for optimizing geothermal resource exploration and subsurface resource evaluation. Conventional predictive methods, such as density porosity models, often have limited accuracy, which can impede effective reservoir characterization. To overcome these limitations, this study leverages advanced machine learning (ML) techniques including AdaBoost, XGBoost, Gradient Boosting, LightGBM (LGBM), and ExtraTrees to predict porosity using well log data. ML offers a powerful alternative by leveraging data-driven approaches to uncover complex, nonlinear relationships between well log responses and porosity. Feature selection techniques, including correlation analysis from heat maps and feature importance are employed to identify well log data relationships. A comprehensive comparative analysis of well log data from two wells in North Dakota demonstrates that the comparative ML-based approach could be utilized to predict porosity effectively. The robustness of the data-driven models is validated through 5-fold cross-validation to confirm its reliability. Additionally, blind testing on a second well further verifies the model’s generalization capability and practical applicability. The results highlight the strong potential of machine learning in enhancing porosity estimation for geothermal reservoirs. By providing more precise and efficient predictions, this ML-driven framework can support better decision-making in geothermal exploration and development, subsurface characterization, reservoir modeling uncertainty, ultimately contributing to more sustainable and cost-effective energy extraction.

Press the Back button in your browser, or search again.

Copyright 2026, Stanford Geothermal Program: Readers who download papers from this site should honor the copyright of the original authors and may not copy or distribute the work further without the permission of the original publisher.