---

Evaluating the performance of an optimized tree-ensemble learning algorithm for predicting sonic logs in the Gandhar CO2 EOR project.

ZiaS,DabiS,VedantiN.

JOURNAL OF APPLIED GEOPHYSICS

https://doi.org/10.1016/j.jappgeo.2025.105904 

This study presents an optimized machine learning approach to predict missing well-logs, essential for CO₂ storage characterization in the Gandhar oilfield, India. The model produced lower prediction errors with reliable uncertainty estimates, outperformed empirical methods, and effectively captured lithological variations. The predicted logs were then used to model CO₂-saturated velocities, providing a robust and scalable tool for risk-informed reservoir characterization.

Prediction results of test well 27 from (a) optimized GBR algorithm with the 90 % prediction interval. (b) Plot of gamma-ray log. Empirical method results from (c) Gardner, (d) Lindseth, and (e) the combination of both. (f) Plot of density log for a better understanding of low to high-density lithology. Lithological boundaries are marked in each figure for enhanced clarity.

---