Full text loading...
-
Integrated Cores and Logs to Predict Rocktype and Permeability using ML, Central Luconia Carbonate Buildup, Malaysia
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, EAGE/FESM Conference on Petrophysics meets Geoscience: Unlocking Reservoir Potential in a Dynamic Energy Landscape, Nov 2025, Volume 2025, p.1 - 3
Abstract
Summary
Carbonate reservoirs in the Central Luconia field, Sarawak, Malaysia, present considerable challenges due to their heterogeneity and complex pore systems, which complicate reservoir characterization and hydrocarbon recovery. Traditional log-based methods often fail to accurately represent lithofacies variation and pore structure, resulting in discrepancies in dynamic reservoir models. To address these limitations, this research introduces a machine learning (ML)-based approach to enhance rock type classification and permeability estimation.
The proposed methodology follows a three-step workflow. First, rock type classification is performed using lithofacies, depositional facies, and Hydraulic Flow Unit (HFU) approaches. These are derived from capillary pressure data and analyzed using Rock Quality Index (RQI) and Flow Zone Indicator (FZI) methods. Next, well log data are classified into electrofacies using the Self-Organizing Map (SOM) algorithm, with refinement via core-calibrated supervised learning and Multi-Graph Based Clustering (MRGC) to optimize accuracy. Finally, permeability prediction is conducted using lithofacies-based and HFU-based models, integrating key reservoir parameters.
This methodology is applied to the field of carbonate reservoir. The study compares the effectiveness of the three classification schemes in predicting permeability, validated against core data. The results aim to provide insights into the capability of each method in capturing reservoir heterogeneity and improving subsurface model reliability.
© EAGE Publications BV