Sample-Log and ML Facies Workflow for Legacy CO2 Well Prediction

🎙 Sample-Log and Machine Learning Workflow for Legacy Well CO2 Storage Assessment

In this study, Prasad Pothana presents an integrated workflow for predicting petrophysical properties in legacy wells and applying those results to 3D static modeling for subsurface CO2 storage assessment in the Delaware Basin. The approach combines traditional sample-log interpretation with machine-learning-based facies analysis, helping preserve geologic context while extending predictive capability into wells with limited log data.

Sample logs were first used to constrain lithologic interpretation, define formation-specific mineral associations, and support petrophysical evaluation in wells with more complete log suites. These interpreted wells were then used to generate zone-specific lithofacies classes from mineral fractions and calculate facies-dependent mineral statistics. Those facies labels became the training targets for machine learning models built from normalized gamma-ray and sonic logs.

After training, the models were applied to legacy wells with sparse data to predict lithofacies in intervals where direct multimineral evaluation was not possible. The predicted facies were then transformed into mineral-frame volumes, allowing sonic-based porosity to be estimated from a machine-learning-derived mineral framework. The workflow was further expanded to include bound and free water, producing a consistent set of facies-linked petrophysical properties for field-scale modeling.

Overall, the study shows how sample-log interpretation can be carried forward into a predictive facies framework, enabling mineral composition and porosity to be reconstructed in data-limited legacy wells. These results provide geologically constrained inputs for kriging-based 3D static models and improve reservoir and seal characterization for CO2 storage assessment.

Know what’s Happening Next — before everyone else does.

Get App

Host or Publisher The Permian Basin Chapter of the Society of Petrophysicists and Well Log Analysts (SPWLA)