Multimodal AI-Based 3D Reservoir Prediction for Integrated Subsurface Characterization

General Background: Hydrocarbon reservoir characterization remains challenging due to subsurface complexity and fragmented data sources. Specific Background: Conventional seismic-based interpretation often fails to capture fine-scale heterogeneity and uncertainty. Knowledge Gap: Limited studies integrate seismic, well-log, and satellite data within a unified AI-driven 3D framework. Aims: This study develops an AI-based 3D modeling system for accurate reservoir prediction using multimodal data fusion. Results: The proposed framework achieves high predictive accuracy, with LightGBM yielding R² values above 0.85 for porosity and 3D U-Net attaining IoU values exceeding 0.75 for structural segmentation. Novelty: The integration of transformer-based fusion and probabilistic uncertainty quantification distinguishes this approach from existing methods. Implications: The system enhances reservoir delineation, reduces exploration risk, and supports data-driven decision-making in hydrocarbon field development. Highlights:• Multimodal fusion improves reservoir prediction accuracy• AI-driven 3D modeling enhances fault and channel detection• Uncertainty quantification supports risk-aware decisions Keywords: Artificial Intelligence, 3D Reservoir Modeling, Seismic Data, Multimodal Fusion, Hydrocarbon Exploration

Ҳали таржима қилинмаган