An integrated model for conglomerate reservoir quality classification based on graph attention and multiscale...

Area of Science:

• Geoscience
• Artificial Intelligence
• Petroleum Engineering

Background:

• Conglomerate reservoirs exhibit complex heterogeneity due to sedimentary and diagenetic processes.
• Identifying dominant factors and classifying reservoir quality in these formations is challenging.

Purpose of the Study:

• Develop an integrated AI framework for dominant factor screening and reservoir quality evaluation in conglomerate reservoirs.
• Improve the accuracy of identifying sweet spots and optimizing reservoir development.

Main Methods:

• Constructed a sample similarity graph to analyze geological associations.
• Employed a Point Graph Attention Network (Point Graph-GAT) for feature parameter importance evaluation.
• Developed a hierarchical Transformer-based multi-modal fusion (HT-MMF) model for multi-scale reservoir quality classification.

Main Results:

• Determined the importance ranking of characteristic parameters: Poisson's ratio > porosity > permeability > pore pressure > brittleness index > density > gamma-ray (GR) > shale content > tensile strength.
• The HT-MMF model achieved a final loss of 0.0006, outperforming benchmark models.
• Classification predictions for new wells showed high consistency with actual classifications (within 2% discrepancy).

Conclusions:

• The proposed AI framework effectively captures the heterogeneity of complex conglomerate reservoirs.
• This method provides a reliable approach for sweet spot identification and fine-scale reservoir development.
• The study highlights the significance of geomechanical and petrophysical properties in reservoir quality assessment.