Optimizing subsurface carbon-energy synergy by balancing diffusion and convection via physics-informed Bayesian...

Area of Science:

• Petroleum Engineering
• Geoscience
• Computational Science

Background:

• Optimizing enhanced oil recovery (EOR) and CO₂ sequestration in shale reservoirs is complex due to competing diffusion and convection physics.
• Existing data-driven methods lack physical fidelity, failing to address these constraints effectively.

Purpose of the Study:

• To develop a physics-informed adaptive ensemble surrogate-assisted Bayesian optimization framework (AES-BO) for co-optimizing EOR and CO₂ sequestration.
• To address the limitations of black-box models in handling the diffusion-convection trade-off in shale reservoirs.

Main Methods:

• Developed AES-BO, integrating Gaussian process regression, polynomial response surface, and radial basis function networks.
• Dynamically weighted surrogates based on real-time cross-validation error to embed physical priors.
• Applied the framework to a field-scale shale oil model for CO₂-N₂ hybrid huff-n-puff optimization.

Main Results:

• Achieved a global optimum net present value of $64.2 million, outperforming other methods by 2.2-2.8%.
• Accelerated computation by up to 82.7% compared to existing techniques.
• Enhanced recovery factor by 8.23% using CO₂ for oil mobilization and N₂ for pressure maintenance, identifying an economic limit of three huff-n-puff cycles.

Conclusions:

• AES-BO provides a generalizable, physics-guided paradigm for intelligent design of low-carbon subsurface energy systems.
• The framework effectively links operational decisions to fundamental transport physics for optimized EOR and sequestration.
• Demonstrated the economic viability and enhanced efficiency of hybrid CO₂-N₂ huff-n-puff operations in shale reservoirs.