Omarkhan Samarkanov, Masoud Riazi School of Mining and Geosciences

Presented at: 6th EAGE Global Energy Transition Conference & Exhibition (GET 2025), Rotterdam, Netherlands

Researchers at Nazarbayev University have developed a novel Physics-Informed Neural Network (PINN) framework to model complex fluid flows in porous media. This work, presented within the "Carbon Capture & Storage" and student tracks at the prestigious EAGE GET 2025, addresses a critical bottleneck in reservoir engineering: the high computational cost of simulating multiphase flow in heterogeneous reservoirs.

THE CHALLENGE

Traditional numerical simulators (such as Finite Volume methods) require dividing a reservoir into millions of grid cells to track fluid movement accurately. This approach is computationally expensive, especially when modeling complex substances like emulsions, which dynamically alter rock permeability as droplets become trapped in pores.

THE AI SOLUTION

The team, led by Prof. Masoud Riazi and Omarkhan Samarkanov, replaced the traditional grid with a mesh-free deep learning architecture that "knows" physics.

• Physics-Embedded: Unlike standard "black box" AI, this model embeds the physical laws of mass conservation and Darcy’s law directly into the network's training process.

• Heterogeneity Handling: The model successfully predicts saturation profiles in both high-permeability "thief zones" and low-permeability zones, accurately capturing how emulsions divert flow to improve sweep efficiency.

KEY FINDINGS

• High Fidelity: The PINN model successfully replicated complex Finite Volume (FV) benchmarks, proving that AI can accurately model mechanisms like pore clogging without numerical dispersion.

• Speed & Scalability: The differentiable nature of the model allows for rapid inference, paving the way for real-time "Digital Twins" of reservoir systems that are far faster than traditional supercomputing methods.

WHY IT MATTERS: FROM EOR TO CCS

While validated on Emulsion Flooding for Enhanced Oil Recovery (EOR), the team emphasizes that this technology is a bridge to green energy solutions. The underlying physics—multiphase transport, trapping mechanisms, and pressure management—are mathematically identical to those required for Carbon Capture and Storage (CCS).

• Dual Application: This framework can be immediately adapted to model CO₂ plume migration and trapping in saline aquifers.

• Strategic Impact: By providing fast and accurate simulation tools, this innovation supports the optimization of secure carbon sequestration sites and geothermal reservoirs, contributing directly to the global energy transition.