Well2Flow: Reconstruction of reservoir states from sparse wells using score-based generative models

2025-04-07

AI Summary

Accurately reconstructing spatially heterogeneous permeability and saturation fields from sparse well-log data in saline aquifers remains challenging. Method: This paper proposes a physics-embedded fractional-matching generative model that, for the first time, integrates multiphase flow physical constraints and well-log priors into a score-based generative framework (DDPM/Score SDE). High-fidelity numerical simulations generate joint training data, while conditional sampling incorporates vertical well-profile observations. Contribution/Results: The method achieves high-fidelity, physically interpretable joint reconstruction of 3D permeability and saturation fields under sparse two-well configurations. It significantly improves both reconstruction accuracy and physical consistency—ensuring mass conservation and Darcy-compliant flow behavior. Moreover, it demonstrates strong generalization across diverse geological scenarios, enabling robust reservoir management under data-scarce conditions.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories
• Natural Language Processing: Generation

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

This study investigates the use of score-based generative models for reservoir simulation, with a focus on reconstructing spatially varying permeability and saturation fields in saline aquifers, inferred from sparse observations at two well locations. By modeling the joint distribution of permeability and saturation derived from high-fidelity reservoir simulations, the proposed neural network is trained to learn the complex spatiotemporal dynamics governing multiphase fluid flow in porous media. During inference, the framework effectively reconstructs both permeability and saturation fields by conditioning on sparse vertical profiles extracted from well log data. This approach introduces a novel methodology for incorporating physical constraints and well log guidance into generative models, significantly enhancing the accuracy and physical plausibility of the reconstructed subsurface states. Furthermore, the framework demonstrates strong generalization capabilities across varying geological scenarios, highlighting its potential for practical deployment in data-scarce reservoir management tasks.

Problem

• Reconstruct reservoir states from sparse well data
• Model permeability and saturation fields in aquifers
• Enhance accuracy of subsurface state reconstruction

Innovation

• Score-based generative models reconstruct reservoir states
• Neural network learns spatiotemporal fluid flow dynamics
• Physical constraints enhance subsurface state accuracy