Application of Artificial Intelligence in Reservoir Simulation for Characterizing Multi-Phase Fluid Flow through Petroleum Reservoirs

Abstract

Artificial intelligence (AI) has rapidly advanced and influenced all scientific fields, including the petroleum industry, where it is no longer a novel concept. This article explores the evolving role of AI and its integration with traditional reservoir simulation approaches. The first section highlights the diminishing emphasis on conceptual and mathematical modelling in reservoir simulation, with a growing focus on sophisticated numerical solution techniques. This shift often neglects fundamental reservoir physics and mathematics, leading to the superficial characterization of multi-phase fluid flow in petroleum reservoirs. The second section examines the prevalent use of petroleum software packages, which heavily rely on input data without accounting for variations in data scales. These tools treat underlying programming as a black box, often bypassing critical basic sciences, such as reservoir conceptualization, applied mathematics, and numerical techniques, resulting in incomplete reservoir characterization. The third section discusses the role of machine learning (ML) and AI in reservoir applications. While data science plays a pivotal role, the lack of integration with fundamental reservoir physics reduces fluid flow analysis to an art devoid of scientific rigour. The final section proposes a hybrid approach that couples AI/ML with traditional reservoir simulation. This integration bridges the gap between science-based reservoir simulation and AI-driven fluid flow characterization, enabling the petroleum industry to achieve a new paradigm for multi-phase fluid flow analysis. By combining fundamental science with advanced AI techniques, this approach offers a comprehensive framework for accurate reservoir characterization and improved hydrocarbon production.