Abstract
A primary goal of any Digital Oilfield initiative is to transform available data into reliable information that aids in decision-making. This involves analyzing multiple alternatives in operational decision-making to enhance production and reserves while minimizing operational and capital expenditures. As the oil and gas industry continues its digital transformation journey, integrating artificial intelligence (AI), machine learning (ML), and physics-based models has become essential for efficient reservoir management. Traditional reservoir simulation models are powerful but computationally expensive, making it challenging to conduct large-scale scenario evaluations and thereby optimization. Meanwhile, purely data-driven approaches, while fast, often fail to incorporate the underlying physics governing reservoir behavior, limiting their predictive accuracy over extended timeframes. This paper presents the application of modeling and optimization technologies to rapidly create models integrating AI/ML with reservoir physics, a methodology known as Data Physics. These models are designed to operate quickly, making them highly suitable for an optimizer to execute thousands of iterations. The multi-objective optimizer's role is to identify and prioritize the best scenarios for specific objectives such as optimizing water injection, well-flowing pressures, conversions/reactivations or addition of infill wells. Additionally, all these different objectives and decisions can be combined together into a comprehensive portfolio optimization of field development strategy. This strategy accounts for various interventions in the aforementioned areas while always considering operational and financial limitations to maximize Net Present Value (NPV).
