Optimizing Capex/Opex Using DataPhysics Models: A Comprehensive Approach to Conversions, Reactivations, Injection Distribution and In-Fill Drilling

Data Physics Models (DPM) integrate Machine Learning (ML) with traditional physics-based simulation, leveraging the strengths of both methodologies: the predictive ability of numerical simulations and the speed of data-driven, analytical models. DPMs are developed in a matter of months and run in minutes, enabling the possibility of running thousands of scenarios and the creation of a Pareto front. By evaluating multiple trade-offs across different objectives, engineers can implement closed-loop reservoir optimization.

Approximately 70% of oil fields have matured, requiring additional support for drive mechanisms and fluid displacement for improved oil recovery (IOR). This situation requires faster reservoir management practices to enable quick decision-making, such as adjusting water injection strategies to maintain or boost production. DPMs have proven to be ideal for these mature fields due to their ability to rapidly generate and optimize models, adapting to the continuously changing operational and reservoir conditions. Also, this approach has given impactful results for other field development plan activities like in-fill drilling site selection, bottom-hole pressure (BHP) management, well reactivations, and conversions. The novel approach described in this paper enhances these applications by optimizing across multiple workflows simultaneously. A “Pareto of Paretos” is generated by combining and optimizing these scenarios, offering a comprehensive, multi-dimensional solution for Capex/Opex allocation. Each point on this Pareto surface represents an optimal set of strategies for resource distribution, including injection control, in-fill well placement, well conversions, reactivations, and downhole pump pressure adjustments.

The case study presented in this paper demonstrates the practical implementation of this approach in a field in Colombia. A DPM was developed in three months to model field operations, generating individual scenario recommendations. The optimization framework effectively balanced Capex and Opex, resulting in a strategic plan that maximized Net Present Value (NPV) over a defined period. The outcome provided a detailed set of actionable insights across multiple operational areas, leading to more informed decision-making and higher financial returns vs. individual optimizations. The value creation has been identified to be greater than 10% vs the current state.

This combined physics-machine learning methodology exemplifies the next generation of decision support systems in oil and gas, especially for mature fields. It significantly improves the speed and quality of operational planning and capital allocation.