A Fast History Matching and Optimization Tool and its Application to a Full Field with More than 1,000 Wells

Large-scale CO2 injection for geo-sequestration in deep saline aquifers can significantly increase reservoir pressure, which, if not appropriately managed, can lead to potential environmental risk. Brine extraction from the aquifer has been proposed as a method to control the reservoir pressure and increase storage capacity. However, iterative optimization of the well controls for this scenario using high-resolution dynamic simulation models can be computationally expensive. In this paper, we demonstrate the application of a so-called coarse–grid network model (CGNet) as a reduced-order model for efficient simulation and optimization of CO2 sequestration with brine extraction. As a proxy, CGNet is configured by aggressively coarsening the fine-scale grid and then tuning the parameters of the associated simulation graph (transmissibility, pore volumes, well indices, and relative permeability endpoints) by minimizing the mismatch of well-response data (rates, bottom-hole pressure) and saturation distribution from the fine-scale model. Calibration and optimization procedures are automated using gradient-based optimization methods that leverage automatic differentiation capabilities in the reservoir simulator in the same way backpropagation methods are used in training neural networks. Once calibrated, CGNet is employed for well-control optimization. Validation with the fine-scale model shows that CGNet closely matches the optimized net-present value (NPV). Numerical examples using the Johansen model, available as a public dataset, shows that the optimization can be accelerated up to seven times using CGNet compared with a fine-scale model. (Using a compiled language will likely result in significantly larger speedups as small models suffer from a disproportionately high computational overhead when executed in MATLAB.) This study implies that a reduced-order model such as CGNet can be a powerful data-driven tool for faster evaluation of CO2 geo-sequestration simulation, combined with proper reservoir monitoring program.

show abstract

Enhancing Waterflooding Performance Using a combined Data Driven and Physical Modeling Approach

Grijalva,

Tellez,

González

et al. 2023

Day 1 Mon, October 02, 2023

View full text Add to dashboard Cite

The waterflooding implementation in an Amazonian oil field has been a game-changer in the field development strategy, becoming the main production drive mechanism and investment focus. About 40% of the daily oil production comes from waterflooding projects. Hence, it is imperative to preserve integrated reservoir and field operation management through a customized pattern balancing methodology that accounts for a need to optimize the injection-extraction relationship minimizing early water breakthrough and avoiding operational issues. This article presents a waterflooding pattern analysis tool that combines data-driven and physics-based Machine Learning models with a smart optimization workflow. This publication focuses on the theoretical foundation of the deployable prototype, which is based mainly on the application of an innovative physics data driven and ML model as well as its testing procedure. The tool has been tested in an area with nine deviated water injector wells and thirty-six deviated/horizontal producer wells, enabling quick analysis response based on different What-If and optimization scenarios. Users can assess the impact on production and waterflooding response by modifying operational parameters such as injection rates or liquid flow rates, or how to react if an oil-producing/water-injection well fails. The engineering and operation teams use and share a tool that avoids personalized spreadsheets with off-dated information and non-auditable metrics behind the results. The data preparation capabilities of the new tool speed up the interaction of data-driven and physics models and make a more efficient data flow process integrated with Capacitance Resistance Model (CRM) (Yousef et al. 2005) analytic model. The teams experienced a step-change in productivity by reducing a complete iteration analysis from 23 to 5 hours. The optimization workflow generates possible injector-producer relationships for pattern analysis and short (weekly) and mid-term (90-day) forecasts. Users can test different scenarios, choose the optimum, and submit subsurface focused well-operating recommendations to field operations.

show abstract

A Fast History Matching and Optimization Tool and its Application to a Full Field with More than 1,000 Wells

Cited by 7 publications

References 26 publications

Hybrid RGNet: A physics-based model enhanced by machine learning for history matching, characterization and future prediction

Hybrid RGNet: A physics-based model enhanced by machine learning for history matching, characterization and future prediction

Rapid Simulation and Optimization of Geological CO2 Sequestration Using Coarse Grid Network Model

Enhancing Waterflooding Performance Using a combined Data Driven and Physical Modeling Approach

Contact Info

Product

Resources

About