Surrogate Model Optimization for WAG Oil Extraction Schedules
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for enhanced tertiary oil recovery, such as the water-alternating-gas (WAG) process, face challenges in optimizing oil extraction schedules due to limited CO2 supply and the need for time-consuming reservoir analysis, leading to inefficient oil extraction and early water or gas breakthrough.
Innovation Solution
A system and method that utilize a control system with processors to generate and refine fluid-and-gas ratio functions, selecting initial and modified trial schedules based on machine learning techniques like Bayesian optimization and Gaussian Process Regression to optimize resource extraction parameters, reducing uncertainty and improving oil extraction metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional reservoir analysis methods are used to develop WAG schedules, then the schedules are based on knowledge and prior experience, but the process requires a significant amount of time to develop
Solution Approach 1:
The patent creates a surrogate model that copies the essential behavior of the complex reservoir simulation model. This surrogate model can be evaluated rapidly without requiring full reservoir simulation, thus replicating the analytical capabilities while dramatically reducing computation time. The surrogate model is trained on simulation data and then used for rapid schedule evaluation and optimization.
Solution Approach 2:
The patent replaces the mechanical reservoir simulation process with a machine learning-based surrogate model. Instead of running time-consuming numerical simulations for each schedule evaluation, the system uses trained ML models that can predict reservoir response instantaneously, substituting the computational mechanics with statistical inference.
2Productivity
If CO2 injection volumes are increased to extract more oil, then oil extraction rate increases, but CO2 supply becomes constrained and more limited
Solution Approach 1:
The patent implements dynamic WAG schedules where the water-to-gas injection ratios and timing are continuously optimized based on real-time reservoir response feedback. This dynamic adjustment allows the system to maximize oil recovery efficiency at each moment, extracting more oil per unit of CO2 injected, thereby reducing overall CO2 consumption while maintaining high productivity.
Solution Approach 2:
The patent optimizes multiple injection parameters including WAG cycle timing, injection rates, and water-to-gas ratios. By carefully tuning these parameters, the system achieves more efficient CO2 utilization, maximizing oil displacement with limited CO2 supply. The optimization finds parameter combinations that prevent early breakthrough and improve sweep efficiency.
3Productivity
If WAG schedules are inappropriately designed to increase oil extraction rate, then production may improve temporarily, but water and gas breakthrough occurs early, making recovery viable only for short periods
Solution Approach 1:
The patent implements a feedback-driven optimization system where reservoir responses to injection schedules are continuously monitored and fed back into the surrogate model. The system learns from actual reservoir behavior and adjusts future injection schedules to prevent early breakthrough. This closed-loop control ensures stable, long-term recovery by adapting to reservoir conditions in real-time.
Solution Approach 2:
The patent uses the surrogate model to predict and evaluate potential breakthrough scenarios before implementing schedules in the field. By simulating various WAG schedules using the trained model, the system can identify schedules that are likely to cause early water or gas breakthrough and avoid them, performing preliminary optimization that prevents problematic behavior before it occurs.
Data Source
AI summary
System includes one or more processors that are configured to perform iterations of the following until a predetermined condition is satisfied. The one or more processors are configured to select a modified trial schedule. The modified trial schedule is selected based on initial fluid-extraction data and initial trial schedules and, if available, prior modified trial schedules and prior modified fluid-extraction data from prior iterations. The one or more processors are configured to receive modified fluid-extraction data generated by execution of the modified trial schedule with a designated model of the reservoir. The one or more processors are also configured to update the surrogate model with the modified fluid-extraction data and the modified trial schedule. For at least a plurality of the iterations, the modified trial schedule is selected, at least in part, to reduce uncertainty in a sample space as characterized by the surrogate model.