A hybrid AI-genetic algorithm framework for the optimization of polymer flooding strategies: a numerical...

Area of Science:

• Petroleum Engineering
• Artificial Intelligence in Energy
• Reservoir Simulation

Background:

• Declining conventional oil resources necessitate advanced recovery techniques.
• Polymer flooding is crucial for enhanced oil recovery but faces optimization challenges due to complex parameters and high simulation costs.

Purpose of the Study:

• To develop an efficient hybrid AI-Genetic Algorithm (GA) framework for optimizing polymer flooding.
• To overcome computational limitations of traditional reservoir simulations using machine learning proxy models.

Main Methods:

• Generation of a 960-case core-scale simulation dataset.
• Development of Feedforward Neural Network (FNN) and Elman Recurrent Neural Network (E-RNN) proxy models.
• Coupling a high-fidelity E-RNN proxy with a GA for multi-objective optimization.

Main Results:

• The E-RNN proxy model achieved high accuracy (R² > 0.99) in forecasting dynamic production data.
• Optimized parameters for maximum oil recovery include high permeability, injection rate, and polymer concentration, with minimized heterogeneity.
• Economic optimization indicated a preference for short, intensive injection periods to maximize profit, revealing a technical-economic trade-off.

Conclusions:

• The hybrid AI-GA framework effectively accelerates polymer flooding design and optimization.
• The study highlights the importance of considering both technical and economic factors in optimization strategies.
• Future work includes integrating laboratory data and scaling the framework to full-field applications.