Machine Learning Enables Data-Driven Predictions of CO₂ EOR Numerical Studies

An open-source framework is introduced for the event and analysis of machine learning- (ML) assisted data-driven fashions of CO₂ enhanced oil restoration (EOR) processes to foretell oil manufacturing and CO₂ retention. The primary goal of the authors was to extend the velocity, robustness, and accuracy of predicting oil restoration and CO₂ retention utilizing a whole open-source method combining Python programming, reservoir simulation, and ML strategies.

Overview of the Open-Supply Framework

The analysis of the predictive fashions was carried out utilizing two CO₂ water-alternating-gas (WAG) simulation circumstances, which have been proposed utilizing the SPE Comparative Answer Venture (CSP) 5 simulation mannequin as a reference. First, a reservoir simulation deck template and a configuration file, together with variable inputs, have been generated to create any variety of simulation jobs for every case. The input-value vary for the simulations was set in a Python script.