Volume 06,Issue 01

Modeling Performance of Foam-CO2 Reservoir Flooding with Hybrid Machine-learning Models Combining a Radial Basis Function and Evolutionary Algorithms

Authors

Seyedeh Raha Moosavi, David A Wood, Seyed Abbas Samadani


Abstract
The primary objective of this work is to demonstrate that a radial basis function (RBF) optimized with four different evaloutionary optimizers can effectively to predict with high accuracy the oil flow rate and oil recovery factor that results when a crude oil reservoir is flooded with foam-CO2. Injecting CO2 together with surfactant in the form of a foam can significantly improve a crude oil reservoir's sweep efficiency. Here, we couple a radial basis function (RBF) with evolutionary algorithms (particle swarm, imperialist competitive, genetic and teaching-learning based) to develop four hybrid-RBF prediction networks and apply them to predict efficiency of foam-CO2 flooding in oil reservoirs. A dataset with 214 published data records was compiled and used to train, optimize and test the four hybrid-RBF networks. The teaching-learning-based optimized (TLBO-RBF) model achieved the most accurate prediction performance, applied to the full dataset, for estimating oil flow rate (RMSE =0.0031, R2 = 0.997) oil recovery factor (RMSE =0.0175, R2 = 0.999) for the foam-CO2 injection EOR dataset. It can therefore be considered as another algorithm for assessing the impacts of foam-CO2 stimulation of oil-bearing reservoirs efficiently in cases where detailed core measurements are not available.

Keyword: Hybrid radial basis functions, Evolutionary optimization algorithms, Prediction performance.

PDF [ 1092.42 Kb ] | Endnote File | XML

CRPASE: TRANSACTIONS of



Follow Us

Google Scholar   Academia

JOURNAL IMPRINT