Supervised Deep Learning-based Paradigm To Screen The Enhanced Oil Recovery Scenarios

This study presents a novel deep learning classifier to select the most efficient enhanced oil recovery (EOR) method based on reservoir’s rock and fluid properties, and temperature. The classifier consists of a one-dimensional (1D) convolutional neural network, long short-term memory (LSTM), and densely connected neural network layers. The genetic algorithm is used to tune the hyperparameters of the hybrid classifier. The classifier is developed and tested using 735 EOR projects on sandstone, unconsolidated sandstone, carbonate, and conglomerate reservoirs in more than 17 countries. The numerical and graphical investigations show that the structure-tuned deep learning classifier is able to accurately select the best EOR method.