EOR screening using optimized artificial neural network by sparrow search algorithm

Enhanced oil recovery (EOR) is a crucial aspect of reservoir engineering, and the use of machine-learning algorithms in the initial stages of screening has been widely accepted as a fast and efficient method for screening the most suitable EOR method. This study presents an artificial neural network (ANN) that recommends the most suitable EOR method based on historical reservoir data. Data from EOR projects worldwide were collected, pre-processed, and then used to build the ANN, which initially achieved a 69% accuracy. The neural network was optimized using the Sparrow Search Algorithm (SSA) and compared with the Particle Swarm Optimization (PSO) algorithm, with a focus on weight and hyperparameter optimization. Validation of the neural network's prediction was done using recall, precision, and the F1 score. Weight optimization yielded an accuracy of 68% with SSA and 34% with PSO, which were insufficient results for EOR prediction. However, hyperparameter optimization was applied, resulting in an accuracy of 94% with SSA and 90% with PSO. The SSA approach demonstrated faster convergence and higher accuracy in both optimization paths, highlighting its potential for optimizing the neural network in predicting the appropriate EOR method for a given reservoir.