Application of nature-inspired algorithms and artificial neural network in waterflooding well control optimization

Abstract With the aid of machine learning method, namely artificial neural networks, we established data-driven proxy models that could be utilized to maximize net present value a waterflooding process by adjusting well control injection rates over production period. These proxies were maneuvered on two different case studies, which included synthetic 2D reservoir model and 3D (the Egg Model). Regarding algorithms, applied nature-inspired metaheuristic i.e., particle swarm optimization grey wolf optimization, perform task. Pertaining development models, demonstrated training blind validation results excellent (with coefficient determination, R 2 being about 0.99). For both studies algorithms employed, obtained using all within 5% error (satisfied level accuracy) compared with simulator. confirm usefulness methodology in developing models. Besides that, computational cost was significantly reduced proxies. This further highlights significant benefits employing for practical use despite subject few constraints.