Advancing Low-Salinity Waterflooding for Improved Oil Recovery: Mechanisms, Machine Learning, and Future Directions

: The efficacy of low-salinity waterflooding for improving hydrocarbon recovery has garnered significant attention. The primary reason for utilising low-salinity waterflooding is its greater efficiency in improving oil recovery compared to conventional high-salinity waterflooding. However, despite its extensive acknowledgement, most of the studies have focused on core size using core flooding and spontaneous imbibition. Therefore, the comprehension of this phenomenon remains limited, leading to uncertainty about the feasibility of implementing large-scale applications. This is primarily attributable to the obscurity of underlying mechanisms and the inconsistency of research results. Furthermore, conventional techniques for estimating oil recovery factors through low-salinity waterflooding have inherent limitations. These methods are time-consuming, require specific data and expertise, and lack built-in uncertainty quantification. Machine learning has gained significance in the petroleum sector in recent years due to its adaptability and its ability to manage large datasets efficiently. The integration of machine learning into low-salinity waterflooding can provide an excellent opportunity to develop predictive models that effectively address constraints associated with conventional methodologies. Machine learning enables fast, accurate, and impartial estimation of oil recovery factors through advanced data-driven strategies. This study provides a systematic review of recent advances in the intelligent prediction of oil recovery factors using machine learning techniques in low-salinity waterflooding. It explores various machine learning models, their effectiveness in addressing existing challenges, and their potential to optimize low-salinity waterflooding performance in large-scale field applications.