Advances and prospects of physical experiments and numerical simulations for microscopic residual oil characterization

The distribution and characterization of microscopic residual oil have long been considered a bottleneck in the development of high-water-cut oilfields. This study presents a systematic review and analysis of the advances and challenges in research methods for microscopic residual oil in the late stage of oilfield development. Accordingly, we propose tailored methods and establish an index system for selecting optimal techniques, thereby providing the foundation for scientifically robust strategies to tap residual oil potential in the oilfields. The results indicate that physical experiments remain central to uncovering the microscopic occurrence patterns and complex seepage mechanisms of residual oil. Their effective application depends on the characterization objectives. Numerical simulation techniques, based on the mechanisms underlying physical experiments, allow for the simulation of dynamic fluid distribution and the prediction of residual oil potential using mathematical and physical models. Compared to traditional methods, machine learning-based techniques offer distinct advantages in the intelligent processing of experimental images, pore structure identification, fluid phase differentiation, residual oil morphology recognition, and the prediction of residual oil distribution. Current research trends show a shift toward the deep integration of physical experiments, numerical simulations, and machine learning. To address the complex challenges of tapping residual oil potential in the late stage of oilfield development, future breakthroughs are required in three key areas. First, physical experiment techniques should be advanced toward higher resolution, more realistic experimental conditions, and broader data dimensions. Second, numerical simulation techniques should be enhanced to achieve greater accuracy and computational efficiency in characterizing seepage mechanisms under complex geological conditions. Third, machine learning-based techniques should highlight the intelligent identification and cognitive understanding of the occurrence morphologies and evolutionary patterns of microscopic residual oil. It is necessary to develop a decision-making model that combines high-precision in situ experiments, multi-scale simulations, and forward-looking strategies. Such a model will play a critical role in surging the tapping efficiency of residual oil potential in high-water-cut oilfields.