The capillary condensation caused by complex microscopic pore structures in shale reservoirs significantly constrains gas flow behavior, which is intuitively reflected as adsorption-desorption hysteresis in low-pressure nitrogen gas adsorption (LP-N2GA) measurements. This study integrates experimental measurements, theoretical analyses, and advanced machine learning algorithm techniques to quantitatively characterize LP-N2GA sorption hysteresis loops and systematically evaluate the effects of various factors on the sorption hysteresis performance of marine-terrestrial shale. The results show that the tested samples exhibit notable hysteresis loops between the LP-N2GA adsorption and desorption curves, with varying hysteresis areas. To quantify sorption hysteresis behavior, a novel computer vision-based image recognition approach has been applied to intelligently determine the area hysteresis index (AHI), which effectively captures the magnitude of the hysteresis loops. The AHI values range from 0.196 to 0.310, allowing for the classification of the selected samples into two distinct types based on their pore structure characteristics. Furthermore, sorption hysteresis is influenced by multiple geological factors, including organic-inorganic components, pore structure, and pore system heterogeneity. Among these factors, pore structure plays the most significant role in determining sorption hysteresis behavior. Additionally, a radial basis function (RBF) predictive model, developed using a data-driven approach, is proposed to quantify AHI under the combined influence of various factors accurately. The estimated results achieve highly accurate predictions with an accuracy of up to 97.6%. The study's findings are expected to provide valuable theoretical insights into gas migration mechanisms in marine-terrestrial shale reservoirs.
Hu et al. (Fri,) studied this question.