Sedimentary facies modeling is a critical approach for understanding geological phenomena, yet the strong heterogeneity of reservoir systems poses a serious challenge for their refined characterization. In this study, we innovatively propose an interpretable attention-guided generative adversarial network framework with dual-domain learning, which achieves precise sedimentary facies modeling under the constraints of well facies and soft probability data. Specifically, we first effectively extract and preserve prior information of sedimentary facies models from both spatial and frequency domain perspectives. Then, during simulation, to enhance the capability of the network model for finely characterizing complex heterogeneous models, cross-spatial attention mechanisms are designed to effectively capture short-range and long-range dependencies between multi-scale pattern features. Additionally, through systematic feature map visualization analysis, we elucidate the processes of conditional fitting and complex sedimentary facies model reconstruction, intuitively demonstrating the functional mechanisms of each module. Finally, systematic experiments are conducted on multiple datasets to validate the effectiveness of the proposed method. The results demonstrate that the generated sedimentary facies models exhibit high consistency with training datasets in terms of visual realism and statistical indicators. Quantitative comparisons reveal remarkable performance of the method, achieving low Wasserstein distance (0.09), Kernel Inception Distance (0.0017) and Kernel Maximum Mean Discrepancy (0.21). These findings further confirm the high realism of the generated realizations regarding pattern features. This study offers a reliable and practical method for geological reservoir modeling, thereby advancing quantitative, precise geological research with broad application prospects.

The study of lacustrine gravity-flow successions, which are regarded as an important reservoir unit of tight oil and shale oil, has been now a hotspot and also a challenge study work. The Triassic Yanchang Formation in Qingcheng oilfield in Ordos Basin, as a typical reservoir of tight oil and shale oil, shows great exploration and development prospects. However, this oilfield did not achieve the expected development efficiency, probably resulting from the poor understanding of the distribution of lacustrine gravity-flow sandbodies. In this work, proper frequency-decomposed seismic attributes were select relying on their correlation to sand thickness, and then fused using machine learning with a supervised algorithm of support vector machine (SVR). A nonlinear mapping relationship (i.e., the trained SVR model) was established between the frequency-decomposed attributes and the thickness of sandbodies interpreted from well logs, and then the quantitative prediction of tight sandstone was realized through the application of the mapping relationship. The research indicates that: low-frequency seismic attributes are suitable for predicting thick sandbodies, while high-frequency seismic attributes are suitable for predicting thin sandbodies. Utilize advantages of seismic information of different frequencies, and consequently significantly reduces the uncertainty of seismic interpretation, and improves the prediction accuracy of sandbodies, and realizes the quantitative prediction of sandbodies. The test results show that the distribution trend and numerical range of intelligent fusion attrinbute are basically consistent with the sandbodies thickness interpreted by well logs, and the reliability of the sandbodies prediction by intelligent fusion attribute is significantly improved. The correlation between the intelligent fusion attribute and sandbodies thickness interpreted by well logs is improved from 0.6 to 0.79, and the prediction error of sandbodies thickness near the wells is less than 5 m. The geological interpretation indicates that the study strata of target formation are lacustrine-fan deposits, consisting of five sedimentary microfacies: branch channel, main lobe, lateral edge of lobe, slump body and inter lobe / inter channel. The main sandbodies is a fan-shaped, continuous deposition, whose thickness gradually decreases along the provenance direction. The branch channel is branched in the shape of narrow strip trees, which is developed above the lobe. Lobe is the dominated sedimentary microfacies in the study area. The slump body is the small scale isolated sandbodies formed by the collapse in the front of lacustrine-fan deposits. And the long axis direction of slump body is parallel to the front of lacustrine-fan deposits. This research results are of great significance for an efficient development of the oilfield in next stage.

Deep-water gravity flow sedimentation is topical in global petroleum and natural gas exploration. In the past 30 years, great achievements have been made in the sedimentary characteristics, controlling factors and sedimentary models of lacustrine gravity flow. However, due to the complex structure of continental sedimentary basins and the diverse development of gravity flow types, lacustrine gravity flow classification is still lacking. By systematically sorting out the development history and research into gravity flow, this paper summarizes a sedimentary classification scheme of lacustrine gravity flows based on sedimentary genesis and the main controlling factors, and summarizes the research into lacustrine gravity flows into the following three aspects. 1) Research into the main controlling factors of gravity flow: the sedimentary characteristics and models of lacustrine gravity flow in different types of basins are obviously different. These are mainly controlled by intra-basin factors and extra-basin factors. The intra-basin factors mainly include basin types (such as topographic slope), sedimentation rate and water density, while the extra-basin factors mainly include the composition and sources of sediments. 2) Gravity flow sand body distribution in source-sink systems: the formation and evolution of lacustrine gravity flow and the distribution pattern of sand bodies are usually controlled by the source-channel-sink system. This system effectively connects internal and external factors within the basin. 3) Lacustrine gravity flow classification: Continental lacustrine basins, compared to marine basins, are characterized by smaller scale, shallower water bodies, and intense tectonic activity. These characteristics result in diverse controlling factors and complex sedimentary features of gravity flow deposition. According to different research purposes, it can be classified according to sedimentary origin, development location or provenance supply, but the classification scheme of lacustrine gravity flow has not been unified. In this paper, lacustrine gravity flow depositions are classified into three major types and seven subtypes, by comprehensively considering the causal mechanisms and main controlling factors of gravity flows. Specifically, lacustrine gravity flow depositions are classified into flood-type (gentle slope sand-rich type, gentle slope sand-mud hybrid type, gentle slope mudrich type, steep slope sand-rich type, steep slope sand-mud hybrid type), slump-type (flexural slump type, fault-controlled slump type), and flood-slump coexisting type. There are plenty of research in the gravity flow of lacustrine facies at present. However, there is still much room for progress in fluid, flow transformation and genetic mechanisms of deep water channels. Therefore, the study of the formation, evolution and genesis mechanism of lacustrine gravity flows by integrating dynamics and sedimentology has become a current research and development trend.

Clastic reservoirs exhibit complex and diverse lithologies. Some lithological heterogeneities, occurring as thin but effectively low-permeability units, have pronounced impact on CO₂ flooding schemes and oil recovery. Thin low-permeability units within permeable sandbodies typically exhibit weak well-log responses, and are therefore of difficult recognition using conventional well-log analysis methods. To address this challenge, a hierarchical method is proposed for interpreting thin lithological heterogeneities by integrating wavelet transform and machine learning. The discrete wavelet transform enhances well-log responses of thin heterogeneities. An automated machine-learning framework is designed, which integrates multiple algorithms and achieves automated parameter optimization. This machine-learning method is then applied to well logs to establish a nonlinear mapping model between lithology and well-log responses. Additionally, the hierarchical nature of the workflow highlights lithological contrasts, facilitating a more accurate lithological differentiation by dividing the recognition of thin heterogeneities into three levels. Benefiting from these three advantages, the proposed method offers potential to significantly enhance the accuracy of well-log interpretations. The results demonstrate that this method yields accurate identification of lithological units as thin as 0.2 m for muddy beds and 0.3 m for diagenetic units, achieving a recognition accuracy exceeding the conventional well-log interpretations. This method also shows significant potential for broader applications, including the identification of other types of geological entities of limited thickness, and determination of reservoir parameters at fine scales.

With the development of unconventional hydrocarbon, how to improve the shale oil and gas recovery become urgent. Therefore hydraulic fracturing becomes the key due to the complicated properties of the reservoirs. The pore structure not only plays an essential role in the formation of complex fracture networks after fracturing but also in resource accumulation mechanism analyses. The lacustrine organic-rich shale samples were selected to carry out petrophysical experiments. Scanning Electron Microscopy (SEM) and X-ray Diffraction were performed to elucidate the geology characteristics. MICP, 2D NMR, CT, and N₂ adsorption were conducted to classify the pore structure types. The contribution of pore structure to oil accumulation and hydrocarbon enrichment was explained through the N₂ adsorption test on the original and extracted state and 2D NMR. The results show that micropores with diameter less than 20 nm are well-developed. The pore structure was divided into three types. Type Ⅰ is characterized by high porosity, lower surface area, and good pore throat connectivity, with free oil existing in large pores, especially lamellation fractures. The dominant nano-pores are spongy organic pores and resources hosted in large pores have been expelled during high thermal evolution. The content of nano-pores (micropores) increases and the pore volume decreases in Type Ⅱ pore structure. In addition, more absorbed oil was enriched. The pore size distribution of type Ⅱ is similar to that of type Ⅰ. However, the maturity and hydrocarbon accumulation is quite different. The oil reserved in large pores was not expelled attributed to the relatively low thermal evolution compared with type Ⅰ. Structural vitrinite was observed through SEM indicating kerogen of type Ⅲ developed in this kind of reservoir while the type of kerogen in pore structure Ⅰ is type Ⅱ. Type Ⅲ pore structure is characterized by the largest surface area, lowest porosity, and almost isolated pores with rarely free oil. Type Ⅰ makes the most contribution to hydrocarbon accumulation and immigration, which shows the best prospect. Of all of these experiments, N₂ adsorption exhibits the best in characterizing pores in shales due to its high resolution for the assessment of nano-scale pores. MICP and NMR have a better advantage in characterizing pore space of sandstone reservoirs, even tight sandstone reservoirs. 2D NMR plays an essential role in fluid recognition and saturation calculation. CT scanning provides a 3D visualization of reservoir space and directly shows the relationship between pores and throats and the characteristics of fractures. This study hopes to guide experiment selection in pore structure characterization in different reservoirs. This research provides insight into hydrocarbon accumulation of shales and guidance in the exploration and development of unconventional resources, for example for geothermal and CCUS reservoirs.

The applicability of sequence stratigraphic models to continental fluvial successions has long been topic for debate. To improve our understanding of how fluvial architectures record responses to changes in the ratio between accommodation rate and sediment-supply rate (A/S), two case studies are analyzed, including a densely drilled subsurface fluvial reservoir imaged with a seismic cube, and an outcropping fluvial succession. The subsurface dataset provides a larger, three-dimensional perspective, whereas the outcrop dataset enables observation at higher resolution. On the basis of both datasets, channel-body density, channel-body stacking patterns and their formative river types are interpreted at different scales, and how these may reflect responses to A/S change (the rate of accommodation creation relative to the rate of sediment supply) are discussed. The results indicate that (ⅰ) channel-body stacking patterns undergo four evolutionary stages along with the A/S increase, i.e., multi-story, mixed multi- and two-story, two-story, and isolated patterns; (ⅱ) channel-body density decreases along with the channel-body stacking patterns varying from multi-story to isolated; (ⅲ) formative rivers types are interpreted as evolving from braided planforms to braided-meandering planforms and then to meandering ones, with the increase of A/S.