To address the challenges of complex micro-fracture morphology, scarce annotated data, and susceptibility to mineral and noise interference in thin-section images, this study proposes a high-precision, few-shot intelligent fracture extraction method based on attention mechanisms. A data augmentation strategy combining multi-scale slicing and random RGB channel permutation was designed to enhance the model’s generalization capability for diverse fracture morphologies. Additionally, an attention mechanism module was embedded into U-Net to strengthen fracture edge feature extraction while suppressing interference from mineral boundaries and micropores. Experimental results demonstrate that with only three annotated samples, the proposed model achieves a 25.5

Buried hill reservoirs constitute a key unconventional hydrocarbon resource in o ffshore China, where fracture networks govern seepage capacity and production performance. However, their great burial depth, low seismic signal-to-noise ratio, and complex geophysical responses make accurate fracture prediction challenging. To overcome the limitations of existing curvature-based approaches-such as high noise sensitivity, poor detection of small-scale fractures, and lack of quantitative multiscale classification-this study proposes a novel multiscale curvature-attribute-based fracture prediction method. The proposed approach features three main innovations: 1) An attribute enhancement strategy that integrates local dynamic smoothing with a Gaussian distance-weighted grid scanning technique to suppress noise while amplifying weak responses, thereby improving robustness; 2) a multidirectional peak scanning algorithm that effectively extracts fractures without the information loss inherent in traditional binarization-skeletonization methods, enabling more complete and continuous fracture characterization; and 3) a median absolute deviation (MAD)-based dynamic grading model that classifies fractures into large-, medium-, and small-scale categories using curvature statistics. Application of this method to the Bozhong A buried hill gas field in the Bohai Bay Basin verifies its effectiveness: it extracts 5.4 x more cumulative fracture length than conventional methods and substantially enhances fracture detection resolution. Even under 15% noise contamination, the method retains 90.05% of its extraction performance, demonstrating strong noise resistance. The predicted multiscale fracture distributions (large, medium, and small) show high consistency with seismic horizon deformation patterns, drilling fluid loss records, well-log interpretations, and productivity data, confirming the reliability of the results. Importantly, medium-scale fractures are identified as the dominant factor controlling single-well productivity. The proposed method provides an effective pathway for quantitative multiscale fracture prediction in buried hill reservoirs, particularly within the weathering zone-the main hydrocarbon-producing interval.

Research on outcrop fractures and their connectivity provides crucial theoretical support and model-based guidance for fracture modeling, seepage simulation, and development evaluation in low-permeability hydrocarbon reservoirs. However, automated fracture extraction and connectivity analysis face challenges: Traditional methods fail in complex geological settings due to reliance on basic features like color/morphology; Limited labeled data impede deep learning for fine-scale fractures requiring labor-intensive pixel annotations; Existing connectivity approaches rely on time-consuming manual topology analysis. These gaps hinder accurate fracture network automatic characterization. This study proposes a deep learning model with an attention mechanism suitable for limited labeled data, called CBAM-U-Net (Convolution Block Attention Module U-Net), which integrates an attention mechanism to improve fracture details recognition. Additionally, data augmentation strategies are applied to improve the model's adaptability to limited labeled data. Furthermore, a Skeleton Convolution Topology (SCT) method is proposed to quantify the connectivity of extracted fracture networks. This approach first identifies and classifies fracture intersections (e.g., I-, Y-, or X-shaped nodes) using a skeletonbased convolution algorithm, then calculates the average connectivity number (CL) to objectively evaluate network connectivity. The results demonstrate that the proposed method, using only three manually labeled samples, accurately extracted fracture information from outcrop images, achieving a 29.6 % improvement in F1score (balancing precision and recall) and a 48.2 % increase in Intersection over Union (IoU) compared to the original U-Net model. Additionally, the fracture connectivity is quantitatively visualized. The CBAM-U-Net network combined with the SCT algorithm provides an efficient tool for geological fracture analysis, particularly in scenarios with scarce training data.

Thermal shock damage in deep shale hydraulic fracturing can impact fracture propagation behaviors, potentially leading to the formation of complex fractures and enhancing gas recovery. This study introduces a thermal-hydraulic-mechnical (THM) coupled fracture propagation model relying on the phase field method to simulate thermal shock-induced fracturing in the deep shale considering dynamic temperature conditions. The validity of this model is confirmed through comparison of experimental and numerical results concerning the THM-coupled stress field and thermal cracking. Special attention is paid to the interaction of thermal shock-induced fractures in deep shale that contains weak planes. The results indicate that thermal shock-induced stress significantly amplifies the tensile stress range and deteriorates rock strength, resulting in a multi-point failure pattern within a fracture. The thermal shock damage degree is closely related to the fracture cooling efficiency, suggesting that considering downhole temperature conditions in THM-coupled fracture stress field calculations is advisable. Thermal shock can activate pre-existing natural fractures and enhance the penetration ability of hydraulic fractures, thereby leading to a fracture network.

The significant heterogeneity of sandstone and mudstone interbedded reservoirs complicates the accurate evaluation of formation brittleness, thereby affecting the optimization of fracturing stage design. This paper first derives a statistical damage constitutive model to predict rock stress-strain curves under in-situ conditions, and the application of the model has been validated by experimental results. It then develops an energy-based brittleness evaluation procedure that accounts for the effects of lithology, physical properties and in-situ conditions on rock brittleness at a logging scale. The results indicate that highly brittle rocks can store significant elastic energy during loading and release it abruptly without requiring additional energy input. Rocks with a low Young's modulus can exhibit high brittleness, as the post-peak mechanical behavior of rock controls its overall brittleness. As confining pressure increases, the brittleness of rocks decreases, rocks with the same Young's modulus exhibits a 44.4% increase in brittleness when confining pressure decreases by 20 MPa. The newly proposed method effectively evaluates the reservoir energy-based brittleness index at the logging scale. This evaluation method is more sensitive to strongly heterogeneous formations and is more objective considering the cumulative impact of internal and external conditions without biased energy weighting.

The currently fracture prediction methods mainly focus on describing the fracture zone within a specific scale, we develop a new method to predict the multi-scale feature zone of buried-hill fractured reservoir. Firstly, we exploit the post-stack seismic structural attribute and pre-stack AVAz attribute respectively predict the large-scale and middle-scale fracture zone, and use the high-order wavelet transform to get a multi-scale merged attribute as the fracture intensity field. Secondly, we extract the fracture lines from the merged attribute using the Hough transform method, and by using the FMI logs to constrain the orientation of the predicted fracture, we establish a fracture orientation field. Besides, based on the power-law distribution relationship between the length and amount of the fracture, the fractal dimension is estimated. Finally, based on the fracture intensity field, fracture orientation field and fractal dimension, we exploit the distribution function and stochastic process to predict the multi-scale fault zone. The position of fracture zone is constrained by the fracture intensity field and realized by Poisson process. The length of fracture obey exponential distribution and constrained by the fractal dimension. The direction of the fracture is constrained by the fracture orientation field and obey Fisher distribution. The new method is applied in M gas field, Bohai Bay Basin, and the predicted large-scale fracture zone is consistent with the regional reservoir quality, and the middle-scale fracture is consistent with the production test data. The results is used in optimization of the designed development wells, and to some extent confirms the practical significance of the proposed method.

Pore fractal has now become an important research method to study reservoir pore heterogeneity and complexity, and is widely used in unconventional reservoir research. However, its guiding significance for reservoir research and the relationship between fractal dimension and reservoir pore structure parameters are not yet clear. This undoubtedly hinders the further development of fractal in reservoir research. This time, comprehensive research, including pore fractals, was conducted on over 700 high-pressure mercury injection experiments, providing guidance for reservoir pore fractal research. The research results indicate that there are significant differences in the application of fractals in different oil fields. Generally, the curves of high-pressure mercury injection experiments present 1- 4 segment fractals, among which the 2 and 3 segment fractals are the most common, and the 4 segment fractals are the least common. There is a correlation between the number of the fractal segments with permeability and pore throat sorting coefficient. The lower the permeability, the more obvious the multi-segment fractal features. The more fractal segments, the greater the fractal dimension corresponding to each segment. The number of fractal segments can characterize the distribution of reservoir pore size. The fractal dimension of large pore-throats is greater than that of small pore-throats. On the whole, there is a negative correlation between fractal dimension and permeability, porosity, and pore throat radius. But this correlation is unstable in different oilfields. And the average value of the overall correlation coefficient is only about 0.5. Different fractal segments in a multi-segment fractal represent different types of pores. The application of pore fractal dimension needs to be combined with pore types.

Fluvial facies is controlled by high-frequency base-level changes and frequent migration, evolution, erosion, and superimposition, forming structurally complex and strongly heterogeneous reservoirs. This leads to poor efficiency of oil recovery and dispersed remaining oil in fluvial facies oil fields, resulting in a much lower recovery rate compared to marine sandstone reservoirs. Based on the theory of Fluvial reservoir configuration, this paper proposes a set of quality characteristic parameters for complex sand bodies, which describe the characteristics of the reservoir architecture including the parameters related to structural features, geometric features, physical properties, and heterogeneity characteristics. The paper also proposes that the scale of the reservoir is a necessary condition for forming high-quality reservoirs, and the reservoir structure is the key factor controlling reservoir quality. Using the guidance of the complex sand body quality characteristic parameter and the reservoir quality control mechanism, an improved distance clustering method based on principal component analysis is used to establish a quality evaluation process for complex sand bodies.Taking the R23 unit of the lower section of the Minghuazhen Formation in the Qinhuangdao 32–6 oil field as an example, the optimal quality of complex sand bodies is evaluated. The best type of complex sand body has a larger sand body thickness, which is a superimposed or closely stacked complex point-bar sand body. The complex sand body has a simple sand body structure, good connectivity, and high initial and cumulative production, with good development effect. However, the bottom of the sand body is generally quickly flooded with water, and remaining oil is mainly distributed in the upper part of the sand body. The quality of the second and third types of complex sand bodies is moderate, and the exploitation effect is poorer than that of the first type of reservoir. They can be used as targets for adjusting the potential of production in the later stage of oil field development.

Cyclic steam stimulation is an effective thermal recovery method for heavy oil recovery. The key potential mechanism is the growth of the steam chamber after steam injection. Taking the LD5X heavy oil reservoir as an example, besides the interlayer developed in this area, the top water and bottom water distribute above and below the interlayer. These factors may have adverse effects on the development of the steam chamber, thus affecting the final heavy oil exploitation. In this work, our goal is to study the effects of interlayer permeability and well–interlayer distance on CSS performance (in the presence of top and bottom water). We developed a high-temperature-resistant interlayer. Based on the simulated interlayer, the field scale model was converted into a laboratory element model through the similarity criterion. In order to quantitatively evaluate the performance of steam stimulation, a thermal detector was used to measure the dynamic growth of the steam chamber and record the production data. The experimental results show that the self-made interlayer has high-temperature resistance, adjustable permeability, and little difference between the physical parameters and the target interlayer. During the cyclic steam stimulation process, the steam chamber presents two different stages in the presence of the top water area, namely the normal production stage and the top water discharge stage. The bottom water has little effect on the growth of the steam chamber. The small interlayer permeability, the increase in horizontal well–interlayer distance, and the existence of the interlayer will delay the top water leakage during steam stimulation. This study has reference significance for us to develop heavy oil resources with a top water barrier when implementing steam stimulation technology.

渤中26-6油田太古界潜山储层品质受风化程度与裂缝发育共同控制,为实现储层评价与储量动用,形成基于断?貌双控的太古界潜山储层综合预测技术.潜山上覆地层时代关系与沉积厚度决定了潜山风化剥蚀程度,基于此通过古地貌恢复实现储层风化程度平面分区.印支期形成的近东?西向高角度断裂与印支至喜山期长期活动的边界正断层对油田裂缝发育起重要控制作用,基于Radon变换与蚂蚁体技术实现裂缝发育平面预测.以风化程度平面分区为低频,裂缝发育平面预测为高频,利用高阶小波变换技术实现高、低频信息有效融合,综合表征储层品质.基于融合属性将储层划分为I类、II-1类与II-2类,I类储层品质最佳,为东营组地层沉积之前太古界潜山持续出露同时裂缝带发育区域.Ⅱ-1类储层品质次之,为沙河街组地层沉积之前太古界潜山持续出露同时裂缝带发育区域.Ⅱ-2类与II-1类储层品质接近,为东营组地层沉积之前太古界潜山持续出露但裂缝带一般发育区域.分类结果与测井及测试资料认识吻合,指导建立了油田储量动用策略,证实基于风化及裂缝发育主控因素开展地震储层预测,并基于属性融合方法实现结果综合的太古界潜山储层技术思路具有实用意义.

Turbidite sandstone reservoir is a typical gravity flow deposit,which is characterized by lateral variation,vertical multi-stage superposition and frequent migration.Seismic inversion is one of the main methods for fine reservoir description,but the complex fault system,lateral abrupt variation and overlimit thickness of the reservoir in E Oilfield restrict the accuracy of reservoir inversion and its subsequent application.In order to solve the problem of reservoir prediction in E Oilfield,a selffacies-control pre-stack inversion technology with complex fault system is proposed in this paper.Firstly,the deep learning algorithm based on the fault contact relationship chart library is used to construct the complex fault system model,and then the high-precision seismic stratigraphic framework is constructed.Secondly,a high-precision self-facies-control low-frequency model is built using the self-facies-control low-frequency model construction technology.Finally,under the constraints of high-precision stratigraphic framework and self-facies-control low-frequency model,self-facies-control pre-stack inversion is realized,which effectively improves the accuracy of sand body prediction near the fault,overlimit thick reservoir characterization and reservoir lateral boundary identification.The application in E Oilfield shows that this method has achieved good results.The thickness coincidence rate of the horizontal length of 16 new drilled development wells is 91%.

Fractures of multi-scales and multi-origins are primary storage space and effective seepage channels for metamorphic buried-hill reservoirs. They not only allow communication between various pores to enhance storage and seepage capacity, but also are essential for high yields. Fracture development and controlling factors at metamorphic buried-hill reservoirs of Bozhong 19-6 Gas Field were investigated based on imaging log, core data, experiments, e.g., thin sections and scanning electron microscope, and outcrop description. Results show that structural fractures, weathering fractures and dissolution fractures were developed in the metamorphic buried-hill reservoirs in the study area, among which structural fractures and weathering fractures are the most popular ones. However, fracture types varied obviously among different structure positions. Lithology, fault, weathering crust and ancient landform are primary factors affecting fracture distribution. Fractures were prone to be developed at lithology with high bright mineral contents. Faults were characterized by a dual structure including fault core and damage zone. Fracture density followed a decreasing trend with increasing distance from faults. Damage zone could be identified when fracture density was consistent with regional fracture density. The width of the damage zone was determined by factors such as fault scale and structure location. Well-connected weathering fractures were popular at the leached unit, with structural fractures of secondary importance. The unweathered unit was dominated by structural fractures with low density.

In order to efficiently develop deep-water turbidite sandstone oilfields complicated by faults in the block OML130, west Africa, high-precision development seismic technologies have been systematically studied and applied for reservoir fine description and production monitoring. Based on 3D seismic acquisition of high-density streamer and wide azimuth ocean bottom node (OBN), and the application of well data constrain seismic processing technology, seven sets of 3D seismic data with high signal-to-noise ratio, high resolution, high fidelity and wide azimuth were obtained successively. In terms of seismic inversion and interpretation, under the guidance of turbidite sandstone geological model, key technologies such as high-resolution prestack seismic inversion, reservoir fine division and correlation, time lapse seismic reservoir monitoring were proposed for high-precision reservoir description, production monitoring and remaining oil distribution prediction. The statistical results show that the prediction accuracy of seismic inversion of main reservoir thickness reached 95

Channel-feeding fan deltas without any conglomerates have recently been discovered on the steep slope in the Albert basin at the northern end of the west branch of the East African Rift. However, the new deposition system's mechanics and patterns are unknown. To reveal the sedimentary process of the distinctive fan delta, we integrate forward stratigraphic modeling (FSM) technology with interpretations of the core, well logs, and 3D seismic data. Firstly, a reference FSM model is constructed to describe the evolution of the fan delta, taking into account initial bathymetry, source supply, lake level, and fluvial discharge as independent variables. Then, the Orthogonal Experimental Design (OED) is employed based on the uncertainty analysis of each independent factor to generate FSM multi-realizations for formation thickness and Net-to-Gross. Sensitivity analysis of Multi-FSM realizations reveal that source supply is the most influential variable affecting formation thickness and texture. Combined with the tectonic evolution of Albert Basin, the formation dip during the deposition period is much smaller than present. Multiple incised channels were formed in the provenance area, which became the primary source system of the fan delta, facilitating the sediments' transportation over a certain distance to produce well-developed cross-bedding sandstone. The channel-feeding fan delta has a narrower plain and wider front, and the principal sand body is high-quality distributary channel. The discovery of the channel-feeding fan delta without any gravel formed on the steep slope of the rift basin enriches the theory of deposition in rift basins. In addition, the application of FSM technology helps to improve the quantification of sedimentary processes.

Natural fractures are regarded as important reservoir spaces and effective seepage channels at metamorphic buried hills. Continuous networks associated with multi-scale fractures with good connectivity is critical for high-quality reservoirs as well as high and stable production in the tight metamorphic rocks. The multi-scale fractures in Bozhong 19–6 metamorphic buried hills were well characterized through integrating image logs, cores, thin-sections, and scanning electron microscope, etc. After that, power-law distribution of multi-scale fractures was established to understand contribution of fractures to reservoir quality and figure out structure models of fracture networks as well as their impact on production. Results show that parameters of fracture systems vary regularly with fracture scales. Fracture development degree, e.g., cumulative areal density, increases as a power law function with decreasing fracture size from macro to micro (e.g., aperture and/or length), where storage space associated with micro fractures is also increased. Reversely, fracture connectivity and permeability follow a significant decreasing trend. Five structure models of fracture network were established based on combination pattern of multi-scale fractures: multi-scale fracture network with high-density and multi-sets, large-scale fracture network with medium-density and multi-sets, small-scale fracture network with high-density and multi-sets, large-scale fracture network with low-density and multi-sets, and small-scale fracture network with low-density and single-set. The former two fracture networks can be widely developed into high-quality reservoirs, contributing greatly to high and stable yields. Fracturing is required for the third and the fourth fracture networks to obtain stable production, while it is difficult for the fifth fracture network to obtain industrial oil and gas flow.

AbstractComplex fault patterns associated with rift development in the Hailar Basin were largely influenced by the Mongolia–Okhotsk Ocean and Palaeo-Pacific tectonic regimes during the Late Jurassic to Early Cretaceous periods. Based on 3D seismic data from the Beier Depression in the Hailar Basin, we characterized the reactivation history of multi-trend major faults and examined the evolution of the Beier Depression during the Early Cretaceous period. NE–SW-, NW–SE- and ENE–WSW-oriented major faults originated from strike-slip-associated structures that were pre-existing fabrics and then were reactivated and propagated upward under extensional regimes in the Late Jurassic. During the syn-rift stage (K1t–K1n), the Hailar Basin was in a NNW–SSE- to NW–SE-oriented extensional setting, and major faults of all orientations were active. There was tectonic quiescence (K1n1L) between the syn-rift stages (a rifting transition stage). The short compression stage after the syn-rift stage caused regional compressional deformation. During the post-rift stage (K1d–K1y), the extension direction rotated to an E–W orientation, and a new population of N–S-trending faults formed together with the reactivation of NE–SW- and ENE–WSW-trending major faults. Structural analysis shows that the major ENE–WSW-trending major faults were polycyclic growth faults reactivated via an upward propagation mode and that the NE–SW-trending faults were dip linkage faults reactivated via a dip linkage mode. The reactivation intensity of the NE–SW-trending major faults was stronger than that of the ENE–WSW-trending major faults. These results demonstrate the differences in the evolution of the different trending faults in the same tectonic regime, and the complexity of the final fault patterns in the Beier Depression was produced by differences in the reactivation of major faults. The originate interpretation of the multi-trend major faults in the Hailar Basin provides new insights into fault generation, and the classification of fault growth also has useful implications for future research on multiphase rifts.

通过测井、优势地震属性融合、频谱属性趋势分析等方法,对秦皇岛32-6油田北区新近系明化镇组下段Ⅱ油组(NmⅡ)复合砂体储层构型特征进行了研究,并厘清了基准面旋回对复合砂体构型的控制作用.研究结果表明:①秦皇岛32-6油田北区新近系明化镇组下段复合砂体主要为点坝-决口扇沉积,渗流屏障主要为河漫滩-废弃河道沉积,按测井相和砂体结构韵律划分内部结构,点坝砂体头部—中部为箱形,点坝尾部砂体为钟形,决口扇砂体主要为漏斗形,河漫滩和废弃河道主要表现为尖峰形和低幅齿形;复合砂体外部叠置样式包括孤立型、紧密侧叠型、疏散侧叠型和堆叠型.②研究区NmⅡ油组沉积早期,河道呈交织条带状分布,主要沉积点坝与决口扇,堆叠式和紧密侧叠型复合砂体广泛发育;沉积晚期河道演化为单一条带状,决口扇零星分布于河道凸岸,复合砂体规模变小,主要为疏散侧叠型和孤立型,河漫滩和废弃河道规模变大,成为复合砂体间的渗流屏障.③研究区长期—中期旋回中,构造活动等异旋回因素通过调整可容空间和沉积物供给,影响复合砂体外部形态和叠置样式;短期旋回通过控制沉积物类型和水动力等自旋回因素影响复合砂体内部结构.

现有叠前宽方位裂缝预测方法基于数据驱动,地质模式认识无法干预预测过程;渤海湾盆地M气田太古界潜山裂缝性储层主要受内幕带高角度断裂控制.为此,基于M气田宽方位地震资料,提出基于高角度断裂约束的方位傅里叶系数叠前反演裂缝预测方法.首先,利用f-k滤波与高分辨率Radon变换从叠后地震资料中提取高角度断裂信息,进一步映射为方位傅里叶系数反演方法中的先验权重,以此建立高角度断裂约束的裂缝预测反演目标泛函,最终求解得到裂缝密度与方位角信息.所提方法在M气田的应用效果表明,裂缝预测结果更具地质意义,与井上解释的裂缝特征以及各井生产测试情况吻合较好,可用于开发井井位设计与优化.

西湖凹陷是东海陆架盆地中重要的含油气凹陷,近期勘探证实渐新统花港组是当前主力产气层段;截至目前,西湖凹陷花港组层序地层划分方案还存在较大争议,严重制约了砂体等时对比及油气精细勘探进程.依据层序地层学结合天文旋回理论,利用小波变换和INPEFA技术对自然伽马曲线进行数学分析得到小波信号曲线、频谱及INPEFA曲线,建立了西湖凹陷花港组年代标尺及高精度层序界面格架.通过小波变换将GR曲线重构成不同阶次的小波信号曲线和频谱,以异常振动和能量团变化趋势凸显层序界面和旋回信息;同时,利用最大熵谱分析技术对GR曲线进行积分处理获得INPEFA曲线,以拐点和正负趋势定量拾取层序界面和旋回特征,将花港组划分为5套三级层序(自下而上分别命名为SQ1—SQ5).利用频谱分析明确三级层序内部地层叠置旋回,结合小波信号曲线变化特征,可将花港组划分为12套四级层序(自下而上分别命名为H12—H1).花港组GR曲线的多窗口频谱分析结果表明,5套三级层序发育持续时间介于1～3 Myr.通过多尺度多方法融合测井技术对缺少岩心和古生物化石区构建多级次等时层序地层格架具有一定推广意义,为渐新统花港组后期勘探开发提供一定的技术支持.