The southwest margin of Ordos Basin has developed faults and fractures to varying scales, significantly enhancing the permeability of tight reservoirs and forming high-quality fault-fracture reservoirs against a backdrop of low permeability reservoir types. However, this also complicates the reservoir's homogeneity and the variability of its capacity, posing challenges for fine characterization of the reservoir's internal structure and for researching capacity control factors. To address these issues, the study employs spatial characterization of the fracture system, fracture classification, and capacity comparative analysis. Further, the development characteristics of the fault-fracture reservoir are defined through joint well-seismic and reservoir research. Key findings from this research include: 1） Establishment of the “binary four zones” model for fault-fracture reservoirs, which divides the reservoir into four distinct zones: the core fracture zone, the induced fracture zone, the micro-fracture transformation zone, and the matrix pore zone. Among these, the core fracture zone is identified as the primary contributor to production yield. 2） It is observed that the longer the fault extension length and the higher the solid drilling structure position, the higher the single well production capacity in the core fracture zone. The induced fracture zone's proximity to the fault correlates with higher production capacity, demonstrating the spatial impact of fault structures on reservoir productivity. 3） The production characteristics of fracture wells are categorized into three stages: ① the fracture system acts as the high yield stage; ② the fracture system serves as the storage stage and plays the role of diversion; ③ the fracture's primarily function in diversion. This research significantly increases the proportion of high-yield wells in the fault-fracture reservoir, providing crucial insights for guiding efficient, ongoing exploration and development activities in the Mesozoic reservoirs on the southwest margin of the basin. This strategic approach enhances understanding and management of fault-fracture reservoirs, optimizing resource extraction and improving operational efficiencies.

Significant progress has been made in shale oil exploration within the second member of the Funing Formation in the Qintong Sag, Subei Basin. However, geologists have noted that the measured Total Organic Carbon（TOC） contents are generally below 2%. Despite this, it is believed that source rocks in saline lacustrine basins can generate substantial amounts of petroleum even with low organic matter abundance, a concept known as the “low TOC” and highly efficient hydrocarbon generation mode of saline lacustrine source rocks. As hydrocarbon generation and expulsion occur during thermal maturation, the TOC levels in source rocks decrease. Therefore, accurately restoring the original TOC of these source rocks is crucial for a proper understanding of their hydrocarbon generation capacity, as well as for evaluating petroleum resources and supporting reserve growth and production enhancement. This study focuses on the typical mudstones and shales of the second member of the Funing Formation in the Qintong Sag. Techniques such as hydrocarbon generation simulation experiments, rock pyrolysis data, TOC and productive carbon content evolution during thermal simulation, and an element mass conservation method were employed to restore the original TOC of these mudstones and shales. The findings reveal that the TOC recovery coefficient of these saline source rocks can exceed values between 3 to 4, and this coefficient is significantly influenced by the lithology. The study found that: the TOC recovery coefficient increases with thermal maturity, the coefficient for laminated shale is higher than that for massive mudstone, and the TOC recovery coefficients for mudstones in the Ⅳ sub-member and shales in the Ⅰ to Ⅲ sub-members of the second member of the Funing Formation range from 1.1~1.5 and 1.5~3.0, respectively. These variations are attributed to differences in the type of organic matter and the hydrocarbon generating activation energy, leading to a higher hydrocarbon transformation rate and TOC recovery coefficient in shales compared to mudstones.

The objective of this study is to delve deeper into the hydrocarbon generation potential and oil-bearing characteristics of various source rocks within the second section of the Funing Formation in Qintong Sag, Subei Basin, and to assess the degree of shale oil enrichment. Through geochemical analysis and simulation experiment research, this investigation proposes a method for characterizing shale oil content that is adapted to the geological features of the second section of the Funing Formation. The study establishes the oil content, occurrence state, and distribution characteristics of shale oil within this section. Findings indicate that the mudstone and calcium mudstone of the fourth and fifth sub-members demonstrate higher total oil content, lower free hydrocarbon content, medium to high adsorbed hydrocarbon content, and limited mobility. Meanwhile, the calcium shale, calcareous shale, and laminated marlstone of the first to third sub-members exhibit higher total oil content and free hydrocarbon content, but lower adsorbed hydrocarbon content, which enhances their mobility. The calcium mudstone and calcareous mudstone of the first to third sub-members display a moderate level of total oil content and adsorbed hydrocarbon content but relatively low free hydrocarbon content. The marlstone from these sub-members shows comparatively low levels of total oil content as well as free and adsorbed hydrocarbon contents. The study also reveals the distribution of shale oil enrichment intervals. Class Ⅰ and Class Ⅱ shale oil enrichment intervals are primarily located in the middle and lower sections of the first and second sub-members in the deep depression zone. In contrast, the slope zone features limited development of shale oil enrichment strata but contains Class Ⅱ and Class Ⅲ shale oil.

The identification and classification of shale lithofacies are crucial for both theoretical understanding and practical applications in shale gas exploration and exploitation. This study focuses on the shale of the second member of the Paleogene Funing Formation in the Qintong Sag, Subei Basin, using core samples from a typical drilling well, Well-Qinye-1. The research involves whole rock/clay X-ray diffraction analysis on these core samples and employs a previously developed three-terminal diagram of shale mineral components to categorize the types present in this area. Additionally, a BP neural network method optimized by the ASO（Atom Search Optimization） algorithm was utilized to perform data mining on logging information. This process aimed to establish a prediction model for the relative content of clay minerals, siliceous minerals, and carbonate minerals, achieving quantitative characterization of shale mineral content through natural gamma ray spectrometry. Ultimately, the model was applied to predict lithology and identify lithofacies in the second member of Well-Qinye-1 and Well-Shaduo-1. The identification results closely aligned with the data measured from the samples, demonstrating high consistency. This study provides an economical, rapid, and efficient method for predicting shale lithofacies and main mineral components. It also offers a foundational approach for identifying well facies in scenarios where coring and direct testing data are unavailable.

The mud shales of the second member of the Paleogene Funing Formation in the Gaoyou Sag of Subei Basin are distinguished by their complex pore structures, well-developed bedding, and pronounced heterogeneity. These characteristics complicate the accurate evaluation of pore and fluid types using conventional well logging techniques. However, two-dimensional nuclear magnetic resonance（2D NMR） logging presents distinct advantages in fluid identification. The study incorporates the “Blind Source Separation” signal processing technique to perform data clustering analysis. Subsequently, a fluid zone model is applied to determine the fluid content with different movable properties in the reservoir space of the Huazhuang area. The findings demonstrate that 2D NMR logging significantly improves the accuracy of reservoir fluid evaluation, yielding reliable estimates of total porosity, effective porosity, oil saturation, and movable oil volume content. When compared to core analysis, the average absolute errors for effective porosity and oil saturation are minimal, at 0.4% and 7.3%, respectively. Further analysis of fluid properties within the primary lithologies of the second member of the Funing Formation reveals that the felsic-calcitic-dolomitic-mixed shales possess superior physical properties, oil volume content, and mobility. These characteristics render them as prime targets for shale oil exploration in the region. The successful application of 2D NMR logging in the Huazhuang area not only addresses the challenges associated with evaluating shale oil porosity and saturation but also provides essential insights for selecting “sweet spot” districts and forecasting reservoir production.

The intercalated shale oil within the first member of the Qingshankou Formation in the southern Songliao Basin exhibits significant exploration potential, primarily influenced by the quality of its intercalations, which impacts both oil content and productivity. The physical properties of these interlayers are notably heterogeneous, highlighting the necessity to characterize and describe interlayer reservoirs comprehensively and establish a suitable classification scheme. This study utilized advanced techniques such as field emission scanning electron microscopy, high-pressure mercury injection, and nuclear magnetic resonance to characterize the reservoir. Employing the fractal theory associated with mercury injection, a physical property classification standard tailored for sandstone interlayers was developed. This standard was combined with logging data to predict the physical property parameters of the reservoirs, facilitating the identification and mapping of high-quality interlayer reservoirs. The results delineate the first member of the Qingshankou Formation in the Daqingzijing area into regions of varying source rock maturity: low maturity areas with a vitrinite reflectance（R_o） of less than 1.0% and areas with middle to high maturity source rocks（R_o greater than 1.0%）. It was found that interlayer physical properties deteriorate as the maturity of the source rock increases. A grading standard for interlayer physical properties was established, categorizing the sandstone interlayers into Class Ⅰ to Ⅲ, and deeming some as invalid reservoirs. From Class Ⅰ to invalid reservoirs, there is a sequential decrease in the content of large and medium pores, with reservoir space transitioning from intergranular pores and intergranular solution pores to intragranular solution pores and intergranular pores. The mercury injection profiles evolve from weak platforms and slow straight lines to convex shapes, indicating a gradual degradation in oil content. High-quality interlayer reservoirs are predominantly situated along the main body of the estuary bar and the underwater distributary channels, with the thickness decreasing from southwest to northeast. The findings of this research provide crucial insights for targeting interbedded shale oil prospects within the first member of the Qingshankou Formation in the southern Songliao Basin, assisting in the strategic selection of exploration and development sites.

To address the issue of low natural imbibition recovery in horizontal wells within tight reservoirs and the challenge of replenishing formation energy during extended depletion phases, a strategy involving water injection huff and puff in large displacement horizontal wells was proposed. This approach builds on the fundamental characteristics of reservoirs and the established mechanisms of water injection huff and puff oil production in tight reservoirs. The research focused on the Chang-7 tight reservoir in Ordos Basin, employing a combination of natural and artificial cores. This study explored the range of natural imbibition and the dynamics of water injection and huff and puff across various displacements. Methods like nuclear magnetic resonance were used to analyze the characteristics of microscopic pore production, the impact of displacement on pore productivity, and the effects of soaking time. The findings reveal that with smaller displacements, mainly large pores are utilized for water injection, resulting in minimal engagement of small and medium pores. Conversely, higher water injections significantly enhance the involvement of small and medium-sized pores, thus substantially boosting the overall recovery rate. Additionally, as the simmering time is extended, the oil-water displacement effect increases, enhancing the degree of recovery through water injection huff and puff, though the rate of improvement eventually stabilizes. Numerical simulation was used to optimize the parameters for water injection huff and puff development in fracturing horizontal wells. For Well-A9, the optimal parameters were identified as a daily water injection rate of 900 m³ and a simmering time of 24 days. The field test confirmed the effectiveness of these parameters, with an initial daily oil increase of 2.11 tons, an effective period of 365 days, and a cumulative oil increase of 770 tons.

Continental shale oil horizontal wells have fast decreasing natural production and low recovery, for which single well water injection huff-n-puff can effectively replenish formation energy and improve recovery. Taking Ordos Chang7 shale oil as an example, numerical simulation method is used to carry out the optimisation study of water injection huff-n-puff in horizontal wells of continental shale oil. To enhance the accuracy of the numerical simulation model for shale reservoirs after volume fracturing, the Embedded Discrete Fracture Model（EDFM） is introduced. This model characterizes both natural and hydraulic fractures resulting from volume fracturing. Additionally, a conceptual model that considers imbibition and reservoir stress sensitivity is established. The timing, volume, and speed of injection, as well as the soaking period and huff-n-puff cycle, are optimized based on simulation results. These results indicate that a too-rapid injection rate causes water to flow along the fractures, decreasing the utilization rate of the injected water. As the huff-n-puff cycle increases, the oil increment per cycle tends to decrease. For the specific case of a shale reservoir in Ordos, the optimization of huff-n-puff parameters is as follows: Water injection should commence when the pressure coefficient drops to 0.706, with an optimal injection volume of 4 000 m³ at a rate of 300 m³/d. The recommended soak period is 15 days, with a total of six huff-n-puff rounds. This approach can increase the recovery rate by 4.95% and achieve a total oil-water replacement rate of 6.65%. This study provides valuable insights for water injection huff-n-puff strategies in shale reservoirs.

Energy storage volume fracturing is a pivotal early development technique for shale reservoirs, designed to supplement reservoir energy preemptively and significantly boost single well production. A method for predicting the maximum cumulative oil production during the development stage of energy storage fracturing is proposed, based on the mechanisms of imbibition and displacement coupled with the statistical analysis of actual production data. The results demonstrate that following a 30% flowback ratio, the cumulative oil production from energy storage fracturing exhibits a strong linear relationship with the logarithm of the flowback ratio. This relationship can predict the maximum cumulative oil production of a single well after fracturing. Validated by actual production data from other shale reservoirs, this method proves to be more accurate and universal than the decline curve analysis method. It encompasses a comprehensive evaluation of subjective and objective factors such as reservoir conditions, fracturing scale and technology, production system design, and drainage efficiency. Additionally, the method facilitates the determination of the liquid-to-oil ratio and the reasonable flowback rate. By controlling the average rate of discharge and production within the range of 6~8 m³/（d·km）, which aligns with the rates of oil drainage and imbibition, higher oil recovery and a lower liquid-to-oil ratio are achieved. This prediction method for maximum recoverable oil post-single well fracturing provides a basis for the economic benefit evaluation, production system optimization, and fracturing cost control of energy storage fracturing. It holds significant guiding importance for geological-engineering integration, well spacing optimization, and fracturing design.

To address issues such as the significant variance in shut-in effects between wells and the unclear effectiveness and timeliness of shut-ins during fracturing in shale formations, a numerical model integrating fracturing, shut-in, and production processes was developed. This model considers the synergistic effects of fracture closure, oil-water imbibition replacement, and fracturing fluid retention. The model's reliability was verified through simulations, exploring the oil-water migration law in the formation during the shut-in period of shale oil fractured wells and determining the optimal shut-in duration. Research indicates that: ① The oil-water migration characteristics during well shut-in undergo phased changes over time, which can be categorized into three main control stages: fracture closure, imbibition replacement, and energy balance. Additionally, these migration laws are closely related to the complexity of hydraulic fractures in space. ② Under the influence of capillary imbibition, extending the shut-in time appropriately benefits the initial production of fractured wells. However, an increase in fracturing fluid retention within the matrix can also exacerbate oil phase permeability damage. Based on the law of change in incremental oil volume, a reasonable shut-in time is identified to be between 30 to 45 days. ③ Considering real working conditions, “fracturing & well shut-in time” is proposed as an indicator for optimizing well shut-in, which aims to improve time efficiency and reduce differences between wells. This paper proposes an evaluation method and simulation workflow for assessing the well shut-in effects of volume fractured horizontal wells, offering valuable guidance in optimizing the reasonable shut-in time for shale oil fractured wells.

The imbibition displacement of fracturing fluid is a key technique for enhancing oil recovery in shale oil reservoirs. This paper assesses the impact of fracturing fluid wetting modifiers on the efficiency of imbibition displacement and explores the underlying mechanisms at play in shale oil reservoirs. Tests were conducted on the surface tension, interfacial tension, and wettability of the treating agent, alongside investigations into its compatibility with conventional fracturing fluids. Additionally, the imbibition displacement efficiency of rock samples with varying pore sizes was evaluated. The results indicate that the anionic surfactant AOS proved to be the most effective treatment agent for improving the wetting properties of fracturing fluids. The imbibition displacement efficiencies observed were 8.17% for particles, 17.55% for matrix, and 37.37% for fractured rock samples. These findings demonstrate that the imbibition displacement force encompasses buoyancy, buoyancy-capillary, and capillary forces. By altering rock wettability, the modifier significantly enhances the capillary force, thus boosting oil displacement efficiency by approximately 152.9% compared to conventional fracturing fluids. The influence of pore structure on the dominant imbibition displacement force was also noted. Capillary forces were predominant in small pores, and buoyancy is the main force for natural and hydraulic fractures. The research on the imbibition displacement mechanism of fracturing fluid provides valuable guidance for the efficient development of tight shale oil reservoirs.

Significant productivity breakthroughs have been achieved in key production layers of the shale in Jiyang Depression, notably the lower sub-member of the third member and the upper sub-member of the fourth member of Shahejie Formation. Despite these achievements, the development of these layers is relatively recent, and they exhibit considerable variation in individual well production. The primary factors influencing production remain unclear. Currently, a major focus of research is the comprehensive analysis of the main control factors for high production and the selection of reasonable fracturing parameters for shale oil horizontal wells. To better understand the impact of various factors on horizontal well production, factor correlation and pattern analysis are conducted using field data. Techniques such as gray correlation analysis and principal component analysis are employed to quantify the relationships between the average daily oil production over 90, 180, and 270 days and factors like the volume of fracturing fluid used and sand addition. Subsequently, a shale oil productivity prediction model is constructed, and fracturing parameters are optimized using SHAP（SHapley Additive exPlanations）. The research findings suggest that the volume of fracturing fluid, the amount of sand added, and the number of fracture events are the main engineering parameters affecting production. In contrast, geological parameters such as gray matter content, Total Organic Carbon（TOC）, and porosity significantly influence production as well. Over time, the impact of geological factors on production increases, while the influence of engineering factors diminishes during the later stages of production. Optimization analysis of fracturing parameters determined that a stage length of 40~45 meters, a fracturing fluid volume of 2 700 m³, and a sand addition volume of 180 m³ per stage are the optimal settings. These findings offer new insights for development determination and fracturing design in shale oil horizontal wells.

In the fracturing process of shale oil in Subei Basin, there has been a notable incompatibility between the oil displacement agent and the fracturing fluid. This issue leads to a reduction in the viscosity of the fracturing fluid and can even trigger chemical reactions that result in precipitation, adversely affecting both fracturing efficiency and productivity. To address this, a new fracturing-displacement agent（HDFD） has been developed specifically for shale oil. This agent is composed of maleic anhydride（C₄H₂O₃）, polyoxyethylene aliphatic alcohol ether（HO（CH₂CH₂O）_m（CH₂）_nCH₃）, anionic polyacrylamide（（C₃H₅ON）_n）, and white oil, produced through chemical synthesis and physical blending methods. In laboratory evaluations, the HDFD displayed an apparent viscosity of 9~12 mPa·s, a drag reduction rate of over 70%, an oil-water interfacial tension of 5×10^-3 mN/m, and a 40% increase in oil displacement efficiency at a 2×10^-3 kg/L concentration. Compared with the“drag reduction emulsion + high temperature oil displacement agent” system used in the field, these results suggest that HDFD performs exceptionally well in reducing drag in fracturing fluids and enhancing oil displacement. Field tests conducted on two wells using this agent showed a daily oil production increase of 40.6% and 84.6%, respectively. These outcomes confirm that HDFD is effective for use in the shale oil reservoir of the second member in Subei Basin and holds significant promise for future applications in integrated shale oil fracturing and displacement technologies.

3D geological modeling is essential at every stage of oilfield exploration and development. The scale, precision, and requirements of modeling vary depending on the issues at hand. While zonal block geological modeling guides well location deployment and the analysis of well group development and adjustments, the absence of a unified standard for each independent block model hinders the integration of smaller models into larger ones. This limitation affects the overall exploration, development, and planning of the oilfield. To maximize the precision and guiding role of geological modeling throughout the entire lifecycle of an oilfield, this paper uses an integrated oilfield, Gudao Oilfield, in Bohaiwan Basin as a case study. It introduces a multi-stage modeling strategy that progresses from a monolithic oilfield to zonal blocks and finally to reservoir configurations（single sand bodies）. These approach using geological modeling software, Direct and M³, with independent intellectual property rights and applying geostatistics and embedded reservoir configuration modeling with geological constraints to realize the transformation from macro to micro, whole to local, and general to detailed, providing an in-depth description of the oilfield’s structural and depositional characteristics, reservoir configurations, and spatial distribution of oil reservoirs. The 3D geological large-scale model and hierarchical model of the first multi-billion grid nodes of the oilfield in China are built. Such detailed modeling aids in deepening geological understanding, evaluating the potential of developed oil reservoirs, and locating remaining reserves, thereby ensuring the oilfield’s sustained and stable production. 3D geological multi-stage modeling technology not only facilitates multidisciplinary communication among petroleum professionals but also serves as a robust foundation for digital oilfield management. It represents an inevitable trend towards the development of digital and intelligent oilfields and offers significant guidance for oilfield exploration and development.

Underground gas storage（UGS） plays a crucial role in ensuring national energy security. Predicting multi-cycle well control parameters is key to the efficient operation of UGS facilities. Due to the complexities involved in understanding the operation patterns of W23 UGS and the challenges in developing effective schemes, a prediction model for well control parameters in multi-cycle operations has been developed using the rate-transient analysis method, and the change pattern of well control parameters with the development of UGS is analyzed. The process begins with preprocessing the dynamic production data from W23 UGS, which is then inputted into RTA software along with the reservoir parameters. The Dranchuk-Abu-Kassem model calculates the high-pressure physical properties of natural gas. Following this, the seepage characteristics of W23 UGS are examined, and the optimal wellbore conduit flow model is selected to determine the bottomhole pressure in the gas wells. The Blasingame pattern analysis method is then applied to fit the well control parameters and establish a basic model, such as the well control radius. By adjusting the model with production data, a refined prediction model for multi-cycle operation is developed. This model is utilized to categorize the wells of W23 UGS based on their production performance and control characteristics. The results categorize the wells into three types, with distribution ratios of 39%, 34%, and 27%, respectively. The predictive model reveals that during UGS operation, the well control radius of the gas wells initially increases significantly, then stabilizes. Similarly, the average well control volume of the gas reservoir continues to rise, and the effective permeability of the reservoir influenced by the three categories of wells initially increases before stabilizing.

The review of underground gas storage（UGS） capacity is influenced by multiple factors. Optimizing this capacity is crucial for achieving global optimization of UGS design parameters. The W23 UGS in Dongpu sag serves as a case study to explore the effects of formation water intrusion, salt deposition, stress sensitivity, and produced liquid on storage capacity utilization. A new model for reviewing the storage capacity, tailored to the specific conditions of W23 UGS, was developed and its feasibility was assessed. The results indicate that the W23 UGS exhibits low porosity and permeability, strong vertical heterogeneity, but weak lateral heterogeneity, characteristic of a weak edge water-driven gas reservoir. Factors such as the original gas storage volume, salt deposition, stress sensitivity, vertical production degree, formation water invasion, and produced liquid significantly impact the storage capacity of W23 UGS. The final check of the effective gas storage volume was calculated to be 3 679.51×10⁴ m³, which is 3.9% lower than the originally estimated volume. Regionally, the effective volume after the first-stage review was 3 013.47×10⁴ m³, representing 97.3% of the original storage volume. Meanwhile, an additional effective volume of 666.04×10⁴ m³, accounting for 91.1% of the original volume, was identified. The main reasons for the reduced review results include the impacts of stress sensitivity, formation water intrusion, and vertical utilization, which collectively led to a decrease in the effective gas storage volume.

The evaluation of underground gas storage（UGS） tightness is a critical aspect of the construction and operation phases. The Liuzhuang UGS in the Subei Basin, characterized by a fault-controlled fault nose structure and edge water UGS, underwent a comprehensive evaluation of its sealing performance. This assessment focused on both cap rocks and faults, incorporating rock mechanics experiments and theoretical analysis. The evaluation analyzed the sealing strength of the cap rock from macroscopic, microscopic, and mechanical perspectives. It was determined that the thick cap rocks are primarily mudstone from the upper part of the second member and the lower part of the third member of the Funing Formation. These cap rocks are widespread throughout the area with stable regional distribution. Their lithology is favorable, containing over 30% clay minerals. The cap rock's mechanical properties, including a Young's modulus generally below 20 GPa and a brittleness index typically under 45%, contribute to its effective sealing capabilities. The predominant clay minerals in the cap rock are illite-montmorillonite mixed layers, which easily expand in water, enhancing the oil and gas seal effectiveness. Furthermore, the fault analysis focused on both lateral and vertical sealing capabilities. The Liu① fault, characterized by significant fault throw and steep dip angles, demonstrated robust vertical sealing obtained through the combination of the positive pressure method on the fault plane and regional experiences with overlying strata depths. The lateral seal of the fault was substantiated using both qualitative methods, like sand-mud docking, and quantitative assessments, including the shale smear factor. These methodologies confirmed that the Liuzhuang reservoir-controlling fault possesses strong lateral sealing properties.

Fracture-cavity carbonate reservoirs are highly heterogeneous, presenting complex relationships between reservoir pore space and seepage flow. These complexities pose significant challenges for in-situ stress analysis, selection of injection and production parameters, and evaluations of mechanical integrity. In order to further clarify the variation of in-situ stress during the operation of fracture-cavity carbonate underground gas storage（UGS）, ensure the mechanical integrity during the operation of UGS and increase the upper limit pressure, the model was developed to analyze the stress distribution in fracture-cavity carbonate gas storage and to monitor the variations in four-dimensional in-situ stress. This model also assesses the mechanical integrity across different pore spaces. The findings reveal that: ① Stress concentration is more pronounced in fracture-cavity carbonate reservoirs than in homogeneous ones, with the lowest stress levels often occurring at cavity boundaries. ② Pore pressure and stress fluctuations are more severe in fracture-cavity environments, increasing the likelihood of shear or tensile failures at cavity boundaries during UGS operations. ③ During gas production, shear failure tends to occur along the direction of minimum principal stress, whereas tensile failure is more probable along the direction of maximum principal stress during gas injection. ④ Compared to homogeneous reservoirs, fracture-cavity reservoirs are more prone to tension or shear failures during gas injection but are generally safer during gas production, though shear failures around cavities are more likely. These results provide valuable theoretical and methodological insights for in-situ stress analysis and mechanical integrity assessments of fracture-cavity carbonate UGS.