To improve the accuracy of shale gas production forecasting, an attention mechanism-based CNN-BiLSTM prediction model is proposed. First, CNN is used to extract time-series feature vectors, which are then fed into the BiLSTM network for bidirectional training. The attention mechanism focuses on key information, reducing information loss, and finally outputs the prediction results. Applied to Well X, the results indicate that the model achieves a MAPE of 20.69%, outperforming LSTM, CNN-BiLSTM without attention, and the CNN-BiGRU with attention. Moreover, a k-fold cross-validation was conducted on 10 shale gas wells within the same block, where the model achieved a 60-day prediction MAPE of 8.29%, with relatively small fluctuations in RMSE and MAE, demonstrating good stability and cross-well generalization capability.

The crude oil pretreatment system is a critical step in crude oil processing, and it is responsible for removing impurities to ensure that the quality meets the standard. Its fault can lead to equipment damage, decreased product quality, production interruptions, and environmental pollution. Therefore, it is crucial to promptly identify the cause of failures and analyze the current reliability of the system. To address this urgent issue, this article designs a fault diagnosis model based on object-oriented Bayesian networks and a system reliability assessment model based on dynamic Bayesian networks. By utilizing a fault diagnosis Bayesian network, both single and multiple faults in the system are successfully identified. Subsequently, the output results of the fault diagnosis network are used as key information input into a reliability assessment Bayesian network, enabling an in-depth and comprehensive analysis of the system's reliability status during fault occurrences. This integrated method not only possesses fault diagnosis and reliability assessment capabilities but also precisely analyzes the weak links in the system, providing valuable and practical suggestions and guidance to on-site operators and maintenance teams.

Summary Under actual reservoir conditions, characteristics such as the pore size and distribution of the matrix blocks, as well as throat connectivity, are highly complex. The complex heterogeneity of reservoir characteristics poses significant challenges to the study of the mechanisms underlying imbibition and the flow laws governed by imbibition processes. Therefore, based on the assumption of an ideal connected double-capillary model, we developed an imbibition double-capillary model with different capillary diameter distributions at the oil/water interface. By integrating the ideal connected double-capillary model with Hagen-Poiseuille’s law, the fundamental principles of capillary action, the law of mass conservation, and the continuity equation, we elucidate the mechanisms underlying countercurrent and cocurrent imbibition. Further, we analyze the imbibition behavior of the ideal imbibition double-capillary model in detail under various conditions of capillary radius, wettability, oil/water viscosity ratio, and interfacial tension (IFT), providing an in-depth understanding of the specific effects of capillary diameter distribution on imbibition phenomena. The research findings provide a theoretical foundation for analyzing low flowback rates after fracturing and well shut-in in low-permeability tight reservoirs, the distribution characteristics of fluids in porous media, and the study of imbibition behavior and associated flow laws at different scales. The results also offer guidance for accurately predicting oil and gas recovery rates.

This study employs coarse-grained molecular dynamics (CGMD) simulations to investigate the dynamic interaction mechanisms between hydrophobic silica nanoparticles (H-SiO2 NPs) and the cationic surfactant cetyltrimethylammonium chloride (CTAC) in an aqueous environment. A coarse-grained model based on the MARTINI 2.2 force field was constructed, and simulations were performed for 250 ns at 298 K in a 20 × 20 × 20 nm3 system containing one H-SiO2 NP, 1200 CTAC molecules, 1200 Cl- counterions, and 64,428 water beads. Key findings reveal that CTAC molecules assemble onto the nanoparticle surface via a stepwise mechanism of "monomer adsorption-micelle migration-cluster attachment," ultimately forming a core-shell structure: an H-SiO2 NP core surrounded sequentially by CTAC hydrophobic tails and hydrophilic headgroups. Radial distribution function (RDF) analysis demonstrates spatial segregation between CTAC tails (peak at 0.5-2 nm) and headgroups (peak at 4 nm), confirming their oriented alignment. The adsorption quantity progressively increased, exhibiting a stepwise transition at 205 ns due to micellar cluster attachment, culminating in approximately 350 CTAC molecules covering a surface area of 110 nm2 (20% coverage). A continuous decrease in system potential energy verifies the spontaneity of the adsorption process. The resulting H-SiO2 NP/CTAC aggregate attained a hydrated diameter of 5.5 nm, with partial exposure of hydrophobic C2 beads indicating incomplete surface coverage. By elucidating the dynamic adsorption pathway and core-shell formation mechanism of CTAC on hydrophobic nanoparticles, this CGMD study provides a theoretical foundation for optimizing nanofluid stability in petroleum extraction, designing high-efficiency heat-transfer working fluids, and enabling pollutant adsorption remediation.

In response to the complex multiphase flow problem caused by the in-situ precipitation of dissolved CO2 induced by pressure depletion in CCUS reservoirs, this paper constructs a microfluidic visualization experimental system at room temperature and high pressure (40MPa), which realizes the in-situ observation of the entire process of "dissolution depressurization precipitation", and introduces the time-varying surface laws of dynamic fluid pore fractal dimension and dynamic tortuosity fractal dimension to quantitatively characterize the topological complexity of the flow path. On this basis, a cross scale relative permeability fractal model coupled with microscopic phase transition dynamics mechanism was established, achieving a quantitative correlation between microstructure reconstruction and macroscopic seepage degradation. The research focuses on revealing the mechanical essence of channel reconstruction and flow resistance transition caused by pressure drop and gas separation: the gas phase center occupies a large pore throat, triggering the Jamin effect and forming a high resistance gas lock barrier; The water phase is forced to retreat to the corner of the pore and flow around the wall, resulting in a sudden increase in the specific surface area and extreme elongation of the actual streamline. This leads to a sharp increase in viscous shear dissipation in the boundary layer, causing a cliff like drop in the relative permeability of the water phase in the low water saturation range. The one-dimensional high-precision empirical formula set (R2>0.96) established by regression has a refined mathematical structure, which can not only provide digital theoretical support for evaluating the endogenous self plugging efficiency and cap rock leakage risk in CO2 geological storage at the mine level, but also seamlessly integrate it into the macroscopic Darcy flow simulator as a constitutive equation, providing key engineering value for quantitatively evaluating the endogenous self plugging efficiency of the formation, finely optimizing on-site injection and shut in schemes, and accurately predicting the long-term evolution trajectory of the storage plume across scales.

Machine learning (ML) leverages rapid response and high accuracy for widespread application in predicting elbow particle erosion, thereby facilitating pipeline safety monitoring. However, existing prediction models lack the capability to find the essential mechanism, resulting in insufficient reliability and generalization under data distribution shifts, which hinders their direct industrial application. Therefore, SHAP-ALE-ICE interpretable framework that explains the ML model to visualize erosion influencing factors and their acting behaviors was established in this study. Based on a computational fluid dynamics (CFD) generated dataset comprising 584 gas-solid flow conditions with nine input parameters covering pipe geometry, material properties, gas flow and particle characteristics, Support Vector Regression (SVR) model was selected due to outperforming evaluation metrics and robustness through combined measurement deviations. Shapley Additive exPlanations (SHAP) and input feature simplification analysis acquired that gas inlet velocity and pipe diameter are decisive features, with pipe bend radius ratio, pipe material hardness and particle diameter unimportant. Accumulated Local Effects (ALE) unveiled the decoupled effects of important features on erosion rate. Furthermore, Individual Conditional Expectation (ICE) plots discovered decisive factors have their interaction effect based on mass conservation. This study advances pipeline erosion research by de-black-boxing ML models for industrial practice.

Machine learning (ML) models offer rapid and low-cost prediction of methane (CH4) and carbon dioxide (CO2) adsorption in shale, which is crucial for enhancing recovery and achieving CO2 geological sequestration. However, with adsorption mechanisms model not yet fully established, existing purely data-driven ML lacks reliable physical constraints and exhibits weak interpretability, limited accuracy, and poor generalization. To address this gap, a novel fractal supercritical Dubinin-Radushkevich-Langmuir (FSDR-L) model was derived to describe the adsorption behaviors of CH4 and CO2 in shale and to directly quantify the critical pore size for gas adsorption mechanism transition. The results indicate that increasing temperature shifts CH4/CO2 molecules in shale toward monolayer adsorption, while reducing the contribution of pore-filling. Subsequently, a physicsinformed neural network (PINN) model guided by the insights of the FSDR-L model was developed for the first time to predict CH4/CO2 adsorption amounts in shale. The findings reveal that the PINN model achieved reductions of 38.93 % in mean absolute percentage error, 39.47 % in mean absolute error, and 57.46 % in root mean square error compared to the best-performing conventional ML model, demonstrating superior predictive performance and generalization capability in capturing complex shale-gas adsorption behaviors. Finally, to enhance the interpretability of the PINN model, a variance-based sensitivity analysis was conducted, revealing that total organic carbon, pressure, temperature, and pore volume are the key factors governing CH4/CO2 adsorption capacity in shale.

Spontaneous imbibition plays a significant role in numerous engineering applications, particularly in the oil and gas reservoir recovery over recent decades. In this study, a capillary-based spontaneous imbibition model is developed, incorporating tortuous, non-circular, and axially irregular capillaries. Based on fractal theory and the capillary imbibition model, a semi-analytical mathematical model is proposed to characterize spontaneous imbibition in porous media. By accounting for different flow velocities in pores of varying sizes, the model captures the dynamic heterogeneity of the imbibition front. The results show that for spontaneous imbibition in dry cores, sorptivity remains constant initially and then gradually decreases. In contrast, for oil-saturated cores, sorptivity first increases rapidly to a maximum value before gradually declining. This behavior addresses a key limitation of previous models, which assumed uniform imbibition dynamics prior to reaching maximum uptake. The new model is validated against published experimental data, demonstrating excellent predictive capability. Finally, a quantitative analysis is conducted on key parameters related to fluid properties and pore structure that influence the spontaneous imbibition process.

Gas adsorption in coalbed methane occurs primarily through micropore filling, a process complicated by pronounced structural heterogeneity. This complexity limits the accuracy of conventional adsorption models. To overcome these limitations, we developed a novel fractal adsorption model. This equation uses fractal dimensions, derived from low-temperature gas adsorption and high-pressure mercury intrusion data, to quantitatively describe the complex pore structure. By integrating this structural information with the Kelvin equation, the novel fractal adsorption equation elevates the R 2 value for adsorption capacity prediction to above 0.98. The results demonstrate that while adsorption capacity is influenced by pressure, temperature, and pore complexity, the spatial distribution of the pores themselves is the decisive factor. Ultimately, this model provides a critical distinction between the spaces where gas is stored (adsorption) and the pathways through which it flows. This study establishes a foundational framework for advancing our understanding of gas flow mechanisms in coalbed methane reservoirs.

In this study, a fractal-based model for apparent permeability is developed by embedding the tree-like branching network and capillaries within a porous matrix. The model incorporates multiple gas flow mechanisms, including slip flow and Knudsen diffusion. The results indicate that apparent permeability decreases with increasing pore pressure or decreasing maximum parent branch width. At low pressure and small maximum parent branch width, the flow is predominantly governed by Knudsen diffusion, while at high pressure and large maximum parent branch width, slip flow dominates. As pressure decreases or the maximum parent branch width decreases, the flow transitions from slip flow to Knudsen diffusion. In addition, higher porosity and larger fractal dimensions both contribute to higher apparent permeability. Moreover, the geometric parameters of the branching structure, such as the length ratio, width ratio, branch aspect ratio, and branch level, significantly influence apparent permeability. This model reveals the coupled effects of multiple transport mechanisms within the branching structure. It enhances our understanding of gas flow in the complex porous media of shale reservoirs.

Accurately evaluating the production deliverability of gas wells helps to understand the overall resource potential of a gas reservoir and to optimize production, which is of far-reaching significance in ensuring the sustainable development and utilization of natural gas resources. Conventional well testing methods, which require obtaining stable production rates and corresponding bottomhole flow pressures in the working regime, are not suitable for gas wells where only short blowout and shut-in data are available in the early stages of well construction. The empirical formulas established on the basis of the factors influencing the production deliverability of gas wells are not universally applicable. The deliverability equations for different gas reservoir types were established based on the seepage model during the early blowout of gas wells. The time of stabilization of the absolute open flow (AOF) under different factors was analyzed to determine the main controlling factors of the early production deliverability of gas reservoir establishment. Validated by the data of example wells in deep carbonate reservoirs, the calculated results of the newly established deliverability equation have an error of less than 5% from the theoretical binomial deliverability, which is of some reference significance for determining the initial deliverability of gas wells.

Pt coating has been considered a promising strategy to protect the Ti porous transport layer substrate in a proton exchange membrane electrolyzer from corrosion. However, despite significant advancements in understanding the composition, process, and structure of Pt coatings, the underlying causes of coating failure remains unclear. Here, we systematically investigate the failure processes of magnetron-sputtered Pt coatings on Ti substrates through ex situ electrochemical tests that simulate anode operational conditions. Three different thicknesses of thin Pt coatings were prepared on Ti substrates by controlling the sputtering time. As the polarization time increases, the coating gradually fails and the failure time is directly related to the thickness of the coating. Experimental results indicate that Pt coatings approximately 6 nm thick begin to fail after 24 h of operation, with interfacial contact resistance reaching 220.61 mΩ cm2 after 48 h, which is significantly higher than that of bare Ti at 28.39 mΩ cm2. Further analysis reveals the role of oxygen in the coating failure: oxygen atoms generated from electrolyzed water first diffuse from the coating surface into the interface between the coating and the substrate, and eventually, the gradual buildup of oxygen pressure results in the detachment of the coating. A deeper understanding of the failure mechanism of coatings can enhance the efficiency and durability of water electrolysis for hydrogen production by guiding the design of more durable coatings.

Spontaneous imbibition plays a significant role in numerous engineering problems, particularly in the context of oil and gas reservoir exploitation over several decades. To date, various models have been proposed to simulate this process. However, none of the existing models simultaneously consider the effects of variable cross-section, fluid viscosity, and gravity. This study assumes that capillaries exhibit tortuous, non-circular, and irregular axial variations. By fully considering the effects of fluid viscosity and gravity on imbibition behavior, it proposes a more generalized spontaneous imbibition model. The results calculated by the new model are compared with both published data and the numerical simulations conducted in this study, validating its predictive capability. The findings indicate that for symmetric C-D and D-C capillaries, the total imbibition time is independent of the arrangement order of capillaries with different diameters. In contrast, for asymmetric capillaries, the total imbibition time decreases as the number of pore-throat structures increases. When the number of pore-throat structures is large, the interface displacement is proportional to the square root of time (t1/2), exhibiting behavior similar to that observed in uniform capillaries. Based on this observation, an expression for the equivalent diameter is derived. Additionally, the imbibition velocity in non-uniform capillaries is significantly lower than that in uniform ones, with greater differences in pore-throat lengths leading to shorter total imbibition times. Under consistent wetting phase viscosity, a higher non-wetting phase results in longer imbibition times.

During the shutdown period of the J-Y refined oil product pipeline, there is a significant difference between the temperature of the transported oil and the outside soil temperature, which leads to a pressure drop in the pipeline after the shutdown. When the pressure at the high point of the pipeline drops below the saturated vapor pressure, air resistance occurs inside the pipeline, making it impossible to maintain pressure. This further poses a potential threat to the safe operation of the pipeline system. The SPS software was used to establish a pipeline model to simulate the equivalent soil temperature distribution along the pipeline. Through simulation and analysis, the inlet temperature error at the terminal station was effectively corrected, thereby obtaining the average soil temperature along the pipeline and the temperature drop range after shutdown. A fitting equation was utilized to reveal the relationship of equivalent soil temperature change over time. Combined with the pre-pump temperature measurement value and the equivalent soil temperature value obtained from the simulation, a shutdown operation was carried out after the system had been running for a period of time, and the trend of temperature and pressure changes after shutdown was simulated by the SPS model. The analysis shows that the temperature difference between oil and soil shows an exponential relationship with the pressure preservation time after shutdown. If the temperature difference between oil and soil before shutdown is small enough, vaporization is less likely to occur in the whole line after shutdown. If the temperature difference between oil and soil before shutdown is negative, the pressure in the pipe will increase after shutdown.

The gas adsorption capacity of porous media is significantly influenced by the complex pore structure and the size of solid particles. Despite this, a comprehensive adsorption model that captures the relationship between pore structure, solid particles, and adsorption capacity is lacking. Therefore, integrating fractal theory, a fractal particle with capillaries model is proposed. This structure is more consistent with the real adsorption of gas in porous media. In the fractal particle with capillaries model, gas adsorption begins with the smallest apertures and progresses to larger ones as the adsorption pressure increases. By integrating Langmuir theory with the Kelvin equation, we establish an adsorption model that accounts for the complexities of solid particles and pore structures. Our research reveals that as adsorption pressure increases, the adsorption isotherm can display three distinct growth patterns: exponential, reverse S-shaped and logarithmic. These patterns address the limitations of the Langmuir equation, which can only characterize adsorption isotherms with logarithmic growth. Additionally, the study explores how factors such as the capillary fractal dimension, solid particle fractal dimension, bending fractal dimension, and solid particle diameter influence adsorption capacity. The results indicate that higher capillary fractal dimensions, solid particle fractal dimensions, and tortuosity fractal dimensions correspond to more intricate pore structures, which lead to a greater internal surface area and enhanced adsorption capacity. And, smaller particle diameters contribute to increased adsorption capacity. Finally, fitting the model to real experimental data reveals that the established adsorption model performs effectively, demonstrating strong applicability and accuracy.

The Shenfu Gas Field faces challenges with uneven wellhead pressures, where low-pressure wells lose discharge capacity and high-pressure wells require throttling, leading to significant energy waste. Ejectors offer potential for energy recovery by utilizing high-pressure gas to boost low-pressure production. A computational fluid dynamics (CFD) model was developed using simulation software to simulate ejector performance. Parametric studies analyzed key structural parameters (mixing chamber length Lm, diameter Dm, nozzle spacing Lc, diffuser length Ld) and operational variables (compression ratio, working/entrained fluid pressures). Model validity was confirmed via grid independence tests and experimental comparisons (error < 10%). Network-level efficacy was verified using pipeline simulation software. Entrainment ratio (ε) and isentropic efficiency (η) exhibited non-linear relationships with structural parameters, with distinct optima depending on compression ratio. Dm had the strongest influence on ε. Higher compression ratios reduced ε, while increasing working fluid pressure or entrained fluid pressure improved ε. Optimal configurations were identified. Network simulations confirmed functional effectiveness, though efficiency diminished over production time. Ejector efficiency is highly sensitive to specific structural and operational parameters. Deployment in gas gathering networks is viable but most beneficial in early production stages.

For gas reservoirs with edge water, water easily transport into the gas reservoir along the water invasion dominant channel, which results that the water production is higher than expected during the development of gas wells, and then affects the production of gas wells. We had derived a water invasion unit numerical simulation model of oil–water two-phase (WINS). It simulated the water invasion dynamic change in the hyperosmotic zone. But the Buckley–Leverett water flooding front propulsion equation is no longer applicable to the saturation tracking method in the gas reservoirs. The main reason is that the fluid transport between the gas–water two-phases is different from the traditional incompressible fluid transport. In this paper, a new water invasion unit numerical simulation method based on intelligent proxies and optimal control theory in complex gas reservoirs (WINS-G) is proposed. Compared with the WINS, the saturation tracking method of water invasion channel is improved. And it demonstrates a new insight into the water transport in water invasion channel. Through the discussion of model fitting efficiency, it is found that the number of invalid grids is reduced by WINS-G based on the idea of proxy model, and the work efficiency is improved by automatic history fitting on the basis that the model can meet the basic requirements of accuracy.

The multi-stage development strategy is often adopted in the gas field. However, when the productivity decline occurs, many large processing stations will be severely idle and underutilized, significantly reducing operating efficiency and revenue. This study proposes a novel operation mode of multiple gathering production systems for gas field multi-stage development, integrating the decisions about processing capacity allocation and infrastructure construction to share processing stations and improve multi-system operating efficiency. A multi-period mixed integer linear programming model for multi-system operation optimization is established to optimize the Net present value (NPV), considering the production of gas wells, time-varying gas prices, and the capacity of processing stations. The decision of processing capacity, location, construction timing, and capacity expansion of processing stations, as well as transmission capacity of pipelines and processing capacity allocation schemes, can be obtained to meet long-term production demand. Furthermore, a real case study indicates that the proposed processing capacity allocation approach not only has a shorter payback period and increases NPV by 4.8%, but also increases the utilization efficiency of processing stations from 27.37% to 48.94%. This work demonstrates that the synergy between the processing capacity allocation and infrastructure construction can hedge against production fluctuations and increase potential profits.

Capturing the adsorption behavior of gases in heterogeneous shale is crucial for CH4 enhanced recovery and CO2 geological sequestration. This study derives a fractal Langmuir adsorption model that incorporates energy heterogeneity across pore sizes to characterize CH4 and CO2 adsorption in real shale. The results show that the fractal Langmuir model provides accurate fitting and high stability, effectively calculating gas adsorption amounts within various shale pore size ranges. It reveals that CH4 and CO2 are primarily adsorbed in micropores, with adsorption amounts influenced by the interplay of pore quantity and adsorption energy. In mesopores and macropores, gas predominantly exists in a free state, and the volume of free gas is dependent on pore volume. Additionally, a variance-based sensitivity analysis was applied to quantitatively assess the impact of shale pore structure (minimum pore diameter, lambda(min); maximum pore diameter, lambda(max); and porosity) on gas adsorption. The analysis demonstrates that lambda(min) has the greatest effect on adsorption, followed by porosity, while lambda(max) exerts a lesser influence. Increased porosity significantly enhances the adsorption capacity of shale for both CH4 and CO2. Finally, to overcome the physical assumptions of the fractal Langmuir model, a Convolutional Neural Networks-Gaussian Process Regression machine learning framework was developed that directly uses the N-2 adsorption-desorption curve at 77 K as input to predict CH4/CO2 adsorption in shale, achieving comparable accuracy