During deep coal mining, the mine floor can be highly susceptible to water inrush under the combined effects of high in-situ stress and elevated pore pressure. To elucidate the mechanical response and failure mechanisms, a series of triaxial compression tests were conducted under a confining pressure of 38 MPa. The pore pressure was set to 0, 6, 12, and 22 MPa, and specimens of a siltstone mine floor were tested under both natural and saturated conditions. With increasing pore pressure, both the peak strength and elastic modulus of the siltstone decreased continuously. Meanwhile, the Poisson’s ratio increased, indicating less stiffness and enhanced radial dilation. The failure mode gradually changed from high-angle shear failure to a combined shear–tensile failure. This transition was accompanied by an increased number of fractures and the development of multidirectional, interconnected cracks. The elastic strain energy density decreased markedly as pore pressure rose, suggesting a reduced energy storage capacity before failure and a more concentrated energy release during failure. Under high pore pressure and saturated conditions, the plastic strain ratio decreased while the softening rate increased. These results reflect a behavioral transition from ductile slipping to brittle fracturing. This study reveals a coupled evolution mechanism of “stiffness weakening–fracture coalescence–abrupt energy release” in siltstone under high pore pressure. It also clarifies the physical causes of floor water inrushes and provides a theoretical basis for water inrush prevention and disaster prediction in deep coal mines.

The dynamic change of surrounding rock stress field in the roadway during the coal mining process leads to the development of fractures, resulting in the mining damage of overlying rock aquifer. As a result, the upper aquifer enters the mining space, increasing the risk of water inrush disasters in mines. Therefore, investigating the permeability evolution of fractured confined rocks under fluid-solid coupling conditions is of significant theoretical and practical importance. Focusing on the permeability evolution of the coal seam roof strata in the Shennan Mining Area under mining-induced disturbances, triaxial compression and permeability tests under fluid-solid coupling conditions were carried out on the typical weathered bedrock, sandstone, and burnt rock samples in burning area. Combining the acoustic emission monitoring results, the mechanical damage characteristics and permeability evolution patterns of different rock types were systematically analyzed, revealing the intrinsic mechanisms of permeability evolution under fluid-solid coupling effects. The results indicate that: The peak strength and secant elastic modulus of the three rock types exhibit a near linear growth trend with the increasing confining pressure, suggesting that confining pressure can enhance the rock strength while restricting the radial strain; There are differences between the acoustic emission ringing count and energy of different rock types, which is related to the mineral compositions. The confining pressure exerts a certain influence on the acoustic emission ringing count of different rock types; The permeability evolution of the three rock types under varying confining pressures can be roughly divided into four main stages, which correspond to the different stages of the deviatoric stress-strain curves; The variation trend of the acoustic emission damage variable is generally consistent with the permeability evolution. The node of permeability increase is near the crack initiation stress, indicating that the crack initiation and propagation of cracks are the dominant factors controlling the permeability evolution of different rock types. The experimental results can provide a theoretical basis for coal seam water conservation mining and mine water prevention in Shennan Mining Area.

ObjectivesWhile passing through a deeply buried water-rich fault fracture zone, a tunnel boring machine (TBM) frequently faces an extremely high risk of water inrushes, seriously restricting the engineering safety and efficiency. Methods This study aims to reveal the evolution patterns of water inrush disasters in the case where the TBM roadway tunneling passes through faults zones and propose technologies for advance prevention and control to mitigate the water inrush risk. Against the engineering background of the tunneling of a main roadway in a coal mine within a North China-type coalfield, this study established a 3D numerical model using the FLAC3D software to stimulate the TBM tunneling passing through a deeply buried fault fracture zone while considering the stress-damage-seepage coupling effect of surrounding rocks. Then, using the model, this study conducted simulation analyses of the spatiotemporal evolution characteristics and patterns of the displacement, plastic zone, permeability coefficient, and water inflow of surrounding rocks as the TBM approached the deeply buried fault. Based on the mechanisms behind water inrushes, a comprehensive reinforcement scheme centered on advance segmented grouting was proposed. Results and ConclusionsAs the TBM approached the fault, the surrounding rocks behind the tunneling face showed a significant three-stage spatial differentiation in displacement and plastic zone range. In contrast, the surrounding rocks in front of the tunneling face exhibited an exponential increase in displacement and plastic zone range and a surge of approximately 106 times in the permeability coefficient due to intensified damage. As a result, a continuous water-conducting fracture network was formed, ultimately inducing the overall instability and water inrush disasters of the surrounding rocks. The surrounding rocks presented an exponential increase in water inflow with an increase in hydraulic gradient under TBM roadway tunneling. Notably, at the critical water inrush distance (i.e., the distance between the tunneling face and the fault) of 3.0 m, the instantaneous water inflow reached up to 956.1 m³/h. After the reinforcement scheme of surface directional drilling combined with advance segmented grouting was implemented for the surrounding rocks within the influence range of the deeply buried fault fracture zone, the displacement, water inflow, and permeability coefficient of the surrounding rocks were effectively controlled, ensuring the tunneling safety. The results of this study provide a basis for the scientific decision-making regarding the TBM tunneling safety. Furthermore, these results hold significant engineering value for enhancing the environmental adaptability of TBM equipment under complex hydrogeological conditions and for improving the dynamic early warning system and the prevention and control technologies for water inrush mitigation.

Study region: Liulin County, located on the eastern bank of the Yellow River (YR) in Shanxi Province, China, is characterized by complex terrain including mountains and loess plateaus under a semi-arid climate. Study focus: To mitigate seasonal water supply instability and identify reliable groundwater sources, this study proposes a novel machine-learning framework integrating the Sparrow Search Algorithm (SSA) with Extreme Gradient Boosting (XGBoost). Utilizing Geographic Information System (GIS) and field surveys, a comprehensive hydrogeological dataset was constructed. The Boruta Algorithm (BA) was employed to eliminate redundant variables, while Random Forest (RF) evaluated feature importance. The proposed model was rigorously benchmarked against five alternative methods, including hybrids optimized by the Grey Wolf Optimizer (GWO). Furthermore, SHapley Additive exPlanations (SHAP) were applied to decipher the ''black-box'' nature of the models, quantifying feature contributions and non-linear interactions. New hydrological insights for the region: The results demonstrate that the SSA-XGBoost model achieved superior predictive accuracy, yielding a maximum Area Under the Curve (AUC) of 0.8812. Consensus from RF and SHAP analyses identified lithology and altitude as the dominant controlling factors, while the Normalized Difference Vegetation Index (NDVI) and rainfall provided essential spatial variability. GIS-based zonation revealed that approximately 22.06 % of the study area possesses high groundwater potential. This framework effectively balances high predictive accuracy with transparency, providing a scientifically robust tool for sustainable groundwater management in complex terrain regions.

Uncertainty in the spatial configuration of fracture networks is a primary factor influencing and deviating the simulation results of groundwater flow and radionuclide transport processes. Thus, accurately characterizing fracture networks to reduce such uncertainty is a predominant and challenging task in geological applications, particularly for the geological disposal of radionuclide waste. To address this challenge, this study aims to develop a sequential framework for the effective characterization of multi-dimension fracture networks, sequentially integrating the global optimization capabilities of genetic algorithm (GA) and pattern recognition strengths of deep learning (DL). First, the GA inversion model characterizes fracture networks from observation data and priori information, which serve as training datasets for data-driven DL models. However, owing to the stochastic nature of GA, uncertainty remains in the fracture network generations. Therefore, the DL inversion model subsequently reduces the uncertainty from GA and further improves the accuracy of the generations. We also develop a robust surrogate model based on an advanced self-attention mechanism to enhance the capability of capturing multi-scale features and meet the computational needs of GA iterations. The performance evaluation of the proposed framework quantifies the extent of reduction in the mean square error (MSE) loss. The results demonstrate that in the 2D fracture network case, the GA reduces the MSE loss by 61.3% compared to the initial random generations, with the DL models further reducing the loss to 89%. In the 3D fracture network case, the GA achieves a 37.1% reduction, while the DL models further decrease the loss to 68.3%. The uncertainty analysis results indicate that the proposed sequential framework effectively characterizes fracture networks and significantly reduces uncertainty.

Objective The failure depth of a coal seam floor represents a critical parameter for assessing the risk of floor water inrushes. With the widespread application of fully mechanized mining with a large mining height, existing empirical equations suffer from insufficient accuracy when used to calculate the mining-induced floor failure depths of coal seams with moderate thicknesses and above.Methods Based on the compressive strength of coal seam floors (i.e., hard, moderately hard, and weak floors), 107 sets of data samples of the measured floor failure depths of coal seams with moderate thicknesses and above under fully mechanized mining were classified. Subsequently, models for predicting the floor failure depth were established while considering four factors: mining height, the length of the mining face along its dip direction, coal seam dip angle, and mining depth. The resulting four theoretical models consisted of three linear models (i.e., the quasi-classical empirical equation, the linear support vector regression (SVR) model, and the log-linear mixed model) and one nonlinear model (i.e., the backpropagation (BP) neural network model). Then, the reliability of the four models was compared and verified using two evaluation metrics, i.e., goodness of fit (R2) and mean absolute percentage error (EMAP), as well as measured data from nine mines.Results and ConclusionsFor coal seam floors with three lithologies, the R2 values of the four models decreased in the order of the BP neural network model, the log-linear mixed model, the linear SVR model, and the quasi-classical empirical equation model sequentially, suggesting inferior goodness of fit of the latter two models. The log-linear mixed model, the linear SVR model, and the quasi-classical empirical equation model yielded EMAP values greater than the ideal threshold of 20%. However, the three models showed remarkable improvements compared to the classical empirical equation, which is no longer suited to guiding the prediction of the floor failure depths of coal seams with moderate thicknesses and above under fully mechanized mining due to discrepancies between calculated values and actual data. In contrast, the BP neural network model yielded average relative errors of less than 10%, significantly outperforming other models. By systematically comparing the prediction accuracy of the four theoretical models, this study provides a valuable reference for selecting an appropriate method for calculating the floor failure depths of coal seams with moderate thicknesses and above.

With the westward shift of coal resource development in China, downward extraction of close distance coal seams (CDCS) has become increasingly common, accompanied by intensified risks of overburden failure and water inrush. Aiming to replace simplistic empirical predictions with a robust, scientifically grounded model, this study investigates a typical CDCS working face in the Shaozhai Coal Mine using theoretical analysis, numerical simulation, physical similarity modeling, and field monitoring to characterize the progressive deformation of overlying strata and the evolution of the water-conducting fracture zone (WCFZ). Results show that upper-seam mining induces only minor delamination and slow subsidence, whereas downward mining causes severe roof collapse due to the combined effects of large mining height and upper-seam disturbance. The masonry-beam structure is significantly damaged, and fracture connectivity is markedly enhanced. The WCFZ develops through four stages—initial growth, rapid expansion, stabilization, and continuous evolution—and reaches a maximum height of 205.6 m, affecting the Zhiluo–Yan’an and Luohe–Yijun aquifers, which serve as primary water sources during mining. These findings provide a scientific basis for assessing water-inrush risks and guiding water-hazard control in CDCS downward mining in western coalfields. The study establishes a four-phase WCFZ evolution model under the "thin seam (No. 2) + thick seam (No. 8)" combination in Shaozhai Coal Mine. Downward mining of close distance coal seams intensifies collapse and fractures, increasing inrush risk. An integrated method combining empirical, numerical, physical, and field data improves fracture prediction and hazard control. Primary direct water sources were confirmed during downward mining of close-distance seams at Shaozhai Mine.

With the rapid advancement of intelligent mining in the coal industry, traditional water hazard prevention technologies have proven inadequate due to low efficiency and insufficient precision, failing to meet the precise control requirements under complex geological conditions. To address these challenges, an intelligent water hazard prevention system for intelligent mining was proposed. The proposed framework establishes a closed-loop management system covering the entire process from data perception, model construction, dynamic evaluation, integrated governance to emergency response, significantly enhancing both proactive measures and reliability in water hazard prevention. Firstly, through intelligent geophysical exploration (including “long-digging and long-probing”technology), dynamic seismic detection during drilling, and a multi-source monitoring network integrating electrical, microseismic, and hydrological parameters across the “well-ground-hole” triad, the accuracy of anomaly boundary identification was significantly improved. Combined with real-time data processing and dynamic imaging, this enables precise detection of water-conducting structures ahead of working faces. Secondly, leveraging intelligent exploration and monitoring achievements, multi-attribute high-precision modeling technology was developed. By integrating borehole data, seismic information, and real-time drilling data, combined with regional stratigraphic analysis, segmented LiDAR scanning, global calibration algorithms, implicit iterative interpolation algorithms, and TIN-GTP grid modeling techniques, critical hazards like aquifers and collapse columns through parametric modeling was mapped. This resulted in centimeter-level 3D roadway models and comprehensive multi-attribute hydrogeological models, supporting spatial correlation analysis of water hazard risks and enabling dynamic updates to static geological models with transparency. Thirdly, utilizing attribute information such as water-bearing/intermediate layers thickness, elevation, and hydrological monitoring data from the integrated mine hydrogeological model, we employed LSTM-GCN hybrid networks and Bayesian models. Through spatiotemporal information coupling algorithms, we evaluated flood risk assessment, achieved real-time prediction of mine water inflow sources, and conducted dynamic risk analysis via thermal maps of mine roof and floor water hazards. A dynamic evaluation system for mine water hazard risks was established. Subsequently, three innovative algorithms: A 3D geological model-based intelligent drilling trajectory design algorithm, an AI-powered video-driven intelligent identification algorithm for water probing and drainage drill rods, and a natural language report template generation algorithm were developed. A digital hydrogeological information management system and a smart control system for disaster-causing water discharge and grouting were also established. Through spatial overlay analysis of 3D trajectories with coal seam structures, multi-objective optimization models, and adaptive control algorithms, integrated autonomous drilling trajectory generation, remote monitoring and trajectory optimization was achieved, as well as automated dynamic updates of design reports. The efficiency of water hazard prevention and control engineering was significantly improved. Finally, by implementing a dual real-time analysis feedback mechanism for geological and hydrogeological data in water hazard management projects, information updates drived by synchronizeing mining engineering geology and hydrogeological data. Combining personnel positioning with dynamic tunnel model simulations of water surge propagation, a multi-constraint path dynamic optimization algorithm to rapidly generate disaster scenario simulations and rescue plans was utilized. Future research will focus on intelligent control platform for Internet of things and comprehensive smart systems, advancing water hazard prevention toward autonomous “sensing-analysis-decision” processes to provide core support for coal mine safety production and intelligent construction.

Few studies have comprehensively analyzed hydrochemistry and isotopes to assess the effects of mining on groundwater regimes. In this study, 153 samples collected before and after mining activities were categorized into five clusters (C1-C5) by machine learning and principal component analysis (PCA), based on external factors like water–rock interactions and human activities during different mining stages. Hydrochemical analysis showed that the Na⁺ and SO₄2⁻ mainly result from albite and halite dissolution, cation exchange, and potentially pyrite dissolution. The Ca2⁺ primarily originates from calcite and dolomite dissolution, also influenced by cation exchange. Agricultural runoff, fertilizer use, and soil infiltration all affected hydrochemical characteristics before and after mining. Isotopic analysis reveals that the pre-mining groundwater was recharged by atmospheric precipitation with minimal evaporation. The mine water resembles the Quaternary and weathered bedrock groundwater isotopically, suggesting these are its main recharge sources, likely due to enhanced hydraulic connectivity from mining. Post-mining, new recharge pathways emerged, increasing the recharge of mine and weathered bedrock water from evaporation-affected surface water. Entropy weightage quality index analysis indicates that most of the water can be made suitable for industry use, drinking water, and agriculture with treatment. These findings on the coupled hydrochemical and isotopic evolution improve our understanding of the effects of mining on groundwater systems and aid in preventing coalfield water hazards.

The extraction of coal seam could induce a series of ecological environment problems in arid and semi-arid mining area of Northwest China. In order to clarify the driving factors of ecological environment in coal mining areas and form an ecological risk prediction method, Yushen mining area was taken as the research object, the relationships between meteorology, soil, groundwater, aquiclude, mining parameters and ecological environment elements were analyzed, the ecological risk prediction model of coal mine area was formed, and the distribution of ecological risk index under coal mining conditions was predicted. The results showed that NDVI increased with the increase of precipitation and potential evapotranspiration, vegetation water consumption decreased first and then remains stable with the increase of water table depth. Under the condition of coal mining, the groundwater loss of loose aquifer decreased with the increase of the thickness of the aquiclude, the development height of water-conducting fracture zone was positively correlated with mining thickness, width and depth, the maximum surface horizontal deformation was positively correlated with mining thickness and width, and negatively correlated with mining speed. On this basis, the ecological risk prediction index system of coal mine was constructed, which was composed of nine key factors, such as drought index, surface slope, soil type, water table depth, aquiclude thickness, development height of water-conducting fracture zone, ratio of mining depth to mining thickness, ratio of mining width to mining depth, mining speed, and the grade division and corresponding value of each index were given according to the quantitative and qualitative characteristics. Taking the typical water-rich coal mine of Yushen mining area as an example, the comprehensive weight of each index was determined by the game theory based on the subjective and objective weight, the weight of development height of water-conducting fracture zone, water table depth, aquiclude thickness, drought index, ratio of mining depth to mining thickness was relatively large, and the sum of five weights accounted for 99% of the total weight, which played a decisive role in ecological risk prediction. The prediction results show that the areas with medium ecological risk accounted for 58.95%, and the areas with high ecological risk accounted for 41.05%, and there was no low-risk area. It was necessary to reduce the ecological risk of mining by adjusting the mining parameters to inhibit the development of water-conductive fracture zone and surface cracks, carrying out the reconstruction of aquiclude to increase the thickness of aquiclude, and restoring the damaged ecological environment. The research results could provide reference for the ecological environment protection and restoration in arid and semi-arid coal mining area of Northwest China.

[Objective]The mine water associated with coal mining tends to be rich in fluoride ions.If discharged dir-ectly without effective treatment,such water will cause severe pollution to regional ecology,affecting the quality of wa-ter resources and the stability of the ecosystem.[Methods]This study focuses on the challenging treatment of the fluor-ide pollution caused by coal mining-associated mine water.To overcome the bottlenecks including low efficiency and weak anti-interference of traditional methods for fluoride removal,this study designed a setup for fluoride removal us-ing the nucleation crystallization pelleting(NCP)processing and proposed a novel fluoride removal method-NCP chemical precipitation.The fluoride pollution of mine water poses great environmental risks since the resulting fluoride mass concentration in surface water might exceed relevant standards by 8‒15 times.The deep fluoride removal in a com-plex water quality environment is challenging in the prevention and control of fluoride pollution.This study developed a coordinated regulating mechanism integrating multi-phase reactions:chemical precipitation,nucleation induction,and porous adsorption.Process optimization experiments on a laboratory scale were conducted using a continuous flow chemistry system with a hydraulic retention time(HRT)of 45 min and an upward flow rate of 1.8 m/h.Then,this study systematically determined the dynamic process of fluorine migration and transformation using advanced characteriza-tion techniques such as X-ray photoelectron spectroscopy(XPS)and in-situ Fourier transform infrared(FTIR)spectro-scopy.[Results and Conclusions]The results indicate that the mass concentration of fluoride dropped from 12.6 mg/L to 7.6 mg/L(removal rate:39.8%)after only a single stage of processing under the optimized conditions(i.e.,a nucle-ation inducer(CaCl2)dosage of 1 200 mg/L and a seed loading ratio of 1∶50),with efficiency being 2.3 times higher than that of conventional coagulating sedimentation.X-ray diffraction(XRD)corroborated that thermodynamically stable aragonite and vaterite crystals were generated from reactions between Ca2+and F-.Notably,coexisting carbonates enhanced fluoride removal by forming CaCO3·CaF2 composite precipitates(FTIR reveals a characteristic peak at 1 080 cm-1)or porous calcite carriers(SEM images indicate a porosity increase of 62.76%).This study revealed the regulation pattern of interfacial reactions in a carbonate system.Energy dispersive spectroscopy(EDS)confirmed the gradient dis-tribution of fluorine elements in the cross section of formed particles,revealing the progressive removal mechanism from surface adsorption to lattice fixation.The NCP technique can effectively remove fluoride ions in mine water with com-plex water quality and can deal with the complex chemical composition in mine water.The results of this study will lay a foundation for the engineering application of the fluoride removal technology based on the NCP process while also providing a feasible technical route for solving the environmental pollution caused by fluoride-bearing mine water.

Due to an unclear understanding of the crack initiation mechanism during the process of splitting grouting at the top of the Ordovician limestone, there is a lack of a theoretical basis for the selection of key parameters such as grouting pressure, slurry water–cement ratio, and borehole azimuth. This study explored those problems using indoor physical model tests of the real triaxial and numerical analysis, considering factors such as the slurry water–cement ratio, in situ stress level, and crack inclinations and openings. The results of the real, triaxial, physical, similar- material model test were compared with a numerical simulation calculation of splitting grouting. The results show that in the process of splitting grouting, the fracture initiation pressure at the top of the Ordovician limestone decreased with increases in the water–cement ratio, and the secondary fracture initiation pressure was greater than that of initial fracture initiation. The crack initiation pressure increased with increases in the crack opening and the angle between the crack and the maximum principal stress (PS). When the maximum PS, intermediate PS, slurry water–cement ratio, and angle between the crack and maximum PS were fixed, the initiation pressure and secondary initiation pressure decreasd with increases in the difference between the maximum and minimum PSes. However, the decrease in secondary initiation pressure was greater than that of the initiation pressure. Under various slurry water–cement ratios, stress levels, and angles between the crack and maximum PS, the crack started at the tip and showed a characteristic trend of extending parallel to the direction of maximum PS.

[Background and Objective]Limited by locations,as well as economic and technical levels,most of the coal-based solid waste is still accumulated in the form of open-air landfills without treatment,thus occupying large quantities of land resources and causing secondary pollution to the environment.The proper treatment and reduction of high-salin-ity wastewater(e.g.,high-salinity mine water and high-salinity water from the coal chemical industry)represent a key link in the achievement of zero liquid discharge.However,existing technologies for high-salinity wastewater treatment are generally confronted with issues such as great initial investment and high operation costs of the treatment engineer-ing.[Methods]This study developed a technology for the synergistic treatment of coal-based solid waste and high-salin-ity wastewater(also referred to as solid-liquid synergistic waste backfilling).Specifically,high-salinity water,rather than ordinary water and additives such as early strength agent,and solid waste cementitious materials-used to replacing part of cement,were mixed while stirring to produce filling paste,which was then pumped to the underground goaves of coal mines.To analyze the feasibility of this technology,this study investigated the mechanical properties and potential envir-onmental impacts of filling paste prepared using coal-based solid waste and high-salinity wastewater from a certain coal mine in the Ningdong coal base in Ningxia.The mechanical properties,microstructures,and heavy metal leaching char-acteristics of the solidified filling paste were analyzed using the uniaxial compressive strength(UCS)test,scanning elec-tron microscopy(SEM),and inductively coupled plasma-mass spectrometry(ICP-MS).[Results and Conclusions]The results indicate that the strength of all the solidified filling paste increased over time but gradually decreased with an in-crease in the quantity of mineral powders added and a decrease in the proportion of cementitious materials.Notably,after some time,all filling paste prepared using high-salinity water as mixing water exhibited 3-day strength exceeding 0.5 MPa,meeting the minimum requirements specified in Technical specification for coal mine gangue-based solid waste filling(NB/T 11432-2023).Their 14-day strength reached 3.38-5.99 MPa,satisfying the requirements of vari-ous scenarios in most coal mine filling.The assessment results obtained using Nemerow's pollution index and extrac-tion toxicity tests indicate that the leachate from the solidified filling paste exhibited a comprehensive pollution index of heavy metals of 0.25,rated as"Safety"according to the grading criteria for comprehensive pollution assessment.The leaching test results of the solidified filling paste indicate that the primary pollutant concentrations in the leachate all fell below the requirements of Class Ⅲ water standard specified in Identification standards for hazardous wastes-Identifica-tion for extraction toxicity(GB 5085.3-2007)and Standard for groundwater quality(GB/T 14848-2017).Therefore,the technology for synergistic treatment of coal-based solid waste and mine high-salinity water can meet the relevant standards in the assessment of mechanical properties and environmental stability.This technology enables the recyclable,low-cost,and full quantitative utilization of coal-based solid waste and high-salinity water,enjoying significant econom-ic and ecological benefits.The results of this study will provide technological support for the construction of waste-free mines,mining cities,and chemical industry.

The intermittency of renewable energy sources often leads to surplus energy curtailment, emphasizing the need for efficient large-scale energy storage. Hydrogen, with its high energy efficiency and clean combustion, is an attractive energy carrier. However, its low density and stringent phase transition conditions limit large-scale storage applications on the surface. However, its low density and stringent phase transition conditions limit large-scale storage applications on the surface. Underground hydrogen storage (UHS) has been proposed as a solution for large-scale storage and utilization of surplus renewable energy. The hydrogen injection rate is a critical operational parameter, governing hydrogen storage and production efficiency. Balancing dynamic changes in key indicators (hydrogen production rate, dissolution rate, and storage mass) is essential. This study prioritized hydrogen production rate and dissolution rate (or storage mass) as primary objectives, employing multi-objective optimization to determine cycle-specific optimal injection rates. Advanced machine learning algorithms were used to develop and compare surrogate models across varying parameters and neural network architectures, identifying the most accurate predictive framework. This methodology significantly enhanced computational efficiency for both hydrogen storage modeling and optimization. The study established Pareto front for multiple objectives and provided corresponding injection rate schemes. Results demonstrated that the Long Short-Term Memory (LSTM) model achieved superior predictive performance, and dividing the Pareto front into three regions (low hydrogen loss mode or high storage mode, balanced mode, and high production mode) to meet different needs. These findings offer theoretical guidance for practical UHS applications.

The theoretical calculation of the height development of water conducting fracture zones under fully mechanized top coal caving conditions is crucial for the prevention and control of water hazard in coal mine roofs. With the advancement of thick top coal caving technology, existed empirical formulas cannot fully meet the prediction of water conducting fracture zone height in thick and extra-thick coal layers. Based on the measured values of the height of water conducting fracture zones in 185 sets of thick and extra-thick coal seams in China, according to the lithology of the overlying rock, the data is divided into three categories: hard, medium hard, and weak. Four mathematical models, including guideline-like formula, linear regression, quadratic regression, and DoseResp function, are used to fit and analyze the data. The model reliability is analyzed using the goodness of fit R2. Under hard overburden conditions, the goodness-of-fit is DoseResp model > quadratic regression model=linear regression model > guideline-like model. Under medium hard and weak overburden conditions, the performance of the goodness-of-fit is DoseResp model > quadratic regression model > linear regression model > guideline-like model. Furthermore, the mean absolute percentage error (MAPE) method is adopted to compare the reliability of the newly fitted formulas with those specified in the Guidelines for Coal Pillar Layout and Mining Under Buildings. Water Bodies, Railways and Main Shafts/Tunnels (hereinafter referred to as the Guidelines). For hard overburden, the accuracy of the four new fitted formulas was higher than that of the existing formulas in the Guidelines; for medium hard overburden conditions, the accuracy of the four new fitted formulas was better than that of the existing formulas in the Guidelines, and the DoseResp model performed the best. For weak overburden conditions, the guideline-like formulas and the DoseResp model are better than the existing formulas in the Guidelines. Applying the newly fitted formulas to specific mines, the DoseResp function has good applicability in predicting the height development of water conducting fracture zones in thick and extra-thick different overburden rocks (hard, medium hard and weak), followed by the guideline-like model, and both predictions are better than the formulas in the Guidelines.

Climate change has driven a global shift from fossil fuels to renewable energy sources. However, the inherent variability of renewable energy, influenced by temporal and climatic factors, presents significant challenges. Underground hydrogen storage offers a promising solution for retaining surplus energy. The complexity and heterogeneity of geological formations are difficult to accurately quantify, leading to large uncertainties in storage assessment results, and computation of forward modeling for large-scale sites is often time-consuming. This study introduced a numerical modeling framework incorporating the complex geological structures into the uncertainty analysis of formation porosity and permeability. We developed surrogate models to predict the hydrogen storage process using three machine learning (ML) algorithms: Extreme Gradient Boosting (XGBoost), Random Forest (RF), and Support Vector Regression (SVR). The study utilized the Sobol algorithm to examine the impact of variations in porosity and permeability on model output. This study applied and analyzed the numerical modeling framework at Wangjiawan in China. The findings indicated that the average final stability of hydrogen injection mass approximates 1800 tons, with the average production mass of hydrogen reaching approximately 950 tons. The XGBoost model demonstrated excellent predictive performance (R2 = 0.9679 and RMSE = 0.0318). Hydrogen production mass and rate are primarily influenced by the permeability of the formations, including injection and production wells during stable periods, while the impact of formation porosity is relatively minor. This study quickly and accurately predicts hydrogen storage processes under different geological parameters by employing ML algorithms. It also evaluates the importance of various geological parameters, providing crucial insights for effectively designing underground hydrogen storage facilities.

Because the law of slurry diffusion in horizontal grouting holes in deep limestone aquifers is unclear, and the key grouting parameters in grouting design and engineering practice lack a theoretical basis, the grouting process cannot be effectively controlled, and the grouting effect cannot be guaranteed. Therefore, this study comprehensively adopts laboratory tests, theoretical derivation, numerical analysis, comparative verification, and numerical simulation to study the influence characteristics of the theoretically calculated slurry diffusion form and inclination angle, which are similar to those of the physical model test and numerical simulation results. Under different grouting pressures, fracture inclinations (excluding 0°), fracture openings, and slurry water-cement ratios, the slurry diffusion form was approximately oval and symmetrical along the central vertical line, and the maximum and minimum slurry diffusion distances appeared at azimuths of 180°and 0°, respectively. The range of the slurry diffusion trace increased and decreased with an increase in the fracture dip angle in the lower and upper fractures, respectively. Under the same conditions, the slurry diffusion distance in the crack above the horizontal grouting hole was greater than that in the crack below. The crack initiation pressure decreased with an increase in the crack length (0.01-0.5 m), and the decreasing range increased with an increase in the maximum principal stress. Under the same maximum principal stress, the cracking pressure decreases with the increase of crack length. When the crack length is 0.01-0.5 m and the maximum principal stress is 15-25 MPa, the cracking pressure at the top of the Ordovician limestone is 16.73-27.85 MPa. The research results are helpful to provide theoretical support and basis for the practice of grouting reconstruction in the advanced area of karst fractured aquifer in the floor of North China coalfield.

Predicting mine water inflow is a complex challenge where traditional models often fail to accurately capture intricate temporal patterns. In this study, we introduce the Self-Attention mechanism into the prediction model. This model effectively captures long-term dependencies in time series, leading to more accurate predictions, particularly at data change peaks. The Self-Attention model proposed in this study exhibits superior performance in predicting trend components compared to Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) neural networks, particularly for sequences with high variability. This model holds great potential for coal mine water inflow prediction. Additionally, we propose a novel prediction framework that leverages Variational Mode Decomposition (VMD) to combine the strengths of multiple models. Compared to the best results achieved without the framework, the proposed framework and models demonstrate an average improvement of 26.12% in Mean Absolute Error (MAE) and 33.47% in Root Mean Squared Error (RMSE). Overall, the proposed framework harnesses the predictive potential of different models, significantly enhancing the accuracy of coal mine water inflow forecasting. The accurate prediction of mine water inflow can optimize mining production processes, enhance production efficiency, ensure safety and environmental protection, thereby leading to positive economic and social impacts.

The electrical characteristics of cement-based grouting materials are the foundation for implementing geophysical electrical exploration in coal mine grouting engineering. However, the preparation methods for high-resistivity grouting materials, which act as “contrast agents” in geophysical surveys, remain unclear due to the multiple influencing factors of resistivity. To address this, resistivity experiments were conducted using a self-developed apparatus to investigate the effects of curing age, admixture type, and dosage on the resistivity characteristics of grouting materials. This led to identifying optimal mix proportions for conventional cement-based high-resistivity grouting materials. Concurrently, mechanical strength tests were performed to analyze the impact of admixture dosage and curing age on compressive strength. The results indicate that the resistivity enhancement effects of four common cement additives can be ranked as follows: pyrophyllite powder >polyvinyl alcohol > air-entraining agent > fly ash. Considering both resistivity increase and mechanical strength, talc powder and polyvinyl alcohol emerge as viable candidates for use as additives in cement-based high-resistivity grouting materials. Specifically, with a talc powder dosage of 15%, the electrical resistivity of the grouted body after 28 days reached 4,966.7 Ω m, which is 119.1 times that of the control group (41.7 Ω m). Similarly, with a polyvinyl alcohol dosage of 1%, the resistivity reached 7,070.6 Ω m, which is 169.6 times that of the control group. These findings provide critical insights for developing high resistivity grouting materials with dual functionality as geophysical contrast agents and structural reinforcements.

Permeability (k) is crucial for subsurface fluid flow, but predicting k-values in tight sandstones remains challenging due to their complex pore structure and heterogeneity. Although machine learning (ML) has shown promise, it faces significant challenges, including limited high-quality data, high computational costs, and unclear prediction mechanisms. This study proposes a sparse data-driven knowledge discovery framework aimed at enhancing the accuracy and interpretability of k-value predictions in tight sandstone formations. We integrate ML models with data augmentation (ML-DA), using Extreme Gradient Boosting (XGBoost-DA) and Least Squares Support Vector Regression (LSSVR-DA), optimized through genetic algorithms (GA), particle swarm optimization (PSO), and Bayesian optimization (BO). SHapley Additive Explanations (SHAP) are employed to elucidate the interactions between key factors influencing predictions. Monte Carlo simulations demonstrate the robust performance of our ML-DA models, even under data constraints. SHAP analysis identifies key predictors, including porosity, displacement pressure, median pore throat radius, median pressure, and carbonate content. Partial dependence plots (PDPs) reveal a significant interaction between porosity and carbonate content, as well as a decrease in model stability at low carbonate content. This study presents an interpretable ML framework with data augmentation, enabling improved predictions from sparse data while exploring the interactions between key factors. The framework can be adapted to other domains facing similar challenges, enhancing the accuracy and transparency of model predictions.