Geothermal energy is a sustainable energy source that meets the needs of the climate crisis and global warming caused by fossil fuel burning. Geothermal resources are found in complex geological settings, with faults and interconnected networks of fractures acting as pathways for fluid circulation. Identifying faults and fractures is an essential component of exploiting geothermal resources. However, accurately predicting fractures without high-resolution geophysical logs (e.g., image logs) and well-core samples is challenging. Soft computing techniques, such as machine learning, make it possible to map fracture networks at a finer resolution. This study employed four supervised machine learning techniques (multilayer perceptron (MLP), random forests (RF), extreme gradient boosting (XGBoost), and support vector regression (SVR)) to identify fractures in geothermal carbonate reservoirs in the sub-basins of East China. The models were trained and tested on a diverse well-logging dataset collected at the field scale. A comparison of the predicted results revealed that XGBoost with optimized hyperparameters and data division achieved the best performance than RF, MLP, and SVR with RMSE = 0.02 and R2 = 0.92. The Q-learning algorithm outperformed grid search, Bayesian, and ant colony optimizations. The blind well test demonstrates that it is possible to accurately identify fractures by applying machine learning algorithms to standard well logs. In addition, the comparative analysis indicates that XGBoost was able to handle the complex relationship between input parameters (e.g., DTP > RD > DEN > GR > CAL > RS > U > CNL) and fracture in geologically complex geothermal carbonate reservoirs. Furthermore, comparing the XGBoost model with previous studies proved superior in training and testing. This study suggests that XGBoost with Q-learning-based optimized hyperparameters and data division is a suitable algorithm for identifying fractures using well-log data to explore complex geothermal systems in carbonate rocks. Graphical abstract: (Figure presented.)

The oil and gas industry relies on accurately predicting profitable clusters in subsurface formations for geophysical reservoir analysis. It is challenging to predict payable clusters in complicated geological settings like the Lower Indus Basin, Pakistan. In complex, high-dimensional heterogeneous geological settings, traditional statistical methods seldom provide correct results. Therefore, this paper introduces a robust unsupervised AI strategy designed to identify and classify profitable zones using self-organizing maps (SOM) and K-means clustering techniques. Results of SOM and K-means clustering provided the reservoir potentials of six depositional facies types (MBSD, DCSD, MBSMD, SSiCL, SMDFM, MBSh) based on cluster distributions. The depositional facies MBSD and DCSD exhibited high similarity and achieved a maximum effective porosity (PHIE) value of ≥ 15%, indicating good reservoir rock typing (RRT) features. The density-based spatial clustering of applications with noise (DBSCAN) showed minimum outliers through meta cluster attributes and confirmed the reliability of the generated cluster results. Shapley Additive Explanations (SHAP) model identified PHIE as the most significant parameter and was beneficial in identifying payable and non-payable clustering zones. Additionally, this strategy highlights the importance of unsupervised AI in managing profitable cluster distribution across various geological formations, going beyond simple reservoir characterization.

Natural barriers, encompassing stable geological formations that serve as the final bastion against radionuclide transport, are paramount in mitigating the long-term contamination risks associated with the nuclear waste disposal. Therefore, it is important to simulate and predict the processes and spatial-temporal distributions of radionuclide transport within these barriers. However, accurately predicting radionuclide transport on the field scale is challenging due to uncertainties associated with parameter scaling. This study develops an integrated evaluation framework that combines upscaled parameters, streamline transport models, and response surface techniques to systematically assess environmental risk metrics and parameter uncertainties across different scales. Initially, upscaling methods are established to estimate the prior interval of critical transport parameters at the field scale, and streamline models are derived by considering the radionuclides transport with a variety of physicochemical mechanisms and geological characterizations in natural barriers. To assess uncertainty ranges of the risk metrics related to upscaled parameters, uncertainty quantification is performed on the ground of 5000 Monte Carlo simulations. The results indicate that the upscaled dispersivity of fractured media (αLf) has a relatively high sensitivity ranking on release dose for all nuclides, and upscaled matrix sorption coefficient (Kd) of Pu-242 strongly affects breakthrough time and release dose of Pu-242. Facilitated by robust response surface with the lowest R2 of 0.89, it is shown that the release doses of Pu-242 and Pb-210 increase under conditions of low Kd and αLf, respectively. Furthermore, statistical analysis reveals that employing limited laboratory-scale parameters results in narrower confidence intervals for risk metrics, while upscaling methods better account for the highly heterogeneous properties of large-scale field conditions. The developed risk evaluation framework provides valuable insights for utilizing upscaled parameters and modeling radionuclide transport within natural barriers under various scenarios.

Coal spontaneous combustion (CSC) threatens the safety of the coal industry, with moisture content and particle size being pivotal factors. This study examines the heating dynamics and critical self-ignition temperatures (CSITs) of Baiyinhua lignite stockpiles through wire-mesh basket (WMB) tests at two scales. The CSC process in coal stockpiles unfolds in four stages. Notably, Stage II is notable for significant moisture evaporation between 43 and 84 °C, while Stage IV marks the onset of self-heating. Moisture evaporation absorbs heat, linearly prolonging the Stage II duration, which accounts for 0%–70 % of the total time. Conversely, larger particle sizes enhance pore seepage, effectively shortening the heating time. The time for d = 20 mm (particle size) coal sample to reach ambient temperature is roughly half that of 1.5 mm. CSITs of raw coal increase by 10–15 °C compared to the dry coal, and CSIT of coal samples with d = 50 mm has increased by 17.5 °C compared to 10 mm. Therefore, both an increase in particle size and moisture content increase the CSIT, thereby reducing the propensity for CSC. Frank-Kamenetskii theory and dimensionless analysis predict spontaneous combustion risks in field-scale coal stockpiles. This investigation contributes valuable insights to the estimation and prevention of CSC.

This study investigates the application of machine learning techniques—specifically convolutional neural networks, multilayer perceptrons and cascaded forward neural networks —to understand the wettability of the CO2/brine/rock system, a critical factor in carbon dioxide (CO2) capture, utilization, and storage in deep saline aquifers. Understanding wettability is essential for improving the efficacy of CO2 storage. The study incorporates variables such as salinity, mineral types, measurement methods, pressure, and temperature into the machine learning models. Using a dataset of 876 samples from existing literature, the proposed models were trained and optimized using the Adam optimizer, Levenberg-Marquardt algorithm, and particle swarm optimization respectively. The performance of these models was evaluated through plot analysis, statistical indicators, and the Taylor diagram, demonstrating a high level of accuracy compared to experimental data. The specifically convolutional neural networks model showed exceptional accuracy in predicting CO2 wettability in brine, with a root mean square error of 0.9612 and coefficient of determination value of 0.9982. The minimal presence of outliers in the specifically convolutional neural networks model further confirms its robustness. This research highlights the effectiveness of deep learning in modeling complex wettability behaviors in CO2-brine-mineral systems, offering substantial insights for enhancing carbon dioxide (CO2) capture, utilization, and storage strategies. The novelty of this work lies in its comprehensive integration of multiple variables and the use of advanced machine learning optimization techniques, going beyond previous efforts by achieving higher predictive accuracy and providing a more detailed understanding of wettability dynamics.

This study introduces machine learning (ML) algorithms to predict hydrogen (H2) thermodynamic properties for geological storage, focusing on its mixtures with natural gas, nitrogen (N2), and carbon dioxide (CO2). Employing 1167 data samples, this research utilizes multiple linear regression (MLR), random forest (RF), gradient boosting (GB), and decision tree (DT) models, enhanced by sensitivity analysis for feature engineering. GB model's superior accuracy and efficiency over conventional statistical linear regression methods. The finding reveals that the GB model achieves superior performance, with an R2 value of 0.9998 for thermal capacity and the lowest mean absolute error (MAE)/root mean square error (RMSE) across all H2 properties—density (1.51%, 2.06 kg/m3), viscosity (0.000078%, 0.000117 cp), thermal conductivity (0.000427%, 0.000849 W/m.K), and thermal capacity (8.79%, 26.06 J/g.K). Moreover, this ML-based approach not only demonstrates remarkable accuracy and efficiency but also suggests the potential of smart paradigms to further enhance the safety and transportability of underground hydrogen storage projects. Ultimately, this work will help the reservoir simulation at field scale to be more robust and reliable.

The utilization of CO2 flooding is a widely applied enhanced oil recovery (EOR) technique in mature onshore oil fields. As well as being able to increase oil production and recovery it offers the potential to provide long-term, geological storage for carbon in subsurface reservoirs, thereby contributing to the mitigation of carbon emissions originating from human activities. Substantial research efforts have provided some insight into the uncertainties associated with CO2-EOR projects, but further understanding and the development of more reliable methods are required to accurately predict the outcomes of the complex processes involved in CO2 reservoir flooding. In this study, four machine learning (ML) algorithms were developed to predict CO2 storage mass and cumulative oil production, using the CMG-GEM three-phase flow compositional reservoir simulator with nine reservoir input variables covering a range of uncertainties associated with oil zones. The Mahalanobis distance technique was applied for identifying and excluding outlier data points from the target variable distributions of training data groups. 520 and 439 data records were excluded from the CO2 storage quantity and cumulative oil production dataset, respectively, to generate more reliable predictions. Meticulous training and testing of the ML models revealed that, of the models evaluated, the LSSVM model generated the lowest prediction errors with the test dataset (RMSE = 0.7811 million mt for CO2 storage mass; RMSE = 10.1245 million barrels for cumulative oil production) demonstrating excellent generalization capabilities.

Rock mechanical properties, encompassing the static Young's modulus, bulk modulus, and shear modulus, influence strategic decision-making operations, such as drilling programs and well completion designs. When precisely calculated, these variables can significantly affect drilling cost-efficiencies and production economic returns. Even though these attributes are significant, the conventional estimation method is core-lab testing, which is expensive and limited to covering the entire reservoir strata. To address this challenge, we propose a new semi-analytical model with direct equations to estimate the elastic rock properties primarily anchored around porosity. In particular, we developed an analytical framework using generalized equations for bulk and shear modulus to calculate static rock mechanical properties as porosity-driven functions. We employed lab-tested core data from 55 core samples to validate the new model's efficiency. The model's innovative design allows computing all elastic properties solely based on porosity. Impressively, the results closely matched the lab measurements. The new semi-analytical model's equations yielded a good root mean square error (RMSE) in the calculated rock mechanical properties. Specifically, the new static Young's modulus equation has the lowest RMSE, 1.489, compared to 45.168 for the Edimann equation and 3.043 for the Zhang equation. The newly developed static shear modulus equation had an RMSE of 0.497, lower than Zhang's empirical equation's 1.262. However, the Zhang equation has an RMSE of 0.891 versus 2.029, slightly outperforming the new equation. According to the RMSE values, the discrepancy between the core sample-based measured, new semi-analytical model-calculated static modulus was caused by the fact that the core samples were collected from clean sand oil-bearing reservoir intervals. However, the derived equations used porosity ranges that covered shale and sand. In essence, the new semi-analytical model provides a fast and precise estimation of rock mechanical properties, easily solving drilling concerns. This research not only advances knowledge of drilling and geomechanics, but also makes strategies more efficient and cost-effective.

This study explores the development of predictive models for carbon dioxide (CO2) solubility in ionic liquids based on a compiled dataset of 10,116 experimentally measured data points involving four input variables: pressure (P), temperature (T), cation type, and anion type. The deep-learning (DL) predictive models evaluated are standalone and hybrid versions of convolutional neural network (CNN) and long short-term memory (LSTM) algorithms with cuckoo optimization algorithm (COA) and gradient-based optimization (GBO). The laboratory-measured data was separated into training and test categories, and each category was normalized separately to improve the performance of the deep learning algorithms. The Mahalanobis distance-based quantile method was utilized to identify any outliers in the training data. Once identified, the outlier data points were eliminated from the training dataset. The control parameters of the deep learning algorithms were optimized using COA to enhance their efficiency, and the algorithms were hybridized with optimization algorithms to further improve their performance. The resulting models were analyzed to assess their accuracy, degree of overfitting, and the importance of input features. The study found that using 80% of the data for training and 20% for testing results in more accurate and generalizable models. Using the outlier detection method on the training data led to 307 data points being eliminated as outliers. Developing CO2-solubility predictive model showed that, the CNN[sbnd]COA model had the lowest RMSE and highest R2 among the developed models, indicating high generalizability for data unseen by the trained model. The analysis revealed that using optimization algorithms increased the CO2-solubility prediction performance of DL algorithms and reduced overfitting. T and cation type were the most and least important input features, respectively. Simultaneous changes in cation and anion type on CO2-solubility predictions displayed no systematic pattern. For increases in T, CO2 solubility typically decreased, whereas for increases in P CO2 solubility always increased but at variable rates. The results of this study can be used to develop accurate and generalizable CO2-solubility predictive models for various applications.

The most crucial elements in the oil and gas sector are predicting subsurface lithofacies utilizing geophysical logs for reservoir characterization and sweet spot assessment procedures. Nevertheless, accurately predicting payable lithofacies in a complex heterogeneous geological setting, such as the lower goru formation, poses considerable difficulty because conventional methods fall short in delivering highly accurate outcomes. Hence, this research proposes an advanced cost and time-saving data intelligence strategy using multiple classifiers to predict lithofacies with maximum accuracy that will aid in sweet spot evaluation in oil and gas fields globally. Geophysical log data of five wells from a mature gas field were used. The targeted reservoir formation was classified into seven facies types. We evaluated the performance of seven different models: support vector machine (SVM), K-nearest neighbors (KNN), random forest (RF), decision tree (DTr), naive Bayes (NB), adaptive boosting (AB), and ensemble (an integrated SVM, KNN, RF, and DTr classifier). RF and ensemble classifiers predicted the lithofacies with accuracies of 97.5 and 97.3%, respectively. Their efficacy in lithofacies prediction with high accuracy renders them as valuable tools in the domain of sweet spot evaluation. The proposed digital intelligence strategy could help operators identify drilling sites based on in-depth reservoir characterizations.

Due to the dissimilarity among different producing layers, the influences of inter-layer interference on the production performance of a multi-layer gas reservoir are possible. However, systematic studies of inter-layer interference for tight gas reservoirs are really limited, especially for those reservoirs in the presence of water. In this work, five types of possible inter-layer interferences, including both absence and presence of water, are identified for commingled production of tight gas reservoirs. Subsequently, a series of reservoir-scale and pore-scale numerical simulations are conducted to quantify the degree of influence of each type of interference. Consistent field evidence from the Yan'an tight gas reservoir (Ordos Basin, China) is found to support the simulation results. Additionally, suggestions are proposed to mitigate the potential inter-layer interferences. The results indicate that, in the absence of water, commingled production is favorable in two situations: when there is a difference in physical properties and when there is a difference in the pressure system of each layer. For reservoirs with a multi-pressure system, the backflow phenomenon, which significantly influences the production performance, only occurs under extreme conditions (such as very low production rates or well shut-in periods). When water is introduced into the multi-layer system, inter-layer interference becomes nearly inevitable. Perforating both the gas-rich layer and water-rich layer for commingled production is not desirable, as it can trigger water invasion from the water-rich layer into the gas-rich layer. The gas-rich layer might also be interfered with by water from the neighboring unperforated water-rich layer, where the water might break the barrier (eg weak joint surface, cement in fractures) between the two layers and migrate into the gas-rich layer. Additionally, the gas-rich layer could possibly be interfered with by water that accumulates at the bottom of the wellbore due to gravitational differentiation during shut-in operations.

The failure of rocks in seasonal frozen areas under freeze–thaw cycles (FTCs) is a frequently problem in engineering construction, posing a huge threat to the stability of the engineering. In order to explore the mechanism of rock damage degradation. It is necessary to analyze the damage evolution process of rocks and establish an accurate FTCs rock damage constitutive model. Taking the granite in the seasonal frozen zone of Northeast China as the research object, the macro and meso parameters and microstructure of the FTCs granite were analyzed through indoor tests. A new method is proposed to define damage variables considering mesoscopic parameters to establish a mesoscopic damage constitutive model for rocks, further revealing the damage mechanism and failure law of rocks under freeze–thaw load coupling. The research results indicate that in the early stage of FTCs, 20 cycles contribute significantly to the damage and deterioration of rocks, accounting for about 50% of the 80 cycles. After 20 cycles, the degradation trend of various macroscopic and mesoscopic parameters is relatively slow. The residual strain of the rock accumulates as the number of FTCs increases, the brittleness of the rock weakens, and the plasticity strengthens. Based on the established new method of considering mesoscopic parameter damage variables, the viewpoint of introducing correction coefficients based on statistical constitutive models has been validated. After uniaxial compression test data verification, the model has shown good performance. This model expands the damage model of rock freeze–thaw compression coupling effect, providing reference for research on FTCs related to rocks.

This paper delves into the key factors governing contaminant retardation and the scaling of retardation factor for contaminant transport in naturally fractured media. It employs a geostatistical approach, utilizing hierarchical transition probability and covariance models, to characterize multimodal reactive mineral facies. Subsequently, scale-dependent models are developed to illustrate the transport behaviors of reactive contaminants under equilibrium sorption conditions and quantify the spatial and temporal variations of the effective retardation factor. These models incorporate various factors, including the structural characteristics of reactive minerals (the volume proportions and lengths of different facies types), fracture characteristics (the means and variances of fracture apertures), and the adsorption capacity of contaminants on minerals (sorption coefficients (ln Kd)). The results from these Lagrangian-based models are compared and validated against experimental data of solute transport in a single natural fracture. These findings underscore the remarkable agreement between the derived scale-dependent analytical expressions and the experimental results. Additionally, a global sensitivity analysis was performed using the Monte Carlo-based Sobol’ indices method to understand the influential factors governing contaminant retardation. The study reveals that the primary influential factor is the sorption capacity of contaminants on minerals, while fracture and reactive mineral structural characteristics play a secondary role. These insights provide significant information for understanding the reactive transport processes and the environmental impact of transport parameters in fractured media.

Hydrogen, as a clean and efficient energy source, is important in achieving zero-CO2 targets. This paper explores the potential of hydrogen geologic storage (HGS) in China for large-scale energy storage, crucial for stabilizing intermittent renewable energy sources and managing peak demand. Despite its promise, HGS faces challenges due to hydrogen's low density and viscosity, and its complex interactions with geological formations and microorganisms. This review offers a comprehensive overview of the current research status for HGS, with a particular focus on highlighting the main challenges confronting China. These difficulties and challenges primarily arise from complex geological conditions and the absence of fundamental parameters within potential reservoirs, including depleted oil/gas fields, salt caverns, and brine aquifers. Additionally, we have synthesized the current applications of machine learning (ML)as a potential solution. Key challenges are examined, such as the effects of operational parameters (e.g., cyclical injection-production and injection rates) on HGS efficiency, which can influence phenomena such as hydrogen fingering and caprock integrity. This review also looks into the insufficiently understood hydrogen-water-rock geochemical reactions under diverse temperatures and pressures, a gap that hampers the development of predictive numerical simulations and raises concerns about hydrogen leakage due to changes in porosity and permeability. Additionally, the paper addresses the limited knowledge about the metabolic mechanisms of subsurface microorganisms under extreme conditions, highlighting potential risks of hydrogen leakage and groundwater contamination. These microorganisms can metabolize hydrogen, producing gases such as CH4 and H2S, which may cause steel corrosion. Furthermore, the study assesses the distribution and prospects of three primary methods for pure hydrogen storage in China, considering the current state of hydrogen development and relevant government policies. The role of ML in advancing HGS is also discussed, offering insights into future research directions. This review not only scrutinizes the scientific challenges of HGS but also underscores its potential, guiding future simulation and practical engineering applications in this field.

Fluid flow and mass transport parameters are greatly impacted by the fracture surface's heterogeneous properties. By quantifying the spatial correlation patterns of fracture apertures, this study intends to establish an upscaling framework utilizing Lagrangian-based transport models to estimate dispersivities of naturally fractured granite cores. To do this, the surface morphology of the fragmented core sample was photographed using a 3D laser profile scanner with varied degrees of accuracy. Following a geostatistical analysis of the fracture aperture data, common covariance models were fitted to determine the integral scale of log fracture aperture. This information was used to estimate dispersivity using the Lagrangian-based transport models, which only required the collection of fractured geostatistical data. The study also assessed how the dispersivities are affected by fracture lengths and the measurement accuracy of aperture data in this model. The findings showed that the fracture aperture spatial bivariate correlation structures follow the exponential covariance model. Additionally, while the variance of log fracture apertures is significantly influenced by both the data resolutions and the core samples themselves, the integral scale of log fracture apertures primarily depends on the length of the fracture. In terms of the fracture length, the upscaled dispersivities range from 2.35% to 7.21%. The development of upscaling techniques for estimating dispersivities in fractured and heterogenous granitic rocks and scaling them up to the field size for the accurate prediction of solute transport mechanisms can be guided by these findings.

Hydrogen offers significant potential as a sustainable energy source. However, its storage and transportation pose challenges due to its volatility and low density. Subsurface geological formations, such as shale, sandstone, and coal, have been investigated as potential storage sites for hydrogen. Precise estimation of hydrogen adsorption in these formations necessitates a thorough understanding of kerogens, organic-rich sedimentary rocks prevalent in unconventional formations. In this study, we introduce a novel approach employing Extreme Gradient Boosting (XGBoost), Random forest (RF), Light gradient Boosting (LGBM), and Natural gradient boosting (NGBM) algorithms to intelligently predict hydrogen adsorption in various kerogen types. Among these, the NGBM algorithm, utilizing three input features (temperature, pressure, and kerogen types), demonstrated the highest predictive accuracy, achieving an R2 value of 0.989, RMSE of 0.062, and MAE of 0.037 for all data samples. SHAP diagram analysis identified kerogen types as the most influential parameter within the NGBM model. The presented approach has implications beyond energy storage, highlighting the significance of advanced technologies in addressing complex energy challenges. Precise hydrogen adsorption estimation in subsurface formations is vital for developing sustainable energy solutions, and our approach has the potential to expedite progress in this field. Interdisciplinary collaboration among geologists, chemists, and data scientists is crucial for devising innovative solutions for sustainable energy.

Hydrogen is a clean and sustainable renewable energy source with significant potential for use in energy storage applications because of its high energy density. In particular, underground hydrogen storage via the dissolution of hydrogen gas in an aqueous solution has been identified as a promising strategy to address the difficulties associated with large-scale energy storage. However, this process requires the accurate prediction of the solubility of hydrogen in aqueous solutions, which is affected by a range of factors, including temperature, pressure, and the presence of solutes. The present study thus aimed to effectively predict the solubility of hydrogen in aqueous solutions that vary in their salinity by employing a machine learning approach. Four machine learning models were developed and tested: adaptive gradient boosting (Adaboost), gradient boosting, random forest, and extreme gradient boosting. The performance of each model was quantified in terms of their coefficient of determination (R2), root mean square error (RMSE), and mean absolute error (MAE). The Adaboost algorithm exhibited superior performance across all metrics, with an R2 of 0.994, MAE of 0.006, and RMSE of 0.018. A Williams plot detected only 18 outliers in the Adaboost predictions from a total of 255 data points. These results indicate that machine learning techniques have the potential to serve as a valuable tool in the prediction of hydrogen solubility in aqueous solutions for underground hydrogen storage, facilitating the development of smart, cost-effective, and safe hydrogen storage technologies.

For complex and multi-layered clastic oil reservoir formations, modeling lithofacies and petrophysical parameters is essential for reservoir characterization, history matching, and uncertainty quantification. This study introduces a real oilfield case study that conducted high-resolution geostatistical modeling of 3D lithofacies and petrophysical properties for rapid and reliable history matching of the Luhais oil reservoir in southern Iraq. For capturing the reservoir's tidal depositional setting using data collected from 47 wells, the lithofacies distribution (sand, shaly sand, and shale) of a 3D geomodel was constructed using sequential indicator simulation (SISIM). Based on the lithofacies modeling results, 50 sets of porosity and permeability distributions were generated using sequential Gaussian simulation (SGSIM) to provide insight into the spatial geological uncertainty and stochastic history matching. For each rock type, distinct variograms were created in the 0° azimuth direction, representing the shoreface line. The standard deviation between every pair of spatial realizations justified the number of variograms employed. An upscaled version of the geomodel, incorporating the lithofacies, permeability, and porosity, was used to construct a reservoir-flow model capable of providing rapid, accurate, and reliable production history matching, including well and field production rates.

Immiscible fluid-fluid displacement dynamics is a crucial element to understanding and engineering many subsurface flow applications, including enhanced oil recovery and carbon dioxide geological sequestration. Although there are several interfacial properties that govern such a displacement dynamic, the wettability has been considered a dominant factor. Owing to its complex coaffinity among the three phases (i.e., solid-fluid-fluid) and difficulty to be characterized accurately and efficiently, the wettability (defined as the contact angle: θ) determination is of interest in the current study with aim toward machine learning (ML) approach. In the current research, four experimental packages of fluid displacement at 1D capillary scale served as data sets for ML examination on the θ predictability. Via digital image processing, fluid traveling length at a given time was extracted, and the theoretical θ was calculated as ground truth for the modeling, with input features being fluid traveling length, displacing velocity, and the interfacial tension. Random forest (RF) and multilayer perception (MLP) were selected for the modeling due to their appropriate characteristics to the investigated data (being nonlinear relation). The prediction results showed that RF apparently outperformed MLP on the θ prediction, reflecting its best ability to manage missing values and outliers. Although more input features analyzed (from two to three features) did yield a better prediction, the best model remains RF. Sensitivity on the key parameters of displacing velocity and the interfacial tension was also analyzed, where the results confirm the model prediction agreement with theories. The study demonstrated how ML model can be an alternative tool to elucidate the fluid displacement in subsurface, with additional potential for autonomously improving the deep underground flows, converging a new concept of “augmented” artificial intelligence.

Water is ubiquitous within organic-rich shale in cases of connate water occurrence and during hydraulic fracturing treatment. Understanding multiphase transport behaviors in organic nanopores is crucial for the efficient development of shale gas reservoirs. However, current studies have predominantly focused on single-phase or two-phase transport behaviors in ideal graphite nanopores, leaving the understanding of multiphase transport processes within realistic kerogen-based nanopores limited. In this study, we conducted molecular dynamic simulations to investigate shale gas transport behaviors through organic nanopores constructed with realistic kerogen. The results reveal that, due to the complex composition in the chemistry and physics of kerogen macromolecules, gas transport through kerogen nanopores manifests parabolic-shaped velocity distributions with a negligible slip length at the walls, in contrast to the observations of fast mass transport in previous studies using smooth carbon-based skeleton nanopores. In water-saturated nanopores, H2O molecules tend to aggregate at the walls, forming water clusters, and eventually, a water pillar across the pore can be observed. As a result, a water blockage is formed, while the water film or water bridge dominates in some inorganic minerals. The presence of H2O molecules has a dramatic impact on shale gas transport capacity. On this basis, an analytical model was proposed to quantitatively characterize shale gas transport behaviors under different water saturations. The results demonstrate that the traditional continuous model with no-slip assumption remains applicable because of the rough kerogen surface and hindrance of water clusters, advancing the understanding of multiphase transport behaviors in shale nanopores and exploitation of shale gas reservoirs.

Crucial for carbon capture, utilization, and storage (CCUS) initiatives and diverse industries, heat transfer underscores the need for a precise assessment of carbon dioxide (CO2) and nitrogen (N2) viscosities in gaseous blends across various temperatures. This research pioneers an intelligent model by enhancing the dendritic neural regression (DNR) framework, employing the Seagull Optimization Algorithm with Marine Predator Algorithm (SOAMPA) for optimal predictions. Leveraging recent advancements in metaheuristic optimization techniques, the study reveals the superior performance of the novel SOAMPA approach in predictive accuracy, marking a significant breakthrough in predicting CO2-N2 mixture viscosities with implications for advancing CCUS projects and diverse industries. The optimized DNR model, empowered by the modified SOAMPA optimization technique, contributes to estimating the viscosity of N2-CO2 mixture gases. Utilizing inputs like pressure, temperature, mole fraction of N2, and model fraction of CO2, the models are trained and tested on a dataset comprising over 3030 data samples from public literature. Key contributions encompass proposing an optimized DNR approach, introducing the modified SOAMPA technique, and demonstrating its superiority over established optimization methods in conjunction with the traditional DNR model for predicting viscosity based on real experimental datasets.

Enhanced oil recovery (EOR) using CO2 injection is promising with economic and environmental benefits as an active climate-change mitigation approach. Nevertheless, the low sweep efficiency of CO2 injection remains a challenge. CO2-foam injection has been proposed as a remedy, but its laboratory screening for specific reservoirs is costly and time-consuming. In this study, machine-learning models are employed to predict oil recovery factor (ORF) during CO2-foam flooding cost-effectively and accurately. Four models, including general regression neural network (GRNN), cascade forward neural network with Levenberg–Marquardt optimization (CFNN-LM), cascade forward neural network with Bayesian regularization (CFNN-BR), and extreme gradient boosting (XGBoost), are evaluated based on experimental data from previous studies. Results demonstrate that the GRNN model outperforms the others, with an overall mean absolute error of 0.059 and an R2 of 0.9999. The GRNN model's applicability domain is verified using a Williams plot, and an uncertainty analysis for CO2-foam flooding projects is conducted. The novelty of this study lies in developing a machine-learning-based approach that provides an accurate and cost-effective prediction of ORF in CO2-foam experiments. This approach has the potential to significantly reduce screening costs and time required for CO2-foam injection, making it a more viable carbon utilization and EOR strategy.

This study explores the use of machine learning algorithms to predict hydrogen wettability in underground storage sites. The motivation for this research is the need to find a safe and efficient way to store hydrogen, which has become increasingly important as the world shifts toward using more clean energy sources. The study used four different machine learning algorithms including XGBoost, RF, LGRB, and Adaboost_DT to analyze 513 data points collected from previous literature. The input features included pressure, temperature, salinity, and substrate types, while the target output was hydrogen wettability. This study found that the XGBoost algorithm with four inputs produced the most precise predictions with the highest R2 value of 0.941, the lowest RMSE value of 4.455, and MAE of 2.861 for the overall databank. Based on SHAP values, the substrates are the most impactful variables of the XGBoost model. The predicted hydrogen column height was also evaluated for a specific storage site in Australia's basalt formation. At 308 K, the predicted hydrogen column height decreased from 1991 to 1319 m, while at 343 K, it decreased from 1510 to 784 m. These predictions were compared to the real column height that fell from 1660 to 928 m in the same pressure range. Overall, the study's findings provide a valuable guide for predicting wettability and evaluating hydrogen column height in specific storage sites. The use of machine learning algorithms can significantly reduce the time, cost, and unpredictability associated with traditional methods of assessing hydrogen wettability in underground storage sites.

Emissions of carbon dioxide (CO2) are a major source of atmospheric pollution contributing to global warming. Carbon geological sequestration (CGS) in saline aquifers offers a feasible solution to reduce the atmospheric buildup of CO2. The direct determination of the trapping efficiency of CO2 in potential storage formations requires extensive, time-consuming simulations. Machine-learning (ML) models offer a complementary means of determining trapping indexes, thereby reducing the number of simulations required. However, ML models have to date found it difficult to accurately predict two specific reservoir CO2 indexes: residual-trapping index (RTI) and solubility-trapping index (STI). Hybridizing ML models with optimizers (HML) demonstrate better RTI and STI prediction performance by selecting the ML model's hyperparameters more precisely. This study develops and evaluates six HML models, combining a least-squares-support-vector machine (LSSVM) and a radial-basis-function neural network (RBFNN) with three effective optimizer algorithms: genetic (GA), cuckoo optimization (COA), and particle-swarm optimization (PSO). 6810 geological-formation simulation records for RTI and STI were compiled from published studies and evaluated with the six HML models. Error and score analysis reveal that the HML models outperform standalone ML models in predicting RTI and STI for this dataset, with the LSSVM-COA model achieving the lowest root mean squared errors of 0.00421 and 0.00067 for RTI and STI, respectively. Sensitivity analysis identifies residual gas saturation and permeability as the most influential input variables on STI and RTI predictions. The high RTI and STI prediction accuracy achieved by the HML models offers to reduce the uncertainties associated with CGS projects substantially.

Ongoing anthropogenic carbon dioxide (CO2) emissions to the atmosphere cause severe air pollution that leads to complex changes in the climate, which pose threats to human life and ecosystems more generally. Geological CO2 storage (GCS) offers a promising solution to overcome this critical environmental issue by removing some of the CO2 emissions. The performance of GCS projects depends directly on the solubility and residual trapping efficiency of CO2 in a saline aquifer. This study models the solubility trapping index (STI) and residual trapping index (RTI) of CO2 in saline aquifers by applying four robust machine learning (ML) and deep learning (DL) algorithms. Extreme learning machine (ELM), least square support vector machine (LSSVM), general regression neural network (GRNN), and convolutional neural network (CNN) are applied to 6811 compiled simulation records from published studies to provide accurate STI and RTI predictions. Employing different statistical error metrics coupled with supplementary evaluations, involving score and robustness analyses, the prediction accuracy of the models proposed is comparatively assessed. The findings of the study revealed that the LSSVM model delivers the lowest RMSE values: 0.0043 (STI) and 0.0105 (RTI) with few outlying predictions. Presenting the highest STI and RTI prediction scores the LSSVM is distinguished as the most credible model among all the four models studied. The models consider eight input variables, of which the time elapsed and injection rate displays the strongest correlations with STI and RTI, respectively. The results suggest that the proposed LSSVM model is best suited for monitoring CO2 sequestration efficiency from the data variables considered. Applying such models avoids time-consuming complex simulations and offers the potential to generate fast and reliable assessments of GCS project feasibility. Accurate modeling of CO2 storage trapping indexes guarantees successful geological CO2 storage operation, which is, in fact, the cornerstone of properly controlling and managing environmentally polluting gases.

Hydrogen (H2) absorption percentage by porous carbon media (PCM) is important for identifying efficient H2 storage media. PCM with H2-uptakes of greater than 5 wt% are urgently required to improve the performance of H2 fuel tanks for use in fuel-cell-powered transportation vehicles. Machine-learning (ML) methods can provide effective tools for predicting PCM H2-uptakes from influential variables determined by experiments performed on a wide range of PCM. This study evaluates the PCM-H2-uptake prediction performance of four well-established ML models: generalized-regression neural network (GRNN), Least-squares-support-vector machine (LSSVM), adaptive-neuro-fuzzy-inference system (ANFIS), and extreme-learning machine (ELM). A 2072-record database, compiled from literature, comprising eleven independent variables and PCM H2-uptake (dependent variable covering a range of 0 to 8.38 wt%) was evaluated by the four ML models. Each model was trained and validated using 10-fold cross-validation. The LSSVM generates the best PCM-H2-uptake prediction performance when applied to an independent testing subset of data records, achieving a root mean squared error of just 0.2407 wt%. Feature importance sensitivity analysis identifies pressure as the most influential of the independent variable considered. Leverage analysis identified that 96.53% of the data records of the compiled database, when predicted by the LSSVM model, resided within the applicable domain with only seventy-two data records considered as suspected outliers. These results indicate that the LSSVM model developed is highly generalizable for the purpose of predicting PCM H2-uptake from the influential variables.

The utilization of carbon capture utilization and storage (CCUS) in unconventional formations is a promising way for improving hydrocarbon production and combating climate change. Shale wettability plays a crucial factor for successful CCUS projects. In this study, multiple machine learning (ML) techniques, including multilayer perceptron (MLP) and radial basis function neural networks (RBFNN), were used to evaluate shale wettability based on five key features, including formation pressure, temperature, salinity, total organic carbon (TOC), and theta zero. The data were collected from 229 datasets of contact angle in three states of shale/oil/brine, shale/CO2/brine, and shale/CH4/brine systems. Five algorithms were used to tune MLP, while three optimization algorithms were used to optimize the RBFNN computing framework. The results indicate that the RBFNN-MVO model achieved the best predictive accuracy, with a root mean square error (RMSE) value of 0.113 and an R2 of 0.999993. The sensitivity analysis showed that theta zero, TOC, pressure, temperature, and salinity were the most sensitive features. This research demonstrates the effectiveness of RBFNN-MVO model in evaluating shale wettability for CCUS initiatives and cleaner production.

Physical heterogeneities are prevalent features of fracture systems and significantly impact transport processes in aquifers across different spatiotemporal scales. Upscaling solute transport parameter is an effective way of quantifying parameter variability in heterogeneous aquifers including fractured media. This paper develops conceptual models for upscaling conservative transport parameters in fracture media. The focus is on upscaling dispersivity. Lagrangian-based transport model (LBTM) for dispersivity upscaling are derived for the solute transport in two-dimensional fractures surrounded by an impermeable matrix. The LBTM is validated against the random walk particle tracking (RWPT) model, which enables highly efficient and accurate predictions of conservative solute transport. The results show that the derived scale-dependent analytical expressions are in excellent agreement with RWPT model results. In addition, LBTM results are also compared to experimental results from the observed breakthrough curve of a conservative solute transport through a single natural fracture within a granite core. Comparing results from the LBTM and transport experiment shows that LBTM based estimated dispersivity is 10.55% higher than the measured value. Errors introduced by the experiments, the conceptual assumptions in deriving models, and the heterogeneities of fracture apertures not fully sampled by measuring instruments are main factor for such discrepancy. The sensitivity analysis indicates that the longitudinal and transverse dispersivities are positively related to the integral scale and the variance of the log-fracture aperture. The longitudinal dispersivity is strongly contolled by the variance of the log-fracture aperture. The LBTM may be useful for directly predicting solute transports, requiring only the acquisition of fractured geostatistical data. This work provides a better understanding of transport processes in fractured media which ultimately control water quality across scales.

Traditional mine water inflow prediction is characterized by a high degree of uncertainty in model parameters and complex mechanisms involved in the water inflow process. Data-driven models play a key role in predicting inflow mechanisms without considering physical changes. However, the existing models are limited by nonlinearity and non-stationarity. Thus, the principal objective of this study was to propose two robust models, the DIFF-TCN model and the DIFF-LSTM model, for predicting the average water inflow per day. The models consist of three methods, namely Difference Method (DIFF), Temporal Convolutional Neural Network (TCN), and Long Short-Term Memory Neural Network (LSTM). When applied to the Tingnan Coal Mine, Shanxi Province, China, the DIFF-TCN performs better in predicting the average daily water inflow, the model has a MAE of 5.88 m3/h, RMSE of 6.85 m3/h and R2 of 0.96 in the test stage of the water inflow event. Comparison with the other deep learning models (with similar complex structures) and traditional time series model shows the superiority of our proposed DIFF-TCN model. The SHAP value is used to explain the contribution of each model input to the predicted values, and it indicates that the historical time of water inflow data are the most important input, and the advance distance and the groundwater level data also contribute to the model predictions, but groundwater level data for some periods in the past may have a detrimental effect on the model. The findings of this study can provide better understanding about potential of robust deep learning models for smart hydrological forecasting, and they can also provide technical guidance for mining safety production and protection of water resources and water environment around the mining area.

The selection of desirable synthesis procedures to achieve the idea of physiochemical and capacitive properties of activated carbons (ACs) can be carried out by the multi-criteria decision-making (MCDM) technique. The technique of ordering preference by similarity to an ideal solution (TOPSIS) is a well-known MCDM method with a superior selection of ideal materials. The present work elaborates a framework by establishing the TOPSIS method to obtain ideal synthesizing features, materials, and electrochemical measurements of AC electrode. The 21 multi-alternatives are considered from (i) temperature/heating rate of carbonization, chemical activating/doping, and post-purifying procedures of ACs; and (ii) the ratios of components, electrolytes, and potential window from AC-derived electrodes. 12 creteria are obtained from categories include microstructure properties, heteroatoms content of ACs, and gravimetric capacitance of AC electrodes used in electric double-layer capacitors (EDLCs). TOPSIS proposed scores for activating agent ratio, activation temperature, and heating rate of 0.68, 0.65, and 0.57 on the physiochemical criteria of AC, respectively. Also, the electrolyte concentration, type, and ratio of activating agents were ranked with 1, 082, and 062 scores on electrode capacitance, respectively. Moreover, TOPSIS exemplified two-step hydrothermal-assisted synthesis, artificial doping, and 6M HCl for purifying to achieve ACs with ideal physiochemical and capacitive performance. The MCDM technique proved that it can be applied to select the ideal material and process in AC-electrode fabrications with less time, financial, and environmental costs.

For the successful discovery and development of tight sand gas reserves, it is necessary to locate sand with certain features. These features must largely include a significant accumulation of hydrocarbons, rock physics models, and mechanical properties. However, the effective representation of such reservoir properties using applicable parameters is challenging due to the complicated heterogeneous structural characteristics of hydrocarbon sand. Rock physics modeling of sandstone reservoirs from the Lower Goru Basin gas fields represents the link between reservoir parameters and seismic properties. Rock physics diagnostic models have been utilized to describe the reservoir sands of two wells inside this Middle Indus Basin, including contact cement, constant cement, and friable sand. The results showed that sorting the grain and coating cement on the grain’s surface both affected the cementation process. According to the models, the cementation levels in the reservoir sands of the two wells ranged from 2% to more than 6%. The rock physics models established in the study would improve the understanding of characteristics for the relatively high Vp/Vs unconsolidated reservoir sands under study. Integrating rock physics models would improve the prediction of reservoir properties from the elastic properties estimated from seismic data. The velocity–porosity and elastic moduli-porosity patterns for the reservoir zones of the two wells are distinct. To generate a rock physics template (RPT) for the Lower Goru sand from the Early Cretaceous period, an approach based on fluid replacement modeling has been chosen. The ratio of P-wave velocity to S-wave velocity (Vp/Vs) and the P-impedance template can detect cap shale, brine sand, and gas-saturated sand with varying water saturation and porosity from wells in the Rehmat and Miano gas fields, both of which have the same shallow marine depositional characteristics. Conventional neutron-density cross-plot analysis matches up quite well with this RPT’s expected detection of water and gas sands.