The transitional shale system of the Longtan Formation (LTF) is widely distributed in the Sichuan Basin. However, the lithofacies of the LTF shale system exhibit vertical variations, with frequent interbedding of blocks, and shale–sand–coal sequences, which makes identifying “sweet spots” a challenging task. To address this issue, lithofacies associations were investigated based on well log analysis from 30 wells, and experimental data from 19 well samples, including X-ray diffraction, total organic carbon (TOC), pore structure characterization, and methane isothermal adsorption tests. Four lithofacies associations were classified: carbon–shale interbedding (I-1), shale(carbon)–coal interbedding (I-2), shale–sand interbedding (II), and shale–sand–coal assemblage (III). A favorable lithofacies association index (Com) was developed, providing a quantitative method for identifying favorable lithofacies. The results indicate that among the four lithofacies associations, I-2 is the most favorable lithofacies association. The Com index threshold for favorable lithofacies is defined as 0.6, and for the most favorable lithofacies, it is 0.7. Overall, favorable lithofacies are primarily distributed in the Suining-Dazu and Lujiao areas.

Granular samples are often used to characterize the pore structure of shale. To systematically analyze the influence of particle size on pore characteristics, case studies were performed on two groups of organic-rich deep shale samples. Multiple methods, including small-angle neutron scattering (SANS), low-pressure nitrogen gas adsorption (LP-N₂GA), low-pressure carbon dioxide gas adsorption (LP-CO₂GA), and XRD analysis, were adopted to investigate how the crushing process would affect pore structure parameters and the fractal features of deep shale samples. The research indicates that with the decrease in particle size, the measurements from nitrogen adsorption and SANS experiments significantly increase, with relative effects reaching 95.09% and 51.27%, respectively. However, the impact on carbon dioxide adsorption measurements is minor, with a maximum of only 8.97%. This suggests that the comminution process primarily alters the macropore structure, with limited influence on the micropores. Since micropores contribute the majority of the specific surface area in deep shale, the effect of particle size variation on the specific surface area is negligible, averaging only 16.52%. Shales exhibit dual-fractal characteristics. The distribution range of the mass fractal dimension of the experimental samples is 2.658–2.961, which increases as the particle size decreases. The distribution range of the surface fractal dimension is 2.777–2.834, which decreases with the decrease in particle size.

To address the limitations of traditional methods in modeling complex nonlinear relationships in horizontal in situ stress prediction for shale reservoirs, this study proposes an integrated framework that combines well logging interpretation with machine learning to accurately predict horizontal in situ stress in shale reservoirs. Based on the logging data from five wells in the Luzhou Block of the Sichuan Basin (16,000 samples), Recursive Feature Elimination (RF-RFE) was used to identify nine key factors, including Stoneley wave slowness and caliper, from 30 feature parameters. Bayesian optimization was employed to fine-tune the hyperparameters of the XGBoost model globally. Results indicate that the XGBoost model performs optimally in predicting maximum horizontal principal stress (SHmax) and minimum horizontal principal stress (SHmin). It achieves R² values of 0.978 and 0.959, respectively, on the test set. The error metrics (MAE, MSE, RMSE) of the XGBoost model are significantly lower than those of SVM and Random Forest, demonstrating its precise capture of the nonlinear relationships between logging parameters and in situ stress. This framework enhances the model’s adaptability to complex geological conditions through multi-well data training and eliminating redundant features, providing a reliable tool for hydraulic fracturing design and wellbore stability assessment in shale gas development.

Shale gas is a critical energy resource, and estimating its ultimate recoverable reserves (EUR) is a key indicator for evaluating the development potential and effectiveness of gas wells. To address the challenges in accurately predicting shale gas EUR, this study analyzed production data from 200 wells in the CN block. Sixteen potential factors influencing EUR were considered, and key geological, engineering, and production factors were identified using Spearman correlation analysis and mutual information methods to exclude highly linearly correlated variables. An attention mechanism was introduced to weight input features prior to model training, enhancing the interpretability of feature contributions. The hyperparameters of the model were optimized using the Rabbit Optimization Algorithm (ROA), and 10-fold cross-validation was employed to improve the stability and reliability of model evaluation, mitigating overfitting and bias. The performance of four machine learning models was compared, and the optimal model was selected. The results indicated that the ROA-CatBoost-AM model exhibited superior performance in both fitting accuracy and prediction effectiveness. This model was subsequently applied for EUR prediction and for identifying the primary factors controlling productivity, providing effective guidance for development practices. The dominant factors and production forecasts determined by the model offer valuable references for optimizing block development strategies.