阿尔及利亚国家石油公司

Mineral scale deposition remains a major flow-assurance constraint in oil and gas operations, especially in water-flooding and produced-water reinjection, where mixing between incompatible brines promotes super-saturation and precipitation of poorly soluble salts. This work introduces a novel extension of traditional methods used for modeling chemical inhibition and the predictive evaluation of oilfield scale-inhibitor molecules. A systematically optimized Two-Dimensional Quantitative Structure-Activity Relationship Model based on the k-Nearest Neighbors algorithm 2D-QSAR-KNN model was developed to quantitatively link molecular constitution of phosphonate inhibitors, brine chemistry, and operating factors with inhibition efficiency IE %. The optimized model achieved strong accuracy and generalization R2train = 0.9182, R2test = 0.9306, and R2global = 0.9208 with low prediction errors RMSEtrain = 4.7888%, RMSEtest = 4.5485%, and RMSEglobal = 4.7421%. Median absolute errors remained minimal for the train set = 0.80%, and test set = 1.63%, and model stability was confirmed by high correlation with experimental IE % r = 0.94 and R2train/R2test approximate to 0.99, showing no sign of overfitting. Additionally, an inverse-2D-QSAR framework was applied to identify the optimal molecular descriptor profile expected to maximize inhibitory performance within normalized bounds, providing rational rules for next-generation inhibitor design. The findings highlight the practical value of QSAR-inspired AI modeling to accelerate molecule screening and dosage exploration prior to laboratory validation, supporting more cost-effective, interpretable, and environmentally aware sulfate-scale inhibition strategies under high-salinity reservoir conditions.

Featured Application New 2D-VMD-based seismic attributes have been suggested. They provide significant advantages over traditional approaches and that combining complementary methods can further improve seismic interpretation outcomes.Abstract Seismic attributes are widely used to enhance the interpretation of structural, stratigraphic, and lithologic features in subsurface data. Their effectiveness, however, can be limited by noise, resolution constraints, and processing artifacts. This study suggests new seismic attributes computed using 2D-Variational Mode Decomposition (2D-VMD), which are specifically Mode-Weighted Spectral Discontinuity (MWSD) (in Module and Phase modes), VMD-Directionality Coherence (VDC), Instantaneous Frequency Concentration (IFC-VMD), and Instantaneous Bandwidth Dispersion (IBD-VMD). The proposed 2D-VMD-based attributes are compared with seven key conventional seismic attributes: dip, azimuth, chaos, coherence (semblance), curvature (mean curvature), instantaneous frequency, and instantaneous bandwidth (Hilbert transform). Through applications on simulated and real seismic data, each method is compared in terms of its ability to enhance attribute stability, resolution, and interpretability while mitigating limitations such as noise sensitivity and loss of detail. Results indicate that MWSD (Module) is optimal for amplitude stability, MWSD (Phase) for phase-sensitive applications, VDC for high-resolution structural delineation, IFC-VMD for complex geological settings, and IBD-VMD for abrupt feature changes. The findings demonstrate that these new 2D-VMD-based techniques provide significant advantages over traditional approaches and that combining complementary methods can further improve seismic interpretation outcomes.

Natural gas occupies a central position in today's global energy system, acting as a bridge fuel in the transition toward low-carbon systems. Yet, the solidification of carbon dioxide (CO2) during processing and transportation creates significant operational and safety concerns. Accurate prediction of the CO2 frosting temperature (Tf) in natural gas mixtures is therefore essential for ensuring flow assurance and efficient process design. Traditional estimation methods remain constrained: experimental techniques, though precise, are costly and time-intensive, while thermodynamic and empirical approaches often suffer from limited applicability and moderate reliability under diverse conditions. To overcome these limitations, this study proposes an integrated and explainable data-driven framework that employs advanced machine learning paradigms for high-fidelity Tf prediction under complex operating conditions. Three models, namely Categorical Boosting (CatBoost), Tabular Prior-data Fitted Network (TabPFN), and Least Squares Support Vector Machine (LSSVM), were implemented and benchmarked using a comprehensive experimental database to accurately predicting CO2 frosting temperature under extensive operational conditions. Among the applied predictive schemes, the TabPFN model delivered superior accuracy, achieving a mean absolute percentage error of 0.13% and a coefficient of determination of 0.9992. Model interpretability was ensured using Shapley Additive Explanations (SHAP), providing transparent insight into feature influence and reinforcing the physical credibility of the predictions. Statistical reliability was further evaluated using the leverage approach, confirming model robustness. Overall, the framework demonstrates the rigorous integration of high-accuracy prediction, explainable artificial intelligence, and statistical reliability assessment within a unified platform, offering a practical and reliable tool for industrial applications.

Wax precipitation can present a persistent challenge in natural gas systems, leading to blockages, reduced flow efficiency, and costly interruptions across production, transportation, and processing units. Understanding and predicting the solubility of paraffin waxes in supercritical gases, such as CO2 and ethane, is crucial to managing these flow assurance issues. However, conventional experimental techniques, while accurate, are resourceintensive and time-consuming. Similarly, traditional thermodynamic models often struggle with generalization when applied to complex multi-variable systems, particularly under varying operational conditions. In response to these limitations, this study explores the potential of advanced data-driven techniques to model wax solubility more efficiently and accurately. Three intelligent algorithms, including Extreme Trees (ET), Multi-Layer Perceptron optimized with Levenberg-Marquardt Algorithm (MLP-LMA) and Bayesian Regularization (MLP-BR), were trained and tested on a comprehensive experimental dataset that includes pressure, temperature, and critical temperatures of both the gas phase and solid wax compounds. Among these, the MLP-LMA model emerged as the top performer, achieving an outstanding prediction accuracy with an RMSE of 53.7729 and an R2 of 0.9996. Further validation through trend analysis and XAI techniques (SHapley Additive exPlanations) confirmed the model's ability to capture underlying physical patterns and variable importance. Leverage diagnostics also indicated strong statistical reliability, with only 1.98 % of observations classified as potential outliers. Beyond predictive accuracy, the model holds significant promise for real-world deployment. Its integration into industrial flow assurance systems could enable rapid solubility estimation, support operational decisions, reduce downtime, and optimize the design of wax management strategies in CO2 and ethane-rich systems.