2026· Management Strategies and Engineering Sciences· Vol 8, pp. 1-9· 0 citations
Abstract
This study aims to develop and validate machine learning models to predict completion fluid stability and quantitatively assess its impact on well productivity in challenging reservoir environments. The study employed a quantitative, applied research design based on historical completion and production data from onshore oil and gas reservoirs in Iran. The dataset integrated reservoir properties, completion fluid physicochemical characteristics, operational parameters, and post-completion productivity indicators. After data preprocessing, feature engineering, and normalization, multiple supervised machine learning algorithms—including linear, kernel-based, and ensemble models—were trained and evaluated. Robust cross-validation and hyperparameter optimization strategies were applied to ensure model generalizability and prevent overfitting. Model interpretability was addressed through feature importance analysis and sensitivity evaluation. Inferential results indicated that nonlinear ensemble models significantly outperformed linear approaches in predicting completion fluid stability, achieving high explanatory power and low prediction error. Reservoir temperature and formation water salinity emerged as the most influential predictors, followed by fluid thermal stability limits and filtration loss characteristics. Predicted stability classes exhibited statistically meaningful differences in productivity outcomes, with high-stability completions associated with substantially higher normalized productivity indices and initial production rates. The relationship between predicted stability and productivity was nonlinear, revealing a threshold beyond which incremental stability improvements yielded diminishing productivity gains. The findings confirm that machine learning provides a robust and interpretable framework for predicting completion fluid stability and its productivity implications under complex reservoir conditions. By linking stability predictions to measurable production outcomes, the proposed approach offers a practical decision-support tool for optimizing completion fluid design, reducing formation damage risk, and enhancing economic performance in challenging reservoirs.
Relevance. The oil and gas industry remains strategically important to the global economy, and the efficiency of field development is directly determined by the accuracy of reservoir potential assessment and geological system behavior prediction. However, reliable quantitative assessment of reservoir productivity is hampered by high geological heterogeneity, limited and diverse geological and production data, and the difficulty of integrating static and dynamic parameters. With the industry digitalization, machine learning methods are emerging as a promising tool capable of complementing or partially replacing classic hydrodynamic simulators by identifying complex nonlinear relationships between input geological characteristics and output development indicators. Aim. To critically systematize global experience in applying machine learning methods to assess reservoir potential and forecast oil and gas field development indicators. Methods. Analysis of scientific publications devoted to the application of machine learning methods, including neural networks, ensemble algorithms, probabilistic and hybrid models, to forecasting flow rates, cumulative production, and pressure. The characteristics of the input data used, approaches to model training, and metrics for assessing forecast quality are discussed. Results and conclusions. Machine learning methods have been shown to significantly accelerate reservoir potential assessment while maintaining accuracy comparable to hydrodynamic models. They effectively identify complex relationships between geological characteristics and dynamic system responses, making them a promising tool for the rapid analysis of development scenarios. However, key limitations remain sensitivity to the quality of input data and the limited nature of training samples. The authors note a trend toward hybrid approaches combining machine learning and physical and mathematical modeling, improving the robustness and interpretability of results. Despite the high level of automation, expert participation remains critical at the stages of data preparation, feature selection, and results interpretation.
For citation: Piskunov S.A., Truhachev M.S., Rukavishnikov V.S., Davoodi S. Machine learning for reservoir potential assessment: a critical review of established and promising methods. Bulletin of the Tomsk Polytechnic University. Geo Assets Engineering, 2026, vol. 337, no. 8, pp. 168–183. http://doi.org/10.18799/24131830/2026/8/5657
Unknown authors· Bulletin of the Tomsk Polyte...· 0 citations
Reliable estimation of the minimum miscibility pressure (MMP) is a critical requirement for the successful design of gas injection processes in enhanced oil recovery, as miscibility strongly controls displacement efficiency and recovery performance. Direct laboratory measurements of MMP are expensive and time-intensive, while traditional empirical correlations often fail to account for the complex interactions between fluid composition and reservoir conditions, leading to limited predictive reliability. To overcome these challenges, an interpretable data-driven framework based on a stacking ensemble learning strategy is developed for MMP prediction. The framework combines Random Forest, Gradient Boosting, Support Vector Regression, and XGBoost models and incorporates polynomial feature expansion together with recursive feature elimination to explicitly capture nonlinear and interaction effects among gas composition, oil composition, and thermodynamic variables. Model performance was assessed using an independent test dataset and evaluated using the root mean squared error (RMSE), mean absolute error (MAE), and coefficient of determination (R²). The optimized ensemble model demonstrates a substantial improvement over individual base learners, achieving an RMSE of 2.21, an MAE of 1.317, and an R² of 0.912. Beyond predictive accuracy, model interpretability was enhanced through SHapley Additive exPlanations and partial dependence analysis. The results reveal that the interaction between reservoir temperature and the molecular weight of heavy oil fractions exerts the strongest influence on MMP, followed by key compositional interactions involving nitrogen and carbon dioxide. These trends are physically consistent with established miscibility and phase-behavior principles. The main contribution of this work lies in the integration of an optimized stacking ensemble with interaction-focused feature engineering and explainable artificial intelligence techniques. This combination enables accurate, transparent, and physically meaningful MMP predictions, advancing beyond previous studies that emphasize either accuracy or interpretability alone. The proposed framework offers a practical and reliable tool for supporting gas injection design and decision-making in diverse reservoir systems.
M. Ahmadi· Journal of Petroleum Explora...· 0 citations
Accurately predicting bio-oil yield from biomass pyrolysis is a real challenge due to nonlinear interactions between feedstock physicochemical properties and operating conditions. In addition, high feature dimensionality, uncertainties in experimental measurements, and multicollinearity make prediction accuracy and interpretability even more difficult, which explains the demand for a feature-driven and explainable modeling framework. We developed an advanced regression feature-based machine learning framework trained on a very comprehensive dataset of both biomass characteristics and pyrolysis conditions. Correlation-based feature screening was performed to identify informative variables and examine potential redundancy among input parameters before predictive modeling. The model’s performance in the training, validation, and testing phases was assessed using a variety of statistical metrics. To keep the model understandable, global sensitivity analysis and feature attribution methods were used to determine the relative impacts of each input and their combined effects on yield variation. The suggested system design attained very accurate predictions, demonstrating a test-stage coefficient of determination over 0.91 and consistently low prediction errors. Sensitivity analysis identified working temperature and volatile content as the key substrates for yield, followed by fixed carbon and ash content, while elemental hydrogen and oxygen showed condition-dependent behavior. The agreement in feature scoring between sensitivity methods demonstrates the reliability of the concept-based interpretation. The primary novelty of the paper is incorporating feature selection, predictive modeling, and sensitivity analysis into one feature-oriented yield prediction framework. Unlike purely accuracy-driven studies, this approach simultaneously enhances predictive performance and interpretability, providing quantitative insight into feature importance and system sensitivity.
S. Almansour, L. Alkwai, Kusum Yadav et al.· Scientific Reports· 0 citations
Results indicate that support vector regression (SVR) provides the most consistent overall performance across all regimes and offers a strong balance between accuracy and computational efficiency, and a Bayesian neural network (BNN) achieves competitive predictive performance while additionally enabling uncertainty estimation.
Muhammad Bilal Jan, Zengchao Wu, Mengyu Chai· Metals· 0 citations
With increasing difficulty in oil and gas field development, accurate prediction of well productivity has become crucial. Traditional methods such as analytical solutions and numerical simulations have limited accuracy under heterogeneous and complex flow conditions. This study develops a CNN-LSTM model combining convolutional neural networks (CNN) and long short-term memory networks (LSTM) based on measured data from well J-1 in a shale oil block in eastern China for short-term multi-dimensional time series production forecasting. The model integrates CNN’s feature extraction with LSTM’s temporal modeling, using inputs including production rate, oil pressure, casing pressure, and production time. Compared with the traditional random forest (RF) model, the CNN-LSTM outperforms across R², MAE, MAPE, and RMSE metrics, achieving an R² of 0.9707 and MAPE below 5.1% on the test set. Results demonstrate strong fitting and predictive capabilities, indicating good applicability and potential for broader use in shale oil production forecasting.
Accurate prediction of concrete compressive strength is essential for effective mix design, quality control, and structural performance assessment. Conventional empirical models often exhibit limited accuracy due to the complex and nonlinear interactions among concrete constituents.
This study investigates the applicability of several machine learning models for predicting the compressive strength of concrete using a publicly available experimental dataset comprising 1030 concrete mixtures. Linear regression was adopted as a baseline model and compared with support vector regression, random forest regression, and artificial neural networks.
The performance of machine learning models was meticulously assessed using the coefficient of determination, root mean square error, and mean absolute error. Additionally, the models underwent five-fold cross-validation to evaluate their robustness and generalization capabilities. The results unambiguously demonstrate that machine learning models significantly outperform linear regression models.
Cross-validation results confirm the stability and reliability of the developed models. Feature importance analysis reveals that curing age and cement content are the most influential parameters affecting compressive strength, followed by water content, which is consistent with established concrete material behavior. The findings demonstrate that machine learning models, particularly random forest regression, can serve as effective supporting tools for preliminary concrete mix design and performance evaluation.
S. Rouabah· ITEGAM- Journal of Engineeri...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
