
Авторы:
Издательство:
Дата издания
17.04.2026На каком языке издана
ISSN
ISSN 2224-5278Область исследования
Тип исследования
Аннотация (краткое описание)
Abstract. Relevance. This study presents a machine learning (ML) approach for petrophysical core classification, applied to data from the southeastern margin of the Precaspian Depression - a geologically complex region characterized by heterogeneous reservoirs and diverse lithofacies. The Random Forest (RF) algorithm was chosen as the primary method due to its proven efficiency in handling nonlinear relationships and high-dimensional datasets, making it particularly suitable for complex geological environments. Objective. A Python based workflow integrating Scikit-learn, Pandas, and Streamlit was developed to support the complete petrophysical analytical cycle — from data preprocessing to interactive visualization and interpretation. Methods. The RF model achieved 89% accuracy and an F1-score of 0.90 in lithotype classification (sandstones and siltstones). Porosity and permeability emerged as the most influential features, with the application of a logarithmic permeability scale significantly enhancing interpretability in low-permeability zones. Feature importance was rigorously quantified using the Gini index, enabling effective dataset optimization and 8 ISSN 2224-5278 2.2026 dimensionality reduction. Results and conclusions. The model exhibited strong generalization capabilities, with over 90% test accuracy, and demonstrated robust resistance to overfitting, ensuring reliable performance on unseen data. The interactive web application further enhances usability by offering tools for hyperparameter tuning, feature importance analysis, and dynamic data visualization, supporting rapid and informed decision-making by petrophysicists. In conclusion, the proposed ML framework offers a practical, scalable, and adaptable solution that significantly improves the speed, accuracy, and reliability of lithotype classification, making it a valuable tool for modern petrophysical workflows in geologically complex and heterogeneous reservoir settings.DOI
https://doi.org/10.32014/2026.2518-170X.612Тип публикаций
Практическое значение
The proposed machine learning framework based on the Random Forest algorithm and an interactive Python-based application has significant practical value for petrophysical analysis of core data in geologically complex and heterogeneous reservoir settings. The use of machine learning techniques substantially reduces the time required for data processing and interpretation compared to conventional expert-driven approaches, while improving classification consistency and minimizing subjective bias. The developed workflow enables automated lithotype identification from laboratory core measurements, contributing to more accurate evaluation of reservoir properties and improved quality of geological reservoir models. The interactive web application provides petrophysicists with convenient tools for data visualization, feature importance assessment, and model hyperparameter optimization, facilitating efficient and informed decision-making during petrophysical interpretation. The proposed approach can be applied in the oil and gas industry for reserve estimation, geological and reservoir modeling, reservoir characterization, and the selection of productive intervals for field development. Furthermore, the flexibility and scalability of the framework allow its adaptation to different reservoir types and geological conditions, as well as future integration with additional datasets and advanced machine learning algorithms, supporting the ongoing digital transformation of petrophysical workflows.Квартиль
Источник квартиля
SCImago (Scopus)Impact Factor
1.9Данный документ располагается в коллекциях
Международные публикации ученых SU
Scopus
Файлы документа
Документ не загружен