Enhancing Lithofacies Classification Reliability Using K-Means Clustering, Cut- Off Method, and AI: Gran Field Case Study

This article addresses the limitations of traditional petrophysical interpretation and lithofacies analysis methods used in commercial software solutions, such as sub-jectivity, insufficient detail, and reliability, particularly in cases of complex reservoir structures. Accordingly, the development of automated lithofacies analysis tools using Artificial Intelligence (AI) and Machine Learning (ML) is a relevant objective for en-hancing the reliability of geological modeling and reservoir evaluation. The authors have developed an innovative methodological approach for auto-mated lithofacies classification of well logging data, demonstrated via case study of Gran Field. The methodology is centered on the k-means unsupervised clustering al-gorithm, specifically adapted for comprehensive petrophysical data analysis. It is demonstrated that the proposed approach effectively partitions the geological section into lithofacies and ensures the reliability of petrophysical interpretation re-sults. The optimal number of clusters (k=3) was determined using the Silhouette Coef-ficient, and the results were visualized using the Principal Component Analysis (PCA) method, confirming that the identified groups correspond to petrophysical patterns. The clustering results, incorporating PCA, showed clear separation into clay, silt-stone, and sandstone lithofacies. The k-means-based approach mitigates the primary limitations of traditional methods reliant on the subjective selection of cut-off values and forms a reliable foundation for building advanced geological and hydrodynamic models. To facilitate practical application, a Python-based web interface was developed using the Streamlit framework. This application offers a user-friendly interface for preprocessing well-log data, performing clustering, and visualizing results, bridging the gap between advanced ML algorithms and specialists without programming ex-pertise. Comparative analysis reveals that the k-means algorithm outperforms alternative methods across several key metrics, notably in interpretability and the structural co-herence of the results. Future development prospects include the integration of densi-ty-based clustering algorithms, such as DBSCAN, to increase the system's adaptability in complex geological sections. This will open new possibilities for intelligent analyti-cal systems in the field of reservoir evaluation and resource assessment.