Analyzing the impact of clay minerals on the reservoir quality of the Lower Goru Formation using Unsupervised Machine Learning.

Abstract

The reservoir quality of the Lower Goru Formation is highly variable due to its heterogeneous nature influenced by sea level fluctuations during the Early Cretaceous period. This study applies an unsupervised machine learning workflow, integrating Principal Component Analysis (PCA) for dimensionality reduction, Self-Organizing Maps (SOM) for clustering, and fuzzy classification for geological labeling, alongside petrophysical evaluation and cross-plot analysis, to assess the impact of clay minerals on the reservoir quality of the Lower Goru Formation in the NIM-Tay block, Lower Indus Basin, Pakistan. Petrophysical analysis delineates a potential reservoir zone (1455-1517 m) characterized by 13.9% effective porosity and 27.3% water saturation. The first four principal components explain approximately 90% of the dataset variance. Electrofacies classification distinguishes four facies-Impermeable Reservoir, Potential Reservoir, Non-Reservoir, and Tight Reservoir-each corresponding to specific clay mineral assemblages. Cross-plot and electrofacies analysis reveal that facies dominated by chlorite and montmorillonite preserve porosity (15%) and permeability (888.87 mD), whereas kaolinite-rich and mixed-layer clay facies significantly reduce reservoir quality. This study provides a reproducible and scalable framework for integrating machine learning with petrophysical workflows, offering improved reservoir characterization not only in the Lower Indus Basin but also in similar heterogeneous sandstone reservoirs globally.