Enhanced clustering techniques for lithofacies and electrofacies identification in post-salt carbonate and siliciclastic reservoirs: a case study from the campos basin, Brazil

The effective characterization and management of hydrocarbon reservoirs require a thorough understanding of their petrophysical properties, from exploration to production. Modern borehole logging and interpretation techniques play a crucial role in reducing operational costs while enhancing reservoir evaluation. Petrophysical analysis is fundamental in this context, enabling the interpretation of subsurface lithology and the assessment of key rock-fluid interactions, including porosity, permeability, and fluid saturation. These parameters are essential for identifying source rocks, seals, reservoir zones, and aquifers. Geophysical well logs are among the most reliable tools for determining geological formations and their petrophysical attributes. This study investigates two post-salt reservoirs in the Campos Basin, southeastern Brazil, utilizing data from four boreholes - two as reference wells and two as blind tests. A conventional suite of well logs was employed to define electrofacies, supported by geological data. Initial cluster analysis was conducted using singular value decomposition, hierarchical clustering (dendrograms), neutron-density lithological cross-plots, and principal component analysis (PCA) to identify inherent groupings within the dataset. Subsequently, nine unsupervised classification techniques-including eight clustering algorithms (e.g., k-means, k-medoids, Gaussian mixture models, spectral clustering, k-nearest neighbors, subtractive fuzzy clustering, fuzzy c-means, and agglomerative hierarchical clustering) and one neural-based mapping method (competitive neural network) - were applied. The exploratory data analysis was essential to understand the statistical behavior and interrelationships among the petrophysical logs, serving as a foundational step for the effective application of the clustering algorithms. Despite the diverse mathematical foundations of the clustering algorithms, the three electrofacies were reliably correlated with lithofacies. In the siliciclastic reservoir, these corresponded to sandstone, shale, and limestone, while in the carbonate reservoir, they were classified as grainstones, wackestones, and cemented grainstones. This study highlights the efficacy of multi-algorithm clustering in petrophysical facies classification, providing a reliable framework for reservoir characterization in analogous geological settings.