Machine Learning-Based Rock Facies Classification for Improved Reservoir Characterization in Niger Delta

Abstract

The Niger Delta, a cornerstone of Nigeria's oil and gas sector, plays a significant
role in the Nation's energy landscape. This research concentrates on enhancing
reservoir characterization, specifically emphasizing advancing rock facies
classification. Employing advanced machine-learning methodologies and a dataset
from 12 wells containing crucial well log parameters, such as Gamma Ray,
Resistivity Micro-Spherical, Volume of Shale, Resistivity Deep, Resistivity
Medium, Density, and Porosity, we conducted a rigorous evaluation of various
classification models. The Random Forest algorithm emerged as the optimal
choice, achieving an impressive F1 score of 0.93 and an accuracy of 0.93 on the
cross-validation set. A meticulous analysis of identified facies classes, including
Shale, Lower Shoreface, Middle Shoreface, Upper Shoreface, Transition
Shoreface, Over Bank, and Channel, through confusion matrices, offered
profound insights into the Model's efficiency. Feature importance analysis
underscored the critical role of variables such as volume of shale, gamma ray,
porosity, and bulk density in driving accurate predictions. This research
significantly advances subsurface exploration in the Niger Delta, highlighting the
effectiveness of machine learning for geologic characterization within the region's
intricate geological landscape.