Enhancing Well Log Data with Artificial Neural Networks: Insights into Carbonate Reservoirs of the Azkand Formation, Khabbaz Oil Field, Northern Iraq

Accurate reservoir characterization is critical for optimizing hydrocarbon exploration and production, particularly in heterogeneous carbonate reservoirs. This study focuses on the Azkand Formation within the tectonically complex Zagros Fold Belt, employing a multidisciplinary approach to bridge knowledge gaps in petrophysical analysis. By integrating conventional well log data with artificial neural networks (ANNs), we enhanced the prediction of porosity and permeability, calibrated against core measurements. Fractures, critical to fluid flow in low-porosity carbonate succession, were identified using caliper, sonic, neutron, and density logs. Lithology and shale content were determined using neutron-density crossplot and gamma-ray data, respectively, while gas zones were detected via neutron-density crossover analysis. The ANN model demonstrated a high coefficient of determination (R² = 0.93) for porosity predictions and a good correlation (R² = 0.8) for permeability predictions. Additionally, secondary porosity was identified as a dominant feature, resulting from diagenetic processes such as fracturing. This comprehensive workflow highlights the potential of advanced computational techniques combined with traditional methods to characterize complex reservoirs like the Azkand Formation effectively.