Estimation of reservoir permeability and free fluid porosity from NMR logs using a hybrid ANN–PSO model and clustering: a case study of the Asmari reservoir, Ahvaz oil field

Nuclear Magnetic Resonance (NMR) logging is a powerful tool for calculating key petrophysical parameters of reservoirs, including bound fluid volume (BFV), permeability, and effective porosity (free fluid porosity) in hydrocarbon reservoirs. In this study, a three-stage hybrid approach is proposed, which integrates conventional petrophysical log data with a combined Artificial Neural Network–Particle Swarm Optimization (ANN–PSO) model and Electrofacies (EF)-based clustering, leading to improved prediction accuracy and generalizability. The input data were obtained from two wells in the Asmari oil reservoir of the Ahvaz field and included corrected gamma-ray logs (CGR), neutron porosity (NPHI), sonic transit time (DT), effective porosity (PHIE), and electrical resistivity (RT). Data quality was ensured through statistical analysis, outlier removal, and cross-validation. First, an artificial neural network was trained to establish correlations between NMR parameters and conventional logs. Then, the ANN was optimized using the PSO algorithm and applied to the entire dataset. Finally, using the Self-Organizing Map (SOM) clustering method, the data were divided into three EF groups, and separate ANN–PSO models were trained for each group. In the test dataset, the R² values for free fluid porosity (FFP) and permeability increased from 0.86 to 0.89 in the ANN model to 0.88 and 0.89 in the ANN–PSO model, and after EF-based clustering, reached 0.92 and 0.95, respectively. The results demonstrated that clustering improved feature separation, model convergence, and its generalizability. The main innovation of this study is the integration of optimized ANN–PSO models with EF-based clustering for significantly enhanced prediction of NMR parameters in oil reservoirs.