Robust Computational Model Development for Permeability Prediction By Coupling Petrophysical Log Data and Extended Genetic Algorithm

Rock permeability (K) is a crucial petrophysical rock parameter to assess the reservoir fluid flow ability, reservoir quality in reservoir characterization during gas field development. Due to the importance of K in minimizing the risk of exploration and production forecasting, geologists and petroleum reservoir engineers should be able to make reliable K distribution of reservoir rocks. This study aims i) to construct a genetic algorithm-based predictive model to obtain permeability, and ii) also develop an improved model to estimate rock permeability using wireline log data. The gene expression programming (GEP)-based models are constructed using 258 datasets of petrophysical logging parameters, with 65% and 35% for training and testing phases, respectively. An iterative fine-tuning process are employed, involving systematically adjusting the parameters, running the GEP-based models for each case, and evaluating its performance. Different models are assessed to find the most contributing variables for the robust K correlation. It is found that an optimal model configuration with vital variables of gamma-ray, neutron porosity and bulk density for predicting permeability, balancing high accuracy, simplicity, and robustness of highest R² (97%) and the least error (6.24%). This finding revealed a strong foundation for practical applications in gas reservoir studies and further research in this gas field using this new improved model of K. Also, it concludes that the extreme genetic algorithm of GEP offers higher accuracy in K estimation compared to traditional models, offering promising guidelines for resilient model development in the petroleum and geo-energy industry.