Data-driven insights into gas reservoir depths determination: application of ۱D-CNN algorithm in the Kangan and Upper-Dalan formations

Accurately determining the depths of gas reservoirs is a critical challenge, particularly in the Kangan andUpper Dalan formations of the South Pars gas field. Conventional methods, exemplified by Archie's equation, face limitations in such tight carbonate reservoirs, prompting the exploration of advanced techniques like NMR logging. However, the high costs and time-consuming nature of NMR logging necessitate alternative approaches. In this study, a solution grounded in data-driven insights by leveraging a ۱D-CNN (One-Dimensional Convolutional Neural Network) algorithm has been proposed. This deep learning approach aims to provide precise depth determination while overcoming the challenges posed by traditional methods. The study methodology involves the individual implementation of the ۱D-CNN algorithm and its integration into a comprehensive model for enhanced accuracy. By applying this algorithm, we intend to predict gas effective porosity profile based on well logs to determine productive zones and intervals in the Kangan and Upper Dalan formations. The dataset includes information from ۵ wells, incorporating both training and testing wells, with an emphasis on validation through a blind well to ensure robustness. Unlike standard procedures, we go beyond mere prediction by comparing the algorithmic results with actual depths in geographically blind well. The study emphasizes the algorithm's industrial implementation capability by showcasing its effectiveness in predicting reservoir depths. Preliminary results indicate promising accuracy and stability, paving the way for a more intelligent model with practical applications in the delineation of production intervals. In conclusion, our research presents a data-driven approach to gas reservoir depth determination, specifically in the Kangan and Upper Dalan formations, utilizing the ۱D-CNN algorithm. This study not only highlights the potential of this algorithm in overcoming traditional limitations but also underscores its practicality and cost-effectiveness as a valuable alternative to conventional methods and expensive logging techniques in complex reservoirs.