Comparison of Genfis-Based ANFIS Models for Permeability Prediction in Hydrocarbon Reservoirs

Permeability prediction in hydrocarbon reservoirs is one of the major challenges in petroleum engineering, as it plays an important role in determining reservoir quality and hydrocarbon production potential. Direct measurement of permeability through core analysis experiments is expensive and time-consuming, and in some wells, such measurements are not possible. The aim of this study was to compare the performance of three Genfis-based ANFIS models (Genfis۱, Genfis۲, and Genfis۳) for permeability prediction using laboratory core data. A total of ۱۹۸ data samples including four independent variables (depth, porosity, density, and compressibility) were used; ۱۴۸ samples for training and ۵۰ samples for testing. Model performance was evaluated using statistical criteria including correlation coefficient (R) and root mean square error (RMSE). The results showed that although Genfis۲ achieved higher accuracy during training, its weak performance on test data indicated overfitting. In contrast, Genfis۳, in which the initial fuzzy inference system is generated using fuzzy C-means clustering, provided more stable performance and superior generalization capability on the test data, with the highest prediction accuracy among the investigated models (R = ۰.۸۳, RMSE=۰.۱۲۱). Therefore, the proposed ANFIS model based on Genfis۳ can be recommended as a fast and low-cost alternative for permeability prediction in reservoirs with limited access to core data.

کلیدواژه ها:

Permeability ، Hydrocarbon Reservoirs ، Neuro-Fuzzy System ، ANFIS ، Genfis