A Performance Analysis and Comparison of Different Fuzzy Inference Models for Advanced Prediction of Reservoir Properties

Abstract

One of the major concerns of the oil and gas industry is the precise interpretation of well logs for the accurate determination of reservoir properties. Porosity and water saturation are the two fundamental properties having a substantial impact on the field operations and reservoir management. The conventional statistical methods of reservoir property estimations are complex procedures and usually need various stages of correction. On the other hand, interpretation of well logs from uncored intervals would probably result in large errors because of lack of information and the consequent misconceptions in the correction stages. In the present study, the performance of different types of fuzzy inference models is compared. The comparisons indicate that adaptive network-based fuzzy inference systems would perform better than the Sugeno or Mamdani models generated by the aid of fuzzy C-means clustering, while in the absence of a network-based system, a Sugeno model showed better performance than the Mamdani one for the especial case of porosity determination.