Retrograde Gas Condensate Reservoirs: Reliable Estimation of Dew Point Pressure by the Hybrid Neuro-Fuzzy Connectionist Paradigm

Abstract

Abstract The dew point pressure (DPP) is a crucial thermodynamic property for gas reservoir performance evaluation, gas/condensate characterization, reservoir development and management, and downstream facility design. However, dew point pressure measurement is an expensive and time-consuming task; its estimation using the thermodynamic approaches has convergency problems, and available empirical correlations often provide high uncertainty levels. In this paper, the hybrid neuro-fuzzy connectionist paradigm is developed using 390 literature measurements. The adaptive neuro-fuzzy inference system (ANFIS) topology, including the training algorithm and cluster radius (radii), was determined by combining trial-and-error and statistical analyses. The hybrid optimization algorithm and radii = 0.675 are distinguished as the best characteristics for the ANFIS model. A high value of observed R2 = 0.97948 confirms the excellent performance of the designed approach for calculating the DPP of retrograde gas condensate reservoirs. Furthermore, visual inspections and statistical indices are employed to compare the ANFIS reliability and available empirical correlations. The results showed that the ANFIS model is more accurate than the well-known empirical correlations and previous intelligent paradigms in the literature. The designed ANFIS model, the best empirical correlation, and the most accurate intelligent paradigm in the literature present the absolute average relative deviation (AARD) of 1.60%, 11.25%, and 2.10%, respectively.