Application of Adaptive Neuro-Fuzzy Inference System and Optimization Algorithms for Predicting Methane Gas Viscosity at High Pressures and High temperatures Conditions

Abstract

Accurate estimation of methane viscosity is extremely important for petroleum engineers. Methane viscosity as an important property is used to model multiphase fluid flow in porous media. The viscosity of methane is usually presented as a function of pseudoreduced pressure ( $${P}_{\mathrm{pr}})$$ and pseudoreduced temperature ( $${T}_{\mathrm{pr}})$$ and can be obtained using correlations and charts. Nevertheless, the prediction of methane viscosity at high pressure and high pressure (HPHT) by correlations is associated with some level of uncertainties. Moreover, in the available charts, the methane viscosities are not presented at HPHT conditions. Therefore, having an accurate model that can predict the viscosity of methane at HPHT conditions is beneficial. Therefore, in this study, adaptive neuro-fuzzy inference system (ANFIS) as a powerful intelligent tool is used to predict the viscosity of methane at HPHT using literature experimental data. Two hybrid ANFIS-based models are developed. In the first model, particle swarm optimization (PSO) is employed to find the optimum ANFIS model parameters (ANFIS-PSO), while in the second model, the genetic algorithm (GA) optimization is applied (ANFIS-GA). The results show a better prediction of methane viscosity by the ANFIS models compared to those by models and correlations from the literature. Moreover, ANFIS-GA model shows slightly better prediction than the ANFIS-PSO model. In fact, ANFIS-GA model possesses the lowest average absolute relative deviation, the lowest mean squared error, and the highest correlation coefficient. The findings from the present work demonstrate that the proposed ANFIS-GA model can be easily implemented in any reservoir simulation software, and it provides superior accuracy and performance in reservoir simulators.