Impact of the Optimization Procedure on the Equation of State Regression

Abstract

Abstract We investigate the impact of different optimization algorithms and initial guesses on the regression of an equation of state (EoS) for reservoir fluid models, using data for a reservoir fluid from the Brazilian pre-salt region. We employ four optimization strategies, including gradient-based and evolutionary techniques, combined with varied initial guesses to derive alternative regressed EoS models. Our results reveal that while the alternative regressed EoS models demonstrate similar levels of fit to the experimental data, their predictions for phase equilibria are markedly different. This highlights the sensitivity of the EoS regression to subjective choices of an optimization algorithm and an initial guess, and their implications for flow assurance and reservoir simulation and management.