History matching and uncertainty quantification for velocity dependent relative permeability parameters in a gas condensate reservoir

In gas condensate reservoirs, gas flow at large velocities enhances the gas permeability due to gas-liquid positive coupling which results in near-miscible flow condition. On the other hand, augmented pressure drop due to non-Darcy flow, reduces the gas permeability. Models for the “Positive Coupling” or “non-Darcy flow” include several parameters, which are rarely known from reliable lab special core analysis. We offer a good alternative for tuning of these parameters in which the observed production history data are reproduced from the readjusted simulation model. In this study, history matching on observed production data was carried out using evolutionary optimization algorithms including genetic algorithms, neighborhood algorithm, differential evolution algorithm and particle swarm optimization algorithm, where a faster convergence and lower misfit value were obtained from a genetic algorithm. Then, the “Neighborhood Algorithm–Bayes” was used to perform Bayesian posterior inference on the history matched models and create the posterior cumulative probability distributions for all uncertain parameters. Finally, Bayesian credible intervals for production rate and wellhead pressure were computed in the long-range forecast. Our new approach enables to not only calibrate the gas effective permeability parameters to dynamic reservoir data, but allows to capture the uncertainty with parameter estimation and production forecast.