Uncertainty quantification and sensitivity analysis for relative permeability models of two-phase flow in porous media

Abstract Computer simulations are usually employed for the prediction of oil reservoir performance under different extraction scenarios. Computational model parameters are adjusted to petrophysical properties of the reservoir and production data to forecast production profiles. However, some key features of these models, such as relative permeabilities of water and oil, are difficult to measure experimentally. As a consequence they can be considered a source of uncertainties, affecting the reliability of predictions. This work presents a study of uncertainty quantification and sensitivity analysis of different relative permeability models to assess the effects of input uncertainty on quantities of interest computed from a model of two-phase flow of water and oil. To explore different wettability regimes two different datasets were used, and two permeability models were employed. The probability distributions of parameters were estimated via the Markov Chain Monte Carlo method. Uncertainty propagation and sensitivity analyses were performed using the polynomial chaos expansion. The paper highlights output quantities that were most impacted by uncertain input data, and also the parameters which most contribute to the output variances.