Constraining Reservoir Facies Models to Dynamic Data - Impact of Spatial Distribution Uncertainty on Production Forecasts

Abstract This paper presents a new approach to constrain reservoir facies models to dynamic data. This approach is mainly based on the combination of an optimization method called the simplex method and parameterization techniques such as the gradual deformation method. This parameterization technique is well known for its efficiency in constraining geostatistical models to dynamic data. Originaly developped for continuous Gaussian models, the gradual deformation method was extended to facies models [12] and more recently to any kind of geostatistical model. However, the numerical behavior of the inversion can be strongly affected by discontinuities in the model changes in case of facies-based models. Our approach proposes a robust inversion method coupled with the gradual deformation method for the calibration of facies models. The methodology we present is directly inspired by the construction of a reservoir facies model. We start by generating a related Gaussian continuous geostatistical model. In a second phase this continuous model is transformed according to the number of facies and their proportions to obtain a consistent facies model. Since a direct parameterization of the facies model does not allow the preservation of the geological properties of the model, we perform the parameterization on the related continuous model and then we proceed to the model truncation. In this manner, we can optimize the deformation parameters involved in the parameterization using the simplex method to obtain an history match. The efficiency of this approach is illustrated by matching an interference test. The parameterization phase was performed using the gradual deformation method. Constrained facies models were thus obtained. A second phase of this study was devoted to quantifying the impact of uncertainty of spatial facies distribution on production forecasts. In an integrated process, we combine the simplex approach with experimental design theory and the joint modeling technique to achieve a probabilistic production forecast which takes into account the uncertainty on spatial facies distribution and classical reservoir parameters while preserving the history match. This application demonstrates the importance of the interference test matching in reducing the uncertainties on production forecasts.