Uncertainty Assessment of Reservoir Flow Performance Using Discrete Smooth Interpolation

Fetel, Emmanuel¹, Jean-Laurent Mallet² (1) Nancy School of Geology, Vandoeuvre-les-Nancy, France (2) École Nationale Supérieure de Géologie, INPL/CRPG, Nancy, France

One of the most challenging problems in reservoir modelling is to forecast reservoir flow performance. Uncertainty exists at each level of the modelling, starting from the measure­ment of raw data and their interpretation, to the specifications (physical process, fluid prop­erties, etc.) of the flow model. To account for these uncertainties, a common approach is to generate, using geostatistical techniques, a large number of “equiprobable” models using. They are, then, expected to represent a uniform sampling of all the possible geological sce­narios based on the available data. Such an approach is efficient for assessing the uncer­tainty on a static reservoir property such as the connectivity or the oil in place. However due to time-consuming calculations and computer limitations it can not be applied for dynamic measurement of the reservoir production. In practice, only a limited number of models are considered for detailed flow simulations.

This paper proposes an approach based on a n-dimensional response surface, to fore­cast the reservoir flow performance and characterize the associated uncertainty. The approach is on the Discrete Smooth Interpolation algorithm, to build non-linear n-dimen-sional response surface. This algorithm designed to work in a n-dimensional space is par­ticularly convenient because uncertainty on the data and contradictory data can be taken into account. Moreover, the generated response surface is not affected by data clustering and can be edited locally. The approach has been validated on realistic model and results are consistent and some time even better than classical techniques such as multivariate regres­sion, kriging or spline interpolation. And, finally, applications of such a response surface are presented.