Uncertainty Reduction in Reservoir Modeling

This paper presents a method by which uncertainty in reservoir modeling can be reduced by integrating observed data in the estimation of model parameters. The problem is inverse and ill-posed. Further, since the observed data and the reservoir properties are non-linearly related, reducing uncertainty becomes a non-trivial task. The methodology involves generating multiple reservoir descriptions, conditioned on the observed data. This is facilitated by sampling in a multidimensional parameter space de ned by Voronoi cells. In geophysics, this method has been successfully applied in solving the inverse and non-linear problem of seismogram inversion. Here, we present its application in petroleum reservoir engineering. The strength of the method in performance prediction is validated using ne scale data from the Tenth SPE Comparative Solution Project. We use an upscaled model to history match the ne scale data and forecast the ne grid performance using the maximum likelihood model.