Model Selection and Inference for Reservoir History Matching

B028 MODEL SELECTION AND INFERENCE FOR RESERVOIR HISTORY MATCHING Abstract 1 This work presents the approach of the model selection and inference for the characterization of complex reservoirs. The Kullback-Leibler information (the measure of departure of the model from the true system) provides the theoretical basis for our methodology of choosing the best parameterized model for the reservoir. The applicability of alternative statistical model selection approaches such as Akaike’s Schwarz’s and Kashyap criteria are also tested. In many cases there is a substantial model selection uncertainty. Therefore as an extension to the model selection approaches the concept of model averaging