New Development for Model Selection Criteria in Reservoir History Matching

History matching, which is to find a suitable model, such that the simulator correctly predicts the future production, is absolutely needed for a real reservoir simulation. From mathematical point of view the selected model should be minimized in terms of objective function -OF- (necessary condition for model selection) and also the condition of parameter uniqueness (sufficient condition for model selection) must be fulfilled in a successful reservoir history match. Although the value of minimized OF in normal practice get smaller with increasing the number of parameters, the comparison of OF values does not show the unique solution. In this work we developed Penalized Objective Function (POF) strategy based on penalization concept of OF and principle of parsimony, to avoid over-parameterization to find out most probable unique solution for reservoir history matching. The new strategy (POF) has been implemented successfully on field case and could be applied generally for different deterministic model concerning the reservoir engineering decision.