Automatic History Matching Using the Integration of Response Surface Modeling with a Genetic Algorithm

Abstract

History matching is the process of updating a petroleum reservoir model using its production history. It is a required step before a model is accepted for forecasting production. This work aims at integrating response surface modeling with a genetic algorithm to realize a global optimization method for the history matching process. To achieve this, a proxy model was constructed for the flow simulator outputs for all measurements that enter a global objective function. A genetic algorithm was applied to minimize the proxy model. Then a reservoir model was built based on this minimized proxy model. Construction of a good proxy model requires only a limited number of simulation runs and its simulation time is a couple of seconds. Therefore, the evolution of proxy model using a genetic algorithm takes a short time. Compared to other global optimization techniques, this method can perform the history matching process in a quick, low-cost manner. The method was validated by a field case study. The simulation model used contains 41 years of production history. During the history matching process, a limited number of simulation runs (79) was used to construct a high-quality proxy model and by application of genetic algorithm, the global objective function was reduced from 581.362 to 9.347. The observed oil rate, shut-in pressure, repeated formation test pressure, and water cut (must be zero) were matched using the proposed method.

Keywords:

• genetic algorithm
• global objective function
• history matching
• proxy model
• response surface modeling