Using the Low Frequency Observation Data in Automatic History Matching

Abstract

TX 75083-3836, U.S.A., fax 01-972-952-9435. AbstractIntegration of production data in reservoir description requires solving an inverse problem. Typically, a manual procedure is used to adjust the input parameters such that the productiodata can be history matched. Recently, many new techniques have emerged which try to accomplish this objective by using automated techniques. The two most common techniques involve gradient method and adjoint method. Both methods suffer from computationally demanding algorithms, except that gradient method is efficient if number of parameters is small and adjoint method is efficient if the number of observations is small.In a typical history matching of mature reservoir, we are faced with large number of input parameters and large number of observations.This would make both the procedures inefficient.This study proposes a technique to reduce the number of observed data so that history matching can be accomplished more efficiently. Using the low frequency information, we demonstrate that the history matching is more efficient. We further show that the low frequency data provides more robust estimate of input parameters compared to the use of all the available data. The proposed method is validated using a synthetic example as well as actual field example.