Applying Genetic Programming to Reservoir History Matching Problem

Abstract

History matching is the process of updating apetroleum reservoir model using production data. It isa required step before a reservoir model is acceptedfor forecasting production. The process is normallycarried out by flow simulation, which is verytime-consuming. As a result, only a small number ofsimulation runs are conducted and the history matchingresults are normally unsatisfactory.In this work, we introduce a methodology using geneticprogramming (GP) to construct a proxy for reservoirsimulator. Acting as a surrogate for the computersimulator, the cheap GP proxy can evaluate a largenumber (millions) of reservoir models within a veryshort time frame. Collectively, the identifiedgood-matching reservoir models provide us withcomprehensive information about the reservoir.Moreover, we can use these models to forecast futureproduction, which is closer to the reality than theforecasts derived from a small number of computersimulation runs.We have applied the proposed technique to a WestAfrican oil field that has complex geology. The resultsshow that GP is able to deliver high quality proxies.Meanwhile, important information about the reservoirswas revealed from the study. Overall, the project hassuccessfully achieved the goal of improving the qualityof history matching results without increasing thenumber of reservoir simulation runs. This resultsuggests this novel history matching approach might beeffective for other reservoirs with complex geology ora significant amount of production data.