Conditioning Integrated Geological Models to Dynamic Flow Data of Giant Saudi Arabian Reservoir

摘要

A fine-scale full-field multimillion cell geologic model of agiant Saudi Arabian carbonate reservoir has been successfullyconditioned to a thirty-year production/injection history. Thisrepresents the first worldwide case study of the technique forsuch a complicated reservoir with extensive flow history.The initial geologic model was created based on log-derivedporosity data, facies information, and a 3-d stochasticseismic inversion model. It was generated geostatisticallyusing a sequential Gaussian simulation with collocatedcokriging algorithm. It was desired to condition this initialmodel based on a thirty-year injection/production history.Fast streamline simulations were performed on thegeological porosity and permeability models using fieldinjection/production constraints. The flow conditioningalgorithm iteratively modified the permeability field such thatsimulated injection/production behavior more closelyapproached the supplied injection/production history. Thealgorithm uses analogies to seismic travel time andtomography, involving iterative linearization of the time-of-flightexpression about a known initial model based onstatic data.The final flow conditioned model was then validated usingmore thorough streamline simulations verifying improvementsin the match of the simulated injection/production response tothat of the historical response. Additionally, inspection of thedifference between the initial and the flow-conditioned finalmodel showed permeability modifications followed a strongdirectional trend consistent with well-known geologicaltrends. Results from the final conditioned models showedmore than 30% improvement in terms of history-matchingthan that of initial model.