Reservoir flow uncertainty assessment using response surface constrained by secondary information

This paper develops a novel response surface method for characterizing the flow and its uncertainty in hydrocarbon reservoirs. Usually, due to computational requirement, the number of flow simulations that can be carried out is limited, which (1) leads to an under sampling of the uncertain space and (2) does not allow to capture curvatures and non-linearities in the reservoir flow behavior in face of uncertain parameters. The proposed method combines conventional flow simulation results with secondary information provided by fast alternatives such as connectivity measurements, single phase tracer or streamline simulation to overcome this sampling problem. Two response surface interpolation algorithms are investigated: discrete smooth interpolation and kriging with an external drift. Both (1) are independent of any predefined regression model, (2) can model complex response surfaces and (3) are able to integrate a secondary “soft” information. Cross validation is then carried out to evaluate the response surface reliability and the contribution of alternative simulation data. A modified version of the SPE 10th benchmark model is used to compare this methodology with standard ones. Results show: (1) the developed approach generates more accurate response surface than conventional techniques and (2) the cost of the method in time is clearly affordable compared to the gain in accuracy which allows a better appraisal of uncertainty on the reservoir flow response.