Sensitivity of Geostatistical Modeling Parameters on Reservoir Flow Behavior

We have examined what the sensitivity is of geostatistical modeling parameters on reservoir flow behavior, which is characterized by net present value (NPV) and water break through time (WBTT). This can be regarded as a first step to examine the possibility of finding a reduced-order parameter set with geological meaning that can be used to calculate the control action that maximizes an objective function, as done in closed-loop reservoir management. A variogram-based method is used to generate realizations of permeability distributions based on 6 parameters (nugget, sill, two ranges, angle and mean permeability). For each parameter a minimum, mean and maximum number is chosen. Due to the assumed high variability of realizations with identical parameters 30 permeability distributions are generated for each parameter set. Experimental design is applied to reduce the number of reservoir models. The models with different permeability distributions are simulated using a black-oil reservoir simulator and NPV and WBTT are calculated. The results are analyzed using Analysis of Variance and response surfaces. WBTT is mainly sensitive to angle, sill and mean permeability, and NPV to angle and sill. The sensitivity of the interaction between parameters, which affect isotropy and heterogeneity, is also quantified.