Abstract
Static or dynamic modeling of hydrocarbon reservoirs requires accurate modeling of the initial water saturation, which is normally a function of local rock quality and the presence of capillary forces. Without understanding how the latter two parameters impact the initial saturation distribution, both initial volumetric estimates and subsequent predictions are potentially meaningless. In the past, well-centric saturation-height modeling was the preferred technique for deriving the saturation model, which was based on regression of selected representative logs or capillary pressure experiments. In brownfields with potentially large numbers of suitable wells, this single-well technique does not leverage on the abundance of available data, which for the purposes of reservoir modeling is also on a more representative scale than core data and potentially covers larger reservoir portions, both aerially and vertically. The idea behind this new approach is to use a full-scale inversion technique integrating all suitable well and core data in a process similar to history matching. Initially, a physical model is developed that describes the saturation anywhere in the reservoir while taking into account different rock types and reservoir regions. With an initial set of parameters based on core observations, the process starts with a sensitivity analysis using Monte Carlo simulation. Insensitive parameters are tied to core data at this stage and the well log inversion procedure commences. For this procedure, an objective function describes the mismatch between the simulated and observed saturation. The model calibration, which is fully automatic using nonlinear solvers, will lead to the set of parameters that best fits core and log-observed saturations. The resulting parametric saturation model can be populated in 3D geocellular models and spatially analyzed in terms of mismatch and uncertainties. A comprehensive statistical analysis of the results helps to understand the role of the individual physical processes such as capillary pressure, as well as to define confidence intervals and cross-correlation between parameters. An example of a large carbonate reservoir with about 100 wells illustrates that the novel saturation modeling approach can contribute to significantly improved reservoir characterization.
