Numerical modeling of mixed wet reservoirs with a thick transition zone typically requires full hysteresis option to be invoked in the reservoir simulator. In this super giant carbonate reservoir with multi-million active cells and twelve reservoir rock types, running the full hysteresis option proved prohibitive due to excessive run times, and in many cases the run fails. Traditionally, engineers resorted to generating scanning curves that mimic the hysteresis option manually using algorithms or spreadsheets. Going further back in history, engineers used imbibition curves which tend to overstate water production. This paper aims at presenting the underlying methodology of developing water-oil capillary pressure scanning curves for mixed wet super-giant carbonate reservoir and to show how significantly this improved the history match quality of water flood and development plan for EOR as well as optimizing the 3D simulation run time.
To develop the appropriate bounding and scanning water-oil capillary pressure curves, Killough Hysteresis model was used. To ensure that the algorithm of this model resulted in reliable scanning curves, commercial simulator sector model was built and the resulted scanning curves from this sector model compared with the mathematical approach used to generate scanning curves over a wide range of water saturation. Validation between full hysteresis option and scanning curves for each Reservoir Rock Type (RRT) was also carried out on simple 1-D & sector models before use of the scanning curves in the full field model.
The scanning curves methodology involved (i) setting up Hysteresis based 1-D model for each RT (ii) assigning initial water saturation in each cell between Swi and Sor and initial pressure for each rock type (iii) run the model for a long time to span the saturation range (iv) iterate on time step (TSTEP) to get smoother scanning curves (v) filter the scanning curves get reduced data points (v) ensure the Krw curves were not crossing the drainage curve for each rock type (vi) created another 1-D model and used scanning curves and (vii) validated the results with use of full Hysteresis options.
Using full hysteresis saturation function curves show significant improvement in the history match quality especially in transition zone where it was difficult to match the flood front with the normal saturation function. It is worth to mention that using scanning curves (121 curves was the optimum number of scanning curves) showed reduction in the simulation run times as compared with the full hysteresis option.