Using Well Windows in Full-Field Reservoir Simulation

Abstract

Summary This paper presents the general and practical applicability of the "windowing technique"1 to model wells in full-field reservoir simulation. Windows modeling the near-wellbore area and the wellblock itself, constructed by the Perpendicular-Bisectional (PEBI) or k-orthogonal PEBI (k-PEBI) method, will be introduced for all wells of a reservoir or for selected "problem wells." The windowing technique, introduced by Deimbacher and Heinemann,1 allows a time-dependent replacement of grids for a defined area during a simulation run. A window can represent any kind of well with a gridded wellbore and an appropriate grid pattern around the well. Such an approach makes the generally used Peaceman well model superfluous. Because the gridded wellbore and the gridblocks around it are small (some cubic feet), the computational stability requires small timesteps and a greater number of Newton-Raphson iterations. It is obvious that this is not feasible if the solution for the full-scale model and the well windows must be performed simultaneously. Therefore, in a first step, the fully implicit solution for the full-scale model will be calculated, but the inner blocks of the windows are solved for the pressure only, without updating the saturations and mole fractions. This solution provides the boundary influx for the windows. In a second step, the windows are calculated for the same overall timestep with up to 1,000 small local steps. This paper presents the general and practical applicability of this method. Windows, constructed by the PEBI (k-PEBI) method, can be introduced automatically for all the wells (vertical, horizontal, and slanted) or for certain "problem wells." For testing purposes, real case field models were used. It will be shown that the quality of the results obtained for the model calculated with integrated radial grids around the wells and small overall timesteps are equal to those obtained for the same model using the windowing and local timestepping techniques. Further, it will be shown that solving the window model with large timestep lengths for the first solution step and small lengths for the second solution step results in equal or smaller CPU times in comparison to the conventional model.