The term "windowing technique" means time- dependent replacement of grids and parameters during a simulation run. The window is a confined area within the block model and it contains an other grid than the surrounding basic grid. The basic grid is Cartesian, while the grids inside windows can be Cartesian but with a different spacing, or 2D or 3D irregular grids. A window may also contain a different parameterization of the same basic grid.
The basic Cartesian block model and the windows are independent entities. A well-defined relation between windows and basic grid permits the exchange of the grids during the run. Dual timestepping allows the use of large global timesteps outside windows and other (usually smaller) timesteps inside windows. Effective parallel processing of the windows is also possible.
Potential applications of windows are well test simulation models embedded in coarse grids, modeling of arbitrary-direction horizontal wells, different realizations of stochastic modeling, effective parameter updating during the life of a model, etc.
"Grid construction is a tedious and time-consuming task". This statement was made by many authors, including ourselves. Albeit the situation has changed over the years and many improvements have been presented for certain problems, most of the developments were stand-alone pieces of work and the results were difficult to apply in every-day work.
While simulation grids for practical examples are static, simulation scenarios and physical processes change during the simulation run and during the life of the model. The initially set up block model becomes inappropriate for the new situations. Using static grids, the whole model has to be regridded, reparameterized and eventually rematched to honor the new situation.
Among the cases which might necessitate a new grid are: the introduction of horizontal wells not parallel to grid lines, fine radial grids for well test simulation, (local) tensorial permeability anisotropy, secondary grid refinement when EOR-processes are simulated in the prediction runs (secondary grid refinement means that the refinement is introduced during the simulation run and is not present at the beginning), or irregular grids for honoring stochastic reservoir data.