Coarse 3D grids: a general discussion

[augustjohansson]

Coarse 3D grids may be an interesting complement to the 1D tubes currently implemented. This issue is for a general discussion on pros and cons with this approach, and also discuss ideas for realizing such grids in flownet. For the latter we have discussed 1) the COARSEN keyword in Eclipse, and 2) our own implementation.

Regarding 1) preliminary studies by @edubarrosTNO show that ECL can handle the COARSEN keyword and with some manual work we can probably get the data needed (such as PORV and transmissibilities) for writing OPM input files. There are some issues (bluntly copied from the pdf of Eduardo from 2021-10-08):
a. Can we modify .EGRID, .INIT files with modified / history matched properties mentioned above? (Eduardo think there are libecl methods for that)
b. We can write connection factors to the COMPDAT section of SCHEDULE file?
c. When adding two wells (or single well with multiple connections) in fine cells belonging to the same coarse cell,
Eclipse seems to move all connections to the representative fine cell of the coarse cell. Can OPM-Flow also handle
this?
d. How to represent this grid for representing geometry (visualization purposes)? Cartesian grid? Dummy network of “tubes” connecting the representative fine cells of the coarser cells (only those kept active)?

Obviously 2) will provide us much more detailed control over how the partitioning is done, but with the drawback that we have to do it ourselves. This is work in progress.