Standard analysis plot wish 2: Recovery rates before and after "project" on field & group/region level

[EivindBakken]

A first step towards "QC" of predictions. A "project" means an new well, a new development or a new IOR projects. Getting a grip on recovery is of value in itself, and can be used for benchmark against analogues (other field or regions/areas within the current field). Need functionality for selecting what level this is wanted for - MVP on field and FIPNUMs?

• To be evaluated as step 2: Consider establishing a plot as used by RAM evaluations on the NCS - recovery ratios vs reservoir complexity index (RCI)

[asnyv]

@EivindBakken Latest release migth already cover some of this, as it includes on-the-fly filtering and aggregation of regional simulation data + calculation of recovery from aggregated in-place. Could you check out the "Regional simulation time series" page at https://webviz-subsurface-example.azurewebsites.net/? Can also see the PR in #348.

Supports region level production, in-place and recovery based on FIPXXX, and also user specified custom filtering (e.g. in the demo: Reek Region 1 is FIPNUM 1, 3 and 5, Region 2 FIPNUM 2, 4 and 6. UpperReek FIPNUM 1 and 2, MidReek FIPNUM 3 and 4, LowerReek FIPNUM 5 and 6).

Have not added delta profiles in this view, but that should be possible to add later (though additional on-the-fly calculations on larger models can get heavy).

Regarding step 2: The RCI requires a decent chunk of user input as far as I know. Would have to do some evaluation of how this should be handled (RCI as direct input, RCI calculated based on a set of user inputs or RCI calculated directly from reservoir model input (if that is even possible).

[EivindBakken]

Thanks for the quick reponse, @asnyv . I checked the Regional simulation time series, and see that a lot of the functionality is there. In the example I can't find the pred ensemble; I believe the functionality I'm pointing to should be considered as part of QC functionality of predictions. (Hist-pred another one).

And yes RCI is a bit messy with a lot of subjectivity: But defining drainage units are similar, and maybe a RCI table could be a different input where the RCIs from RAM is linked to FIPxxx's

[asnyv]

@EivindBakken Unfortunately there is no prediction case in the current test data set, but it is supported in the current plugin. Recovery for a prediction is though currently supported only iffmu-ensemble/libecl is able to catch the restart case, which means that the path in RESTART has to be 72 characters or less (Eclipse SMSPEC limitation, see: https://github.com/equinor/libecl/issues/713). Quite often fails in fmu-projects due to long relative paths. Possible to hardcode a way around it, but not sure if we want to (at least in webviz, should probably be upstream in fmu-ensemble or libecl in that case).

Will have a prediction ensemble (or two scenarios) on the list for when the test data (hopefully quite soon) is replaced with Drogon, but don't think we will prioritize a temporary update of Reek test data.