search

equinor / flownet

FlowNet - Data-Driven Reservoir Predictions
GNU General Public License v3.0
63 stars 29 forks source link

Problems with layering option for Norne after removing unused well connections #328

Closed edubarrosTNO closed 3 years ago

edubarrosTNO commented 3 years ago

For the modified version of the Norne model where the unused (non-production) well connections are removed from the schedule file (https://github.com/equinor/flownet-testdata/pull/36), the FlowNet model cannot be successfully created when using the layers option. The grid seems to be generated but the schedule file cannot be created. The printed error message indicates that one of the wells ends up having no connections associated with it and therefore cannot have any controls prescribed. This occurred when I used layers: [[1, 4], [5, 22]], and it does not occur when the layers option is not used.

edubarrosTNO commented 3 years ago

As far as I know this issue is not solved yet. PR equinor/flownet-testdata#36 was merged, but that only means that the version of the Norne model without the unused well connections is uploaded. Now FlowNet runs for the Norne model that is in the flownet-testdata master when running without the layers option, but it still fails when using layers: [[1, 4], [5, 22]]. Or did you manage to run it, @wouterjdb or @LonnekevB ?

LonnekevB commented 3 years ago

Ok. I thought we had this incorporated in another issue, that's why we removed it. But if it now runs without the layers option and not with the layers than this is relevant again indeed! Will get back at it! 👍

LonnekevB commented 3 years ago

Update: Fails on producing the Schedule file because some wells (E-4AH, D-4AH and E-3AH) in the top layer don't have connections. The first two wells are removed by _remove_long_connections(). Setting max_distance to 1200 should fix these two wells. The thrid well E-3AH is already non-connected before this function is called. Not sure yet why this is. Probably because it is on its own island (although the code says all nodes are connected).

wouterjdb commented 3 years ago

D-3AH is connected in the bottom layer... The others are connected only in the top layer (picture shows top layer).

image

LonnekevB commented 3 years ago

Was D-3AH also a problem then? Hadn't encountered an error for that well yet. Or do you mean E-3AH?

wouterjdb commented 3 years ago

Maybe I'm wrong there. I'm only struggling with E-4AH at the moment.

edubarrosTNO commented 3 years ago

Maybe I'm wrong there. I'm only struggling with E-4AH at the moment.

I would be surprised if all connections to E-4AH are removed due to too-long-connections. I say that because there should be injector F-4H that supports E-4AH in the top layer. That well should not be too far. Maybe that connection is triggered as too long because we are using some default values for max_distance or max_distance_fraction options?