Use well rates from well_state_nupcol instead of previous_well_state in updateWellRates #6128
Conversation
…in updateWellRates
|
jenkins build this failure_report please |
totto82
changed the title
|
For what it's worth: I had a quick look at one of the more severe test failures (0A4_GRCTRL_LRAT_LRAT_GGR_BASE_MODEL2_MSW.DATA), and the behavior with this PR is much more reasonable than with master. Specifically, the early time results are closer to what is obtained with a very short initial time-step. |
I agree. I implemented this some years ago. Back then, it was necessary to avoid severe oscillations, but I don't see that issue anymore. In any case, setting NUPCOL to something reasonable, like 3-5, is a better workaround IMO. Should we try to get this in before the release? |
IMHO, yes. |
|
Thanks for the request. It is something interesting to see that the function prototype is using I will invite @steink regarding this, he might have more insight regarding this part. |
There was a problem hiding this comment.
With more learning and thinking, this is probably fine and should follow what with testing indicates. Not working on this area much and not familiar with the exact comparison between the two different well states. It is difficult for me to understand deeply how it affects the output exactly. Will wait for @steink 's comments, otherwise, it should be able to go in.
|
This PR seems very reasonnable to me. Will just do a test-run, then report back. |
|
jenkins build this update_data please |
Reason: PR OPM/opm-simulators#6128 opm-common = 771e89b8b9ee76ca3789559e687c6582861e9ecf opm-grid = a8afffe952be60f86a5ae60f562a52e68aec22dd opm-simulators = cf97d4cab3693139e7343c192b634f1a102d4cae ### Changed Tests ### * 0a1_grpctl_stw_model2 * 0a1_grpctl_msw_model2 * 0a2_grpctl_stw_model2 * 0a2_grpctl_msw_model2 * 0a3_grpctl_stw_model2 * 0a3_grpctl_msw_model2 * 0a4_grpctl_stw_model2 * 0a4_grpctl_msw_model2 * udq_actionx * udq_wconprod * udq_pyaction * model4_udq_group * model4_gefac
|
jenkins build this opm-tests=1321 please |
Automatic Reference Data Update for PR OPM/opm-simulators#6128
|
the reference has been updated. I am getting the PR in now. |
The previous_well_state was used to avoid oscillations long time ago. Using the nupcol rates aligns more with what we do elsewhere in the code and will be more accurate in particular at startup of wells. Test failures are expected. A qualitative look at the failed tests indicates that results looks more plausible.