Add Conditional Low-Pressure Padding of Gas Property Tables by bska · Pull Request #3779 · OPM/opm-common

bska · 2023-11-17T12:28:35Z

This PR adds logic to extend gas property tables entered using the PVDG or PVTG keywords to low pressure values if needed. When we compute well production potentials (e.g., summary vectors WxPP) and the well's bottom-hole pressure limit is defaulted we end up using pressures in the order of one atmosphere (1 atm) and this, in turn, may generate negative formation volume factors (mass densities) and/or phase viscosities. The problem is most pronounced for producing wells in history matching mode (keyword WCONHIST), but we should have guards in place to avoid negative densities regardless.

If needed we extend the tables down to 1 barsa and also insert a "limiting" pressure value $p_{\mathrm{lim}}$ corresponding to a maximum FVF value in the range $[1, 10\cdot B_0]$. In this case we create new input tables and redo the property table initialisation with this extended input table.

To this end, we add a way of creating empty DeckItem and DeckKeyword objects which preserve the structural elements of their live counterparts. This, in turn, enables programmatic creation of full DeckItem/DeckKeyword objects by first inserting the padding values and then copying the original table values into those objects.

bska · 2023-11-17T12:29:07Z

I am creating this PR in draft mode because it needs quite a lot of testing before we consider enabling this in the mainline simulator.

bska · 2023-11-17T12:31:22Z

jenkins build this please

bska · 2023-11-21T13:08:32Z

jenkins build this please

bska · 2023-11-22T12:19:01Z

jenkins build this please

bska · 2023-11-28T14:42:39Z

jenkins build this please

akva2 · 2024-02-08T11:48:04Z

sure i get that, but i figured the cosmetic stuff is still warranted. in fact the two first commits could probably be extracted and done now.

bska · 2024-11-15T08:09:47Z

jenkins build this failure_report please

bska · 2024-11-15T09:00:32Z

I have downloaded the failure report locally and will analyse the failures when I have the time.

akva2 · 2025-01-30T15:43:08Z

maybe separate out the two first commits here? in particular the first one is a fix after all.

bska · 2025-01-30T15:52:06Z

maybe separate out the two first commits here? in particular the first one is a fix after all.

Yep, agree. That part has lingered for long enough.

tskille · 2025-02-11T08:31:25Z

This PR really helped respect to robustness for a full field model in Equinor. The asset using this model has already implemented OPM Flow. Therefor I would very much appreciate if this PR could be merged.

Let me know if there is anything that I can do to assist.

bska · 2025-02-11T08:35:06Z

Therefor I would very much appreciate if this PR could be merged.

Thank you for testing. The only thing that's been holding this PR back is lack of testing on real field cases. With your finding in hand, I can justify marking this as "ready for review" now. I need to do a couple of smaller things to have the tests pass, but I'll do those forthwith.

akva2

Looks good to me

akva2 · 2025-02-11T08:41:39Z

opm/input/eclipse/Deck/DeckKeyword.cpp

+
+        ret.m_recordList.clear();
+
+        return ret;


this keeps m_location, intentional? i reckon it is but just to confirm

this keeps m_location, intentional?

It is indeed intentional. I want any diagnostic message to refer to the location at which the input table is entered.

akva2 · 2025-02-11T08:44:25Z

opm/input/eclipse/EclipseState/Tables/Tables.cpp

+        /// Whether or not input table needs padding at low pressures
+        bool inputNeedsPadding() const { return this->needPadding_; }
+
+        /// Low pressure padding rows for input table.  Not needed unless


imho it is better with Needed only if inputNeedsPadding() returns true

Needed only if inputNeedsPadding() returns true

Fair enough. I've pushed an update to rephrase this.

bska

Thanks a lot for looking at this. I've pushed an update to address your comments.

bska · 2025-02-11T09:20:40Z

opm/input/eclipse/Deck/DeckKeyword.cpp

+
+        ret.m_recordList.clear();
+
+        return ret;


this keeps m_location, intentional?

It is indeed intentional. I want any diagnostic message to refer to the location at which the input table is entered.

bska · 2025-02-11T09:21:38Z

opm/input/eclipse/EclipseState/Tables/Tables.cpp

+        /// Whether or not input table needs padding at low pressures
+        bool inputNeedsPadding() const { return this->needPadding_; }
+
+        /// Low pressure padding rows for input table.  Not needed unless


Needed only if inputNeedsPadding() returns true

Fair enough. I've pushed an update to rephrase this.

bska · 2025-02-11T15:07:10Z

jenkins build this failure_report please

bska · 2025-02-11T16:48:15Z

benchmark please

bska · 2025-02-12T11:36:27Z

build this failure_report

I have reviewed the regression failures and they're mostly all of the expected form; the contents of the .INIT file's TAB vector changes due to low-pressure table padding. There are some smaller changes to certain XGRP elements–relative changes in the order of 5.0e-5 at most–which I have not drilled into. Other than that, the main difference is that this PR as a side effect appears to make the dew-point calculation more robust. At least we're getting more non-zero values in the PDEW per-cell result array in the restart file now than we used to. That said, the dew-point calculation should probably be revisited regardless of this work.

I recommend updating the reference solutions and merging this, but we should maybe/probably wait for the benchmark results before doing so?

bska · 2025-02-13T11:53:32Z

benchmark please

I issued that request nearly 48 hours ago now. Is there a high system load on the benchmark runner, @blattms, or did I make a mistake somewhere?

akva2 · 2025-02-13T11:55:49Z

last time i triggered it took 50h.

bska · 2025-02-13T12:03:50Z

last time i triggered it took 50h.

Okay. If that's the case now too then we'll have the response around 8pm CET today.

bska · 2025-02-14T11:55:51Z

last time i triggered it took 50h.

Okay. If that's the case now too then we'll have the response around 8pm CET today.

Still no results. I'm tempted to just update the reference solutions now and merge this. Any objections, @akva2, @atgeirr, @steink?

akva2 · 2025-02-14T11:57:22Z

no objections.

bska · 2025-02-14T12:00:33Z

jenkins build this update_data please

bska · 2025-02-14T13:32:55Z

jenkins build this opm-tests=1292 please

This commit adds a new member function, X emptyStructuralCopy() const to the DeckKeyword and DeckItem classes. This member function will form a "structural" copy of "*this", thus preserving aspects such as the currently active "Dimensions", but removing all data values. This is in preparation of adding low-pressure table expansion for the gas PVT tables (i.e., keywords 'PVDG' and 'PVTG').

This commit adds logic to extend gas property tables entered using the PVDG or PVTG to low pressure values if needed. When we compute well production potentials (e.g., summary vectors WxPP) and the well's bottom-hole pressure limit is defaulted we end up using pressures in the order of one atmosphere (1 atm) and this, in turn, may generate negative formation volume factors (mass densities) and/or phase viscosities. The problem is most pronounced for producing wells in history matching mode (keyword WCONHIST), but we should have guards in place to avoid negative densities regardless. If needed we extend the tables down to 1 barsa and also insert a "limiting" pressure value PLim corresponding to a maximum FVF value in the range [1, 2]. In this case we create new input tables and redo the property table initialisation with this extended input table.

bska · 2025-02-14T14:31:34Z

jenkins build this opm-tests=1292 please

bska · 2025-02-14T15:26:20Z

PR approved, build check is green, and the new reference solutions have been installed on the CI system. I'll merge this into the master branch now to enable the feature.

bska force-pushed the low-pressure-gas-table-padding branch 2 times, most recently from 678035b to b0772f0 Compare November 21, 2023 13:07

bska force-pushed the low-pressure-gas-table-padding branch from b0772f0 to 0615b69 Compare November 22, 2023 12:18

bska force-pushed the low-pressure-gas-table-padding branch 9 times, most recently from 8bcb7d7 to 99e8833 Compare November 28, 2023 10:42

bska force-pushed the low-pressure-gas-table-padding branch 10 times, most recently from 1647b1d to bca7723 Compare December 7, 2023 08:18

bska force-pushed the low-pressure-gas-table-padding branch 3 times, most recently from b4fb4f9 to 2194a00 Compare December 12, 2023 09:03

bska force-pushed the low-pressure-gas-table-padding branch 2 times, most recently from 4eb0e02 to 09bdbab Compare February 19, 2024 08:12

akva2 approved these changes Feb 11, 2025

View reviewed changes

bska commented Feb 11, 2025

View reviewed changes

jenkins4opm mentioned this pull request Feb 14, 2025

Automatic Reference Data Update for PR https://github.com/OPM/opm-common/pull/3779 OPM/opm-tests#1292

Merged

bska added 2 commits February 14, 2025 15:30

vkip mentioned this pull request Feb 20, 2025

set minimum pressure in segments OPM/opm-simulators#6019

Closed

bska mentioned this pull request Mar 7, 2025

avoid negative rs/rv max OPM/opm-simulators#5790

Draft

bska mentioned this pull request Apr 4, 2025

Expand Low Pressure Table Padding to Oil Property Tables #4555

Draft

bska mentioned this pull request May 27, 2025

Add Undersaturated State Handling in Table Initialisation #4621

Draft

Conversation

bska commented Nov 17, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bska commented Nov 17, 2023

Uh oh!

bska commented Nov 17, 2023

Uh oh!

bska commented Nov 21, 2023

Uh oh!

bska commented Nov 22, 2023

Uh oh!

bska commented Nov 28, 2023

Uh oh!

akva2 commented Feb 8, 2024

Uh oh!

bska commented Nov 15, 2024

Uh oh!

bska commented Nov 15, 2024

Uh oh!

akva2 commented Jan 30, 2025

Uh oh!

bska commented Jan 30, 2025

Uh oh!

tskille commented Feb 11, 2025

Uh oh!

bska commented Feb 11, 2025

Uh oh!

akva2 left a comment

Choose a reason for hiding this comment

Uh oh!

akva2 Feb 11, 2025

Choose a reason for hiding this comment

Uh oh!

bska Feb 11, 2025

Choose a reason for hiding this comment

Uh oh!

akva2 Feb 11, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bska Feb 11, 2025

Choose a reason for hiding this comment

Uh oh!

bska left a comment

Choose a reason for hiding this comment

Uh oh!

bska Feb 11, 2025

Choose a reason for hiding this comment

Uh oh!

bska Feb 11, 2025

Choose a reason for hiding this comment

Uh oh!

bska commented Feb 11, 2025

Uh oh!

bska commented Feb 11, 2025

Uh oh!

bska commented Feb 12, 2025

Uh oh!

bska commented Feb 13, 2025

Uh oh!

akva2 commented Feb 13, 2025

Uh oh!

bska commented Feb 13, 2025

Uh oh!

bska commented Feb 14, 2025

Uh oh!

akva2 commented Feb 14, 2025

Uh oh!

bska commented Feb 14, 2025

Uh oh!

bska commented Feb 14, 2025

Uh oh!

bska commented Feb 14, 2025

Uh oh!

bska commented Feb 14, 2025

Uh oh!

Reviewers

Assignees

Labels

bska commented Nov 17, 2023 •

edited

Loading

akva2 Feb 11, 2025 •

edited

Loading