search

OPM / opm-simulators

Simulator programs and utilities for automatic differentiation.
http://www.opm-project.org
GNU General Public License v3.0
111 stars 122 forks source link

ACTIONX: NEXTSTEP Not Working as Planned #3745

Open OPMUSER opened 2 years ago

OPMUSER commented 2 years ago

First I don't have a E100 run to compare with but I have this in an ACTIONX block:

ACTIONX
        ACT-01                                             /
        DAY  >= 15            AND                          /
        MNTH = MAR            AND                          /
        YEAR = 2021                                        /
/
--
--       NEXT   ALL
--       STEP   TIME
--       ----   ----
NEXTSTEP
         0.5    NO                                         / 
--
--       WELL PRODUCTION AND INJECTION TARGETS                                                      
--                                                                              
--  WELL WELL   TARGET                                                       
--  NAME TARG   VALUE                                                        
WELTARG                                             
B-1H     LIFT   63000.0                                    /
B-2H     LIFT   63000.0                                    /
B-3H     LIFT   63000.0                                    /
C-1H     LIFT   63000.0                                    /
C-2H     LIFT   63000.0                                    /
/                                                                               

ENDACTIO

And I get this in the RSM file:

 11-MAY-2021  0.355921      1405451
 16-MAY-2021  0.369610      1404569
 21-MAY-2021  0.383299      1403687

When I expected this:

 11-MAY-2021  0.355921      1405451
 16-MAY-2021  0.369610      1404569
 16-MAY-2021  XXXXXX      XXXXXX
 21-MAY-2021  0.383299      1403687

For the half day time step.

The second ACTIONX block is:

ACTIONX
        ACT-02                                             /
        DAY  >= 1             AND                          /
        MNTH = SEP            AND                          /
        YEAR = 2021                                        /
/
--
--       NEXT   ALL
--       STEP   TIME
--       ----   ----
NEXTSTEP
         1.0    YES                                        / 

ENDACTIO

So one day time steps from September 1, 2021 until it is reset in the deck on Nov 1, 2021 to 2.5 days. I get this in the RSM file:

  1-SEP-2021  0.665298      1385694
  6-SEP-2021  0.678987      1384860
 11-SEP-2021  0.692676      1384033
 16-SEP-2021  0.706366      1383210
 21-SEP-2021  0.720055      1382392
 26-SEP-2021  0.733744      1381579

That is 5.0 day time steps as per what is set on the TUNNING keyword.

For the deck change on Nov 1, 2021 to 2.5 days, this partially works as you can see below, but then reverts to the TUNNING value of 5.0 days. Is this because the simulator is ignoring the NEXSTEP(NSTEP2) = YES option?

  1-NOV-2021  0.832307      1375844
  3-NOV-2021  0.839151      1375452
  8-NOV-2021  0.852841      1374673
 13-NOV-2021  0.866530      1373897
 18-NOV-2021  0.880219      1373125
 23-NOV-2021  0.893908      1372356
 28-NOV-2021  0.907598      1371590

Then on December 1, 2021 in the deck we have:

DATES    
         1  DEC   2021  / 
/  
--
--       NEXT   ALL
--       STEP   TIME
--       ----   ----
NEXTSTEP
         10.0   YES                                        /

And we get:

  1-DEC-2021  0.914442      1371208
  6-DEC-2021  0.928131      1370446
 11-DEC-2021  0.941821      1369688
 16-DEC-2021  0.955510      1368933
 21-DEC-2021  0.969199      1368180
 26-DEC-2021  0.982888      1367430
 31-DEC-2021  0.996578      1366683

Which again is using the TUNNING keyword's 5.0 day limit.

Note E100 will ignore NEXTSTEP(NSTEP1) if the value is less than the equivalent value on the TUNNING keyword, but is silent on the maximum value. Once I have an E100 run we can check the latter point.

Test deck attached.

ACTIONX_NEXTSTEP.zip

OPMUSER commented 2 years ago

Okay, thanks to Jan Inge Logstein we have the answers.

For the first ACTIONX block, the answer is:

 11-MAR-2021  0.188912
 16-MAR-2021  0.202601
 16-MAR-2021  0.203970
 18-MAR-2021  0.208077
 22-MAR-2021  0.220397

This is different to OPM Flow that applies NEXSTEP to the next reporting time step, that is:

  1-APR-2021  0.246407
  1-APR-2021  0.247775
  3-APR-2021  0.251882

Note both simulators trigger ACT-01 on March 16, 2021. So there is an inconsistency here. I wounder if this is because we used the NO option here.

The second ACTIONX block gives:

  1-SEP-2021  0.665298
  2-SEP-2021  0.668036
  5-SEP-2021  0.676249
 10-SEP-2021  0.689938
 15-SEP-2021  0.703628
 20-SEP-2021  0.717317
 25-SEP-2021  0.731006
 28-SEP-2021  0.739220
  1-OCT-2021  0.747433
  2-OCT-2021  0.750171

E100 applies NEXTSTEP to the reporting time step, with the YES option which is correct, whereas as OPM Flow ignores this.

Now in the deck on Nov 1, 2021 NEXTSTEP is set to 2.5 days and both simulators give the same results.

For the deck on Dec 1, 2021 NEXTSTEP is set to 10 days and E100 has:

  1-DEC-2021  0.914442
 11-DEC-2021  0.941821
 16-DEC-2021  0.955510
 21-DEC-2021  0.969199
 26-DEC-2021  0.982888
 31-DEC-2021  0.996578

Whereas, OPM Flow appears to ignore this.

In summary, I think some of the issues are associated with the TUNNING keyword always taking precedence over the NEXTSTEP keyword, when in fact NEXTSTEP overrides TUNNING(TSMAXZ). And likewise TUNNING(TSINIT) always takes precedence over NEXTSTEP(NSTEP1),

bska commented 2 years ago

Did you run with --enable-tuning=true?

Please keep in mind that Flow ignores all explicit TUNING-related timestepping controls unless the user supplies --enable-tuning=true when launching the simulator. I don't think we'll change that aspect of how Flow operates.

OPMUSER commented 2 years ago

Good point, as I have messed up with this before. But no I did run with --enable-tuning=true as you can confirm with the maximum time step size of 5.0 days as stipulated by TUNNING(TSMAXZ).

gdfldm commented 5 months ago

I have tested this with flow 2024.04-pre.

Flow now agrees and gives the following for the first ACTIONX block:

2021-03-11 00:00:00        0.18891171
2021-03-16 00:00:00        0.20260096
2021-03-16 12:00:00        0.20396988
2021-03-18 00:00:00        0.20807666
2021-03-22 12:00:00          0.220397

Flow now agrees and gives the following for the second ACTIONX block:

2021-09-01 00:00:00        0.66529775
2021-09-02 00:00:00        0.66803557
2021-09-05 00:00:00        0.67624915
2021-09-10 00:00:00        0.68993843
2021-09-15 00:00:00        0.70362765
2021-09-20 00:00:00        0.71731693
2021-09-25 00:00:00        0.73100615
2021-09-28 00:00:00        0.73921973
2021-10-01 00:00:00        0.74743325
2021-10-02 00:00:00        0.75017112

However, flow still seems to override NEXTSTEP=10 with TUNING(TSMAXZ)=5 following Dec 1, 2021:

2021-12-01 00:00:00        0.91444218
2021-12-06 00:00:00         0.9281314
2021-12-11 00:00:00        0.94182068
2021-12-16 00:00:00         0.9555099
2021-12-21 00:00:00        0.96919918
2021-12-26 00:00:00        0.98288846
2021-12-31 00:00:00        0.99657768
blattms commented 4 months ago

Just mentioning OPM/opm-common#3954 here as it might be related (and actually wrong?)