[ENHANCEMENT] - expand benchmarking - 1st paper

[pujolt]

Hi David,

On the back of SPE-218477-ms, there is a scope to expand the benchmarking, not only to TOUGH2 but other simulators. Doing so - some issues and questions are likely to arise.

I'll start with the simplest model I found: a TOUGH2 study attached and referred here: https://www.[osti.gov/servlets/purl/975373](https://www.osti.gov/servlets/purl/975373)

2009-Andre-Salt Precipitation in Saline Formation on Supercritical CO2 Injection.pdf

I am attaching the case generated from pyopmnearwell - with some modifications to capture the relative permeabilities as specified in the paper, and adding a production well for constant pressure boundaries conditions.

2009_ANDRE_ASPAPER.txt

Overall - the results are similar:

(plots generated with

plopm_command_saltp.md
plopm_command_sgas.md
plopm_command_pressure.md)

The porosity reduction in the injection cell is as well an excellent match for the 5 mol.kgw-1 case.

However - I can't seem to manage to honour the Sgr (as you see the gas saturation are up to 1), even when I try more reinforcement with End-Point Scaling

2009_ANDRE_ASPAPER_SGC.txt

(I noticed that sometimes pyopmnearwell was failing at honouring Kr_n max or Sgn_r for the input deck - will try and come with some examples soon).

If we would be able to match this Sgr - this would constitute a very good comparison case - any idea on how to achieve this ?
I am quite keen as well to have an external eye for any input issue from me.

The idea is to progress with more complex benchmarks - the next step in the same paper are some thermal considerations.

All the best

Thierry

[pujolt]

Hi David
The flue got me - Sgcr won’t impact the remaining 5% brine after drainage and salt deposition - it’s a Swr issue at the other end. Yet unsure how to stop the full vaporization of the brine, and keen to have your views.
Best
Thierry

[daavid00]

Hei, the implemented model in OPM results in the residual water saturation to evaporate, which captures the dry out zone (i.e., no liquid at all), resulting in the interesting handling of the capillary pressure in that zone (e.g., logarithmic extension). I am not sure if TOUGH2 have options to stop the residual to evaporate (which seems the case). If this would be of interest, then I would recommend to open and issue in opm-simulators, and having the option of residual to evaporates might be included (although in my understanding and looking at experimental data, the residuals should evaporate).

[pujolt]

Hi David - back from flue and did look a bit deeper at the logarithmic extension.

Are you taking the same approach than Webb ( https://doi.org/10.1029/2000WR900057 ), as it seems the case in SPE-218477-MS ?
The SPE paper is taking a Pcap0 (“ovendry value”) at 10.7 Pa whereas the initial Webb recommends 10.9 Pa.
I did further checks using the saltprec.toml adapted with some example from Webb – and indeed the Pcap0 obtained with pyopmnearwell is different than the Webb paper.

Palouse_VG_WEBB.txt
Royal_VG_WEBB.txt

Is there a control on Pcap0 or the only control is on Sl* ?
(Note – I am yet unsure on how important this is as not yet in the sensitivities world).

On the same kind of idea, do you have any reference on how pyopm generates pcfact when activated, or what are the best practices there ?

Many thanks for bearing with me !

Best

[pujolt]

Hi David -
I am now in a second comparative case, still TOUGH2 from the same author, this time 2D radial, similar to SPE-218477-MS.
https://doi.org/10.1016/j.ijggc.2013.10.030

2014_ANDRE.txt
(I just edit the PCFACT with:
PCFACT
0.0 1.0
1.0 1.0
/
)

Plots in plopm are done with:
plopm -i 2014_ANDRE_EDIT -v sgas -z 0 -x '[0,100]' -xlnum '6' -b '[0,1.0]' -cnum 11 -cformat '.1f'

(and equivalent).

Some observations:

1. plopm fails to plot PERMFACT for the permeability reduction visualisation
2. Paper mentions salt precipitation saturation of 40% - OPM deposits max 6%... possibly linked to Pc backflow according to paper but OPM struggles with the input data description
3. Opm flow fails if Pc extension for dryout zone explicit (from pyopmnearwell ) or use of PCFACT
4. Near well profiles are quite different if following pyopmnearwell and imposing homogeneous injection for each cell, or letting OPM compute the PLT profile ("Model Controlled Injection":
   COMPDAT
   INJ 1 1 1 10 OPEN 2* 0.01 /

Will let you digest this one before I move to the next one which is going to be a GEM case...

All the best

Thierry

[pujolt]

on 3- OPM fails if Pc extension explicit - I think I found why - it is effectively because the brine capillary flow is much more effective and exacerbates salt precipitation so that SS > 0.91 (the limit from paper) and effectively K=0 at wellbore cells (hence no injection possible).

In the case without the dry-out Pc, at crash time we have:

With Pc dry-out:

The 40% Salt Precipitation is hence a possibility... but there is a contradiction with the Verma & Pruess / Kozeny - Carman input which kills K at SS>9%.
Maybe TOUGH2 does not see the issue as I tend to remember the well model was somehow disconnected ?

I'll try and confrim my thoughts with alternative K-C models.
Best

[daavid00]

Hei, the logarithmic extension in our paper was set to have a maximum pressure which is contained in the range in Fig. 1 in the Webb paper (one could set higher pressures, but numerical issues are expected, one of the reasons is that the internal tables for co2 in OPM Flow have a range of 0.100 to 100.000 MPa). One can play with the entry no. 10 in safu (sl* for cplog) to have different extensions.
The permfact plot (similar for pcfact) can be generated if you run the plopm command where the deck .DATA is (e.g., inside the preprocessing folder, since the table is read from the PERMFACT.INC file).
The election of pcfact tables and well model (continuos vs adapted to the connections) are expected to significantly impact the results.

[pujolt]

Hi David - many thanks for the clarifications

On 1. & 3. I am fine now with the logarithmic extension of Pc but yet unclear on pcfact tables (are those well-assigned Pc ?)- any references you can give me on the use of the pcfact ?

On 2. - I understand the PERMFACT.INC file plotting but I was after the PERMFACT dynamic array (in the UNRST file):

This would allow to compare the permeability reduction factor with the reference paper.

Hope I am clearer,

The OPM run currently fails with an error message on vaporisation - will try some other options and will come back to you to see if it's another user issue or something worth reporting to OPM.

Many thanks for you help !

[daavid00]

The pcfact aims to include the increase in capillary pressure due to reduction in porosity, and the examples in pyopmnearwell set the exponent bie the toml files to 0.5, which is the common model for the Leveret J-function in literature (the PCFACT table is generated by pyopmnearwell then evaluating the given exponent using the Leveret J-function). For PERMFACT if you get the latest plopm, now it should generate the 2D maps, thanks for reporting this issue.

[pujolt]

Hi David,
Thanks for the clear explanation again.
I'll have look at the plopm update in the new year - best wishes for 2026!

[daavid00]

I'll have look at the plopm update in the new year - best wishes for 2026!

Thanks Thierry, best wishes for 2026 to you as well :).