Tuning EOS parameters

[hoanglongbk88]

You sent to me the dissertation of Ph.D "New strategic method to tune equation-of-state to match experimental data for compositional simulation" 1 or 2 months ago.
I followed the strategy the Al-Meshari, Ali Abdallah, Ph.D proposed. There are 7 steps. I try to do exactly as he wrote. But it is so hard to do on PVTi exactly.
do you have any gas condensate datta , such as components analysis, CVD test, CCE test that I can do on PVTi?

[Shakespear]

I would look through these papers

[link Point to another website Only the registered members can access]
and look in the text book

"Phase Behavior of Petroleum Reservoir Fluids"

[Shakespear]

Take a look on page 440

Ahmed-T-Equation-of-State-and-PVT-Analysis.pdf

Then go to references

Also a very good explanation of the tuning steps in the Petroleum Experts PVTP manual

[link Point to another website Only the registered members can access]
look at the Worked Examples

If no other factors such as system integration influence the selection we would recommend PATH 1 to find a matched EoS. With integrated systems the surface facilities modeling package may not allow the changing of pure component properties. In this case select the approach outlined in PATH 2.

PATH 1

Start with all Tcs and Pcs selected (except for N2 and CO2).

If BI coefficients are available select the value between C1 and the heaviest component.

The reason AFs are not chosen initially is that they have a tendency to be pushed to extreme values which in turn lifts the low temperature end of the Phase Envelope to near vertical. Using Tcs and Pcs give a gentler and more controlled path to a solution.

If using Tcs and Pcs does not result in convergence, start adding AFs.

If Volume Shift is required select some Si component properties. Please note that Volume Shift is not recommended until no other combination is found to work( see Volume Shift for more details)

There may be some objections to using the properties for pure components below C6 since they are measurable.

We have found the above method to work universally. however, we do not prevent the user from selecting any combination of properties or approaching the solution differently.

PATH 2

The requirement to hold pure component properties means that the overall matching process becomes more difficult and the approach must change almost from the beginning to maximize the matching possibilities. Create a composition with 5 or more pseudo components. Use the Whitson gamma method within the Advanced Splitting Dialog to get the best set of pseudo components. Add BICs where practical.

Use the Preferences Dialog to switch off regression with pure components. In the Regression Parameter Dialog select all pseudo component Tcs and Pcs and any BICs that may be available. Initially match with volume shift off.

If match is not adequate add volume shift and pseudo AFs to the match parameters. With the restrictions placed on the match parameters it may not be possible to fully match all the properties of the fluid.

The problems encountered with heavily volume-shifted compositions basically arise from an inconsistency in mass balance. Much work has been done within the IPM suite to reduce the effects of this. If volume shift cannot be eliminated then its late inclusion to the match will help to reduce its effect.

See also Step by Step Guide (Appendix A) for a detailed methodology in creating a fully matched composition by either path.

[hoanglongbk88]

Shakespears, after you tune the PVT model, How can you validate whether your model is right or not?

[Shakespear]

If you are using PVTi, you will see how good things are through the Regression Report. Another way is through PLOT option for your different experiment data (CCE, DL etc.) which will show a plot of the regression results and the experimental data for different properties and parameters.

I have been going through the "WORKFLOW" example for PVTi and must say it is confusing. It was better described in a much older version of PVTi. Not sure why, but it seems that someone who knew it very well made one too many options of how a fresh user would interpret the written description.

However in the end this is a TRIAL-and-ERROR process.

The person who I thought might be the best source of info on the regression process has so far not responded :-(

Notice that Petroleum Experts PATH1 and PATH2 has lots of similarity to the "WORKFLOW" tutorial for PVTi.

Does anyone have access to this paper

A systematic investigation into the most suitable data for the development of equations of state for petroleum reservoir fluids

Robert C. Merrill, Jr.a and Trevor M.J. Newleyb

aBP Research International, Sunbury Research Centre, Chertsey Road, Sunbury-on-Thames, Middle--- TW16 7LN Gt. Britain

bBP Exploration (Alaska), P.O. Box 196612, Anchorage, Alaska 99519 USA

10 August 2001

[hoanglongbk88]

I have tried to match the PVT of gas condensate with observations from experiment by trial and error process. But I can't match all the observations exactly simultaneously. The errors of Liquid volume ( in CCE and CVD) and vapor viscosity in CCE are very large. Can you take a look of my PVT file? I can send it to you. Can you give me your email? my email is nguyenhoanglong2303@yahoo.com.

[Shakespear]

A good paper

"How many components ... "

[link Point to another website Only the registered members can access]

[Shakespear]

[link Point to another website Only the registered members can access]
hoanglongbk88 - Take a look at the graphs of the tuning results in this paper. Pay special attention to the Liq. Saturation CVD data. Tell me what you see.

Liq. Sat. experiment is IMPORTANT because it is the one that will model how much condensate will come out of the gas. Hence that will decide how much will be left in the reservoir if you pressure drops below Dew Point.

Dew Point is also important as it will effect the point on the Liq. Sat. curve at which condensate (Liq. Sat.) starts to drop out.

In you data these are the things that are the biggest problem from what I saw. The rest looked good.

I would suggest to carefully spend the time DOING tutorials "Regression" (R) and "Workflow" (W) to fully understand how PVTi works. For instance you have "Superimpose" button. Do you know why? Then in "W" they tell you to CLOSE the project without saving. Do you know why? ....

Let me know here as this will help others.

[hoanglongbk88]

Superimpose button helps you plot the 2 different results in 1 diagram ( such as 2 phase envelopes, before and after you split the component). The phase diagram has not changed much at the reservoir temperature, which indicates a good splitting. In Workflow, they tell you to CLOSE the project without saving because the first splitting result is not good, the mole fraction of FRC2 is smaller than FRC1. WorkFLow advises us that We would rather have a splitting that has more of the mole fraction in the heaviest component. SO that we have to do another splitting.We don't accept the first splitting result , we have to CLOSE the project without saving, and open again the initial component without splitting to do splitting again.

[Shakespear]

Yes, Super allows you to compare after you have changed the EOS parameter etc., to see what has changed. Now this can be used for ANY experiment results plot, not just the Phase Diagram.

The "CLOSE" business is done this way because if you do a regression, splitting etc. NEW parameters have been introduced into PVTi and you have MOVED FROM YOUR STARTING POINT. PVTi has no function that I am aware of moving you back to the START point once you except regression rsults. Hence if you change parameters after 2 cycles of tuning much worse YOU HAVE NO WAY to go back to the start. That is why they said to CLOSE the project without saving and leave the plots there. But recall that the PLOTS were still there which allows you to reload the START POINT, make changes, then make a new plot that is SUPERIMPOSED on the old to see the effect.

Hope that is clear. If so then make a spreadsheet with CASES running down one column and various parameters across the sheet (as per suggestion in the "W" tutorial for the 3 Groupings. For each Grouping , each case of parameter permutation that you run see how your RMS Error has changed from either the ORIGINAL starting point (in the RESULTS report) or from a POINT that you chose that you think was your best tuned result that you want to make better. Then you will be able to plot case vs. error and see what set of parameters is helping and which is not.

What about the Dissertation? What did you notice here?

[hoanglongbk88]

I have a PH.D dissertation you post here. the match of Liq saturation in CVD test is also not good. In my project, error of the liq saturation compared with experiment is 5% . I think it is not bad. What do you think about my project result?

[Shakespear]

I have looked over some cases of condensate fluid systems and from what I have seen it appears that Liq. Sat can be a problem. It is stated in some cases that the problem may be contamination. I have no idea if that is so in your case but do not forget that the lab data may have problems as well as the sample take.

Now your Liq. Sat is going in the right direction. However is not quite there and viscosity has gotten worst (Tune6.PVI).

I would recommend following the Regression and Workflow, and try the parameters groups suggested. The grouping in that dissertation above might also help.

The grouping you have chosen is strange. How was this gone? CO2 should be by itself IMO, it is over 2% and is a unique none-HC component.

In Tune6 it looks like you were only tuning VcritV. Is that correct?

I think you can do better.