Tuning EOS parameters

[Shakespear]

hoanglongbk88 - The Model that you are building. Is there gas injection or is it just a depletion process?

[hoanglongbk88]

I have grouped CO2 +C2 into pseudo-component as PVTi advices. The viscosity I tuned the last, so that I only choose Vcrit to regress, others PVT I set weight equals 0.
I will try to read the dissertation carefully.
Thanks Shakepears a lot.

[hoanglongbk88]

I have grouped CO2+C2 into 1 pseudo-component (X2 in Tune6). I tuned viscosity the last after other PVT properties were matched. therefore you see that I only choose a Vcrit to regress.
I will try to read the above dissertation carefully.
thank you so much.

[Shakespear]

I would not group CO2.
N2+C1
C2
C3
IC4+NC4
IC5+NC5+C6
C7+

This is how I would have them grouped

Do the spreadsheet idea. Even if you did the basic procedure suggested in "Workflow" then graph the total error % for various key experimental data vs. case (1,2,3,4 ...) it would visually show you how changing the parameters of the EOS moves you from the start point.

What is your model modeling? Is there gas injection going on in the model?

[hoanglongbk88]

Why didn't you group CO2 into C2?

[Shakespear]

Observe how the groupings are done in various papers and as demonstrated in text books. They go sequentially so as to maintain monotonicity which is reflected in the components properties. CO2 is MUCH different then C2, so you do not group it. Better to go in sequence when making the groups.

[hoanglongbk88]

PVTi advises us to group CO2 to C2 because when you plot the K value of component vs pressure, the CO2 line is next to C2 line. So that we group Co2+ C2. Although molcular weight of CO2 is different so far from C2.

[itag]

Hi
I've been following this thread and am most impressed with the list of quality references offered up by Shakespear here. Great job.
However - there are many question remaining unanswered.
a) Where do the samples come from - good down-hole samples vs recombined separator samples ?
b) If separator samples, the the bubble point of the oil should match the separator P,T - you can often characterize & split this 'oil' first then treat the recombination as a separate step ( BTW the bubble point gas should match your sep gas composition! - another good check.)
c) I'm always a little suspicious of recombination ratios - and would have no problem varying the recombination fractions a little to see how this affected the final (mixture) C5+. This means tuning 2 or 3 mixtures and keeping the 'best'
d) Matching dropout curve - and getting C5+ yields right is probably more important than getting Pdew exactly right ( 5% error in Pdew may be OK) -
e) If the field is undeveloped - then the RFT/MDT tool usually provides for a very accurate in-situ gas density .... this can help your C5+ matching. I'm very surprised when folk don't use this density data when its available. ( I'm an RE and density is important in getting initial Pressures right!!)
( Its easy to include gas density in a PVTi experiment)
f) I prefer the Whitson Splitting ( over the simple mole splitting in PVTi) ... Split C7+ into 2 or 3 fractions out to C45 or C50
g) In my mind splitting only makes sense if subsequent lumping is going to use more than just C7+ ( in other words if you lump everything back to C7+ why bother splitting in the 1st place !). Splitting should help you match experiments - if not stop and read the literature!

Key questions for me: Where do the samples come from ? and Shakespear's question: What is the development plan for this reservoir ? Gas cycling ? pure depletion , Strong aquifer ?

good thread
itag

[Shakespear]

itag - all good points. I did not want to get into the data QC as this is a big issue and not my specialty though I am aware of its significance.

From what I understand hoanglongbk88 is doing a dissertation and has limited time to do his work. So I though it would be best to keep things simple.

The reason I was asking about gas injection is that if there is none he would not need to run the model using E300 which is a Bear.

Please add any knowledge you have in this area as sooner or later someone may have a need for it.

I can not say enough about the Whitson course that is on the net. It is long but well worth the time to go through.

[hoanglongbk88]

Shakespear, I am trying to match my file I sent to you by PVTsim and compare the 2 results. What do you think about this idea?

[Shakespear]

I know you have limited time thus there is only so much that you can do. By referring you to the other dissertations and papers where EOS tuning was done I was trying to let you see what kind of results people get. Sometimes it looks good, sometimes some lab data just is not replicated well and they apparently stopped trying. The Liq. Sat. sometimes simply will be a pain in the butt and unless you are an expert you will have no clue how to fix it.

In such a situation you can only do this EOS tuning in a systematic way and try various permutations of parameters and see how it moves you around. In my opinion the best way to do this is through some form of plotting. Hence the idea of a spreadsheet.

So do your best, come as close as you can and then go on, because there is only 24 hr. in a day and you do need to rest

I am trying to write a C program to summarize the regression report but after a day of monkeying around with Turboc C (running in DOS using DOSbox) I realized that errors I was seeing from the program were not mine but strange memory management issues of DOSbox. In a few words, "This was not the best route for me to take".

However with each round of permutation of parameters a summarization needs to be done to get some summary stats to plot or analyze. Looking in those PVTi results panels is good if you have done this a zillion times and know PVT behavior like your girlfriend, otherwise key data needs to be extracted to allow you to compare to other results.

[hoanglongbk88]

I am trying my best to match the PVT of gas condensate on PVTsim, then compare the results from 2 software. But using PVTsim is a little bit more complicated than PVTi or Maybe I don't have enough time to learn how to use PVTsim as PVTi. It is close to deadline to submit my dissertation. Should I accept the result I tuned on PVTi although I know this result is not good?