 Hi Muritala,

Find below some guidelines about the subject, note that these are basics (I'm still
learning the advanced stuff):
I assume you have a representative PVT report.
I assume the fluid composition is available till C29 + C30+ ( these are the
standard for lab now), otherwise you need splitting.

1. Enter your sample and experiments including BP, CCE, DL & SEPS.

1. Firstly match the BP ( don't use other experiments in regression), you can start
by regression on OMEGA, then OMEGB, then Pc Then Tc . use only one variable
each time , for the next variable always reset. you can also use BIC between C1-
C+.

For each variable, check the applied modifier, choose the variable that gives you
the closest match with smallest modifier.
* is matched, set a weight of let's say 1000 to sat. pressure.

* other experiments for regression, don't use viscosities in regression for now.

* here start the trial and error steps, check your match and pick the worst
matched observation ( for example DL-GOR), the selected variables to regress as
the same for BP ( I read some references for selecting variables and each
reference uses different strategies, however there are some rules they agree on)

In general, when deciding which variables to choose, consider the following

sensitivities:
* (Tc, pc), or Omegas of plus fraction(s): for saturation pressure, liquid dropout,
etc.

* BICs between C1 and plus fraction(s).

* Volume shift: for Z-factors, densities, etc.

* Check monotonicity

1. You may play on observations weight and always keep a log of your workflow
and used variables.

1. Match viscosities using Zc or Vc as regression variables.

Mohammed,

4 months ago
o Like(2)
o Report spam


David Tipping
Hello Muritala
 Before doing anything with PVTi, or any PVT fluid modelling software, here are points to
consider. You do not want to jump into the software first otherwise you risk doing an
analysis that is
• not relevant to your objectives
• not representative of the reservoir fluid
• not valid
* is necessary to define the objectives first: Do you want to use the fluid model in a
reservoir simulator for predicting surface oil rates? The data needed can be different for
that for predcitng recovery from a miscible gas flood.

* the samples representative of the reservoir fluid? If the well is flowing below the
saturation pressure, often the GOR of the samples is much greater than the true
reservoir GOR. This needs to be confirmed. For example, looking at trends in GOR in
time.

* the PVT lab data valid? For example, are the equilibrium ratios of the liquid and gas
components realistic? (Check the publication IPTC 13784 Quality Assessment and
Consistency Evaluation of PVT Data.) Many times it is difficult to match an EOS to data
because the data are not realistic.

* data should you use? Example: if it is a gas-condensate reservoir with a liquid drop-out
below about 6%, the CVD data are generally not important. To calibrate the EOS the
separator GOR and stock tank are priority.

Useful guidelines are posted on the Reservoir PVT forum.

David Tipping

4 months ago
o Like(4)
o Report spam


Brian Moffatt
Muritala,

I fully endorse what agree with David says; before you start anything decide what your
objectives are and how accurate they need to be. Work back from the end result you
need to the studies and modelling you need to achieve that. Unfortunately the industry
focuses on heavily workflows, often driven by software requirements, rather than
considering if the workflows will give you (and other stakeholders) what you need.

Your objectives might be to decide a suitable EOR method, to model depletion, to model
gas or water injection, to balance contaminant levels, optimise production, account for
particular effects such as velocity stripping, any number of aspects. It is a good idea to
ask beyond your home discipline what is useful here.

To answer your specific question, you can add a separator test [SEPS] in the experiments
to match to. DST GOR data are preferable. Personally I would not tune to any low
pressure laboratory separator test as they can be inaccurate [low pressure, weak
volumetrics, viscous oil and long equilibration times]. If the model works for the primary
stage and the CCE I would leave it well alone!

Brian.
 4 months ago
o Like(3)
o Report spam


Maksim F.
Brian,

interesting observation about low separator pressures. What would be your Psep
threshold to go or not for the integration of this data in the EoS tuning?

4 months ago
o Like(0)
o Report spam


Brian Moffatt
Good question Maksim!

I have found below 200 psi results can get dubious, below 100 psi unreliable. If the data
are really needed, and people should question if this is the best way to get it, then I
would not recommend a standard separator test but instead perform a simple flash in a
PVT cell where you can allow sufficient time for equilibration. You can then measure both
the composition and density of each phase. This will provide a far more robust calibration
point for your fluid model and you can check consistency of compositions with an
equilibrium plot.

Another problem can be gauges; if a standard PVT gauge used to span reservoir
pressures is used this may have an appreciable error at low pressures. Low pressure
separator measurements really need a dedicated low pressure gauge. I know this error
applies to a lot of historic data, modern laboratory methods can be much better.

PVT experiments are typically used to tune an EoS rather than provide a direct model in
their own right as they were before computer modelling of fluids. This is fine at reservoir
pressures where the measurements are more reliable; the volumetrics are tight but the
EoS is weaker. At low pressures the opposite holds true; the EoS works very well but the
PVT volumetrics are slack, so tuning the EoS here defeats the purpose of the exercise. If
a separator test gives a liquid composition and density of the last stage I would be
tempted to use that in my model to check the export liquid density prediction, but I
would be wary of accepting the GOR data. The EoS is likely to make a better prediction of
production GOR at equilibrium than a conventional lab separator test. Production
separators seem to achieve equilibrium more easily than a standard lab separator test.

Your question can be looked at from a different perspective. If you have data across all
relevant pressures from reservoir through processing to export and you can predict the
density and GORs of the fluids across the entire process then fine, no problem. If the
model is good apart from at the lowest pressures, then I would ignore the low GOR data
but use prediction. No point in ruining a good model on poor input data points!

Brian.

4 months ago
o Like(2)
o Report spam