Well tests in wells with several layers

[marat18]

Hi all!

I don't know if it is an old problem and all already known and studied, but I work for a small company that does various studies for E&P companies and most of the times when we inquire information about field pressure, the only thing we get is a bunch of excel files with well-test results.
Most of the time those tests are of the poor quality and performed on the wells with 2-3 or even more layers explorated simulatenously.
They try to do a good old Horner plot etc to interpret this data, but it rarely works.
So I wonder, is there any software that can interpret these non-ideal data and give some more or less reliable values for field pressure for each layer?
And if not, then how do you think it can be solved?
I'm thinking about setting up Eclipse or Tempest model with several layers and different properties, then performing a well test and may be I'll see that the pressure plot will bend at some point or anything.

Thanks everyone

[dybydx]

I think software that you mention above for interpretation and predict data of each formation(right? i guess) might be available in reservoir engineer.

[vinomarky]

Welcome to the real world of imperfect and uncertain data

No, there is no magical software package that will auto-deconvolve multi-layer tests into unique single layer results. You will have to incorporate more information, some engineering judgement and accept the the result will still be uncertain

Additional information to incorporate;
Well logs - as a minimum thickness's of the various sands. Ideally the E&P companies have core and/or mobility tests, which have been used to tune well log analyses to (gu)estimate perms of those sands. In this case, you would be best served setting up a simple model to reflect the layered architectural, and history match while controlling the range of perms depending upon how confident you are of them (perhaps +/- 50%?). Note, if you are using numerical simulation to do this, understand that Peaceman well connection factor is based upon steady state equations, meaning you may need to significantly fine-up your gridding around the wellbore to represent transient effects well enough in your relatively short term build up tests.

Other information to consider;
Field wide information - geologically, are your sand units pervasive? High perm? relatively homogeneous?
If yes to all, then you will be looking for similar perms AND P*'s in all tests of similar vintage in a given layer (these questions may be true for one layer, but not necessarily all - in which case, it is still helping your understanding). If not, perhaps look for wells close to each other
Also, if looking at high kh contrast wells, think through the likely ranges and understand that kh as a measure of deliverability and Phi.h as a measure of storativity is important. The relative ratio of kh/Phi.h will generally control the degree of depletion seen in each layer (assuming no vertical connection in the reservoir, or significant aquifer support). The early extrapolated P* will probably be closer to the highest kh layer, while late time P* will point closer to the lower kh layer(s)

Essentially, you have more degrees of freedom than data points to constrain with your single well test results, and you will need to look for ways to collapse these freedoms through application of engineering understanding. Remember, an engineer who can only solve the obvious problems (data in + equation = answer) is really just an overpaid technical assistant - use your understanding to extrapolate, group, and selectively simplify.

You may well STILL not have enough information to reduce the levels of uncertainty enough, if which case, dont be shy to recommend additional single sand tests - is the client doing some infill wells? Then another test in an upcoming drill may be easy. It comes down to what are the implications of the questions you are trying to answer (how important is that reduced uncertainty?) If valuable, then additional tests should be easy to justify. If not, then accept the uncertainty - characterize it, and show the implications of the bookend assumptions along with a 'most likely', and plan for flexibility.

[for_lon]

Certainly, there is no software to magically solve this problem. Even today at 2014, we see a lot of operators developing fields under these old bad practice, producing in commingled with no enough data acquisition, and then asking for answer to their problems. The best way to try find a solution to this is to recommend a field data acquisition campaign running PLTs, BU for individual layers, falloffs for individual layers, sidewall cores analysis after new drilling, pressure points during logging new wells (XPT, SFT, MDT, etc) and all possible data and then start to create new models based on this data trying to recreate old bypassed information. Easy to say, hard to do.

[slhero99]

for well test, esp. well test interpretation, i think it's better do not using multi-layer modeling, if using multilayer modeling, it will be very difficult to get good match. I think just because the well test theroy is zero dimension modeling.

[temr]

The good software to interpret any scatter data is excel
The tool for interpretation is expirience and knowledge
As advise use error corridor for your data if we speak about MDT - gradiend of water and oil ( at sertain reservoir conditions)