Page 3 of 10 FirstFirst 1 2 3 4 5 ... LastLast
Results 25 to 36 of 118
  1. #25
    Very helpful you guys....you guys are all aces in my book.


    See More: Reservoir Simulation and History Matching

  2. # ADS
    Spons Circuit
    Join Date
    Always
    Posts
    Many
     
  3. #26

    Join Date
    Nov 2008
    Location
    Australia
    Posts
    508
    I've had some very good results with the Roxar EnABLE package in conjunction with Eclipse, having used it on a dozen or so projects. Great tool to assess sensitivities and (a) drill down to a number of good matches or (b) establish a probabilistic distribution of predictions

    I've never used SimOPT (I believe it is more aimed at optimizing new drill locations), have heard poor things about Cougar and two colleagues of mine have had average/poor results with MEPO

    Just my two cents worth.

    VM

  4.    Sponsored Links



    -

  5. #27

    Join Date
    Jan 2009
    Location
    ALA, KZ
    Posts
    435
    Shakespear, so 1 version of my assisted history matching program is ready
    the results are
    model 900 k active cells one run on 8 core station 4 hours
    243 wells
    oil
    gas
    Vapoil
    Water
    It tooks me a week to set up 98 percent of wells in the range of +-10 % of history data in terms of total production of liquid oil and gas and +- 20 % of history bhp
    The algo consist of following steps
    1. check well for Liquid total
    2. check for BHP
    3. check Water cut
    4. check for Gor
    The main thing i change is multx,multy, myltz in schedule section and this worries me
    What you think of this ??

  6. #28

    Join Date
    Nov 2008
    Location
    Australia
    Posts
    508
    temr,

    Am interested to know what sort of algorithm you use - care to elaborate?

    Is it something along the lines of an algorithm that contours observed pressure points in time, from this map then calculating the pressure gradients in X and Y directions - doing the same for simulated pressures and dividing one gradient map by the other to figure out the difference and making appropriate MULTx corrections to try to better match?

    If so, I've dabbled in this before, and generally found that (unless the geology is quite simple) I ran the significant risk of missing flags pointing to changes needed to the depositional framework - instead 'fixing' with sometimes unrealistic multipliers. Additionally, if your HC saturations are quite dependent upon perm, then changing PERM (I know you are changing MULTx arrays, but unless you have some other process like fractures/streaks etc you would really need to multiply through your PERMx arrays - at least for those mults greater than 1, for those less than 1 you can write off to baffling) messes with the material balance, often resulting in an iterative process that had problems converging to a satisfactory grid.

    Suggest that you multiply your final MULTx with your PERMx's - and see whether perms (and PORO's - use crossplot) that result are realistic - get with your geologist and see if they buy into it.... Likewise calculate new HC saturations based on the new PERMx xMULTx's and re-run to see what it did with your material balance. Sometimes it worked for me, many times it gave me a PERM grid that had many (unrealistic) bullseyes in it..

    Good luck!
    Last edited by vinomarky; 08-15-2010 at 02:40 AM.

  7. #29

    Join Date
    Jan 2009
    Location
    ALA, KZ
    Posts
    435
    vinomarky
    thanks for you advise
    additional geologist is to expensive for me , so i am doing everything by my self form core to simulation.
    As for algorithm it is quite simple
    basically what i am doing firstly is analysis of reason of water and gas break for each well. The process have two steps. At first step program analyze the wcut and gor ratio by well and choose candidates(slaves) and the possible reason(hosts) the candidate mismatch history (injector , aqufer , gas cap etc) and prepare decision matrix for me , secondly i check all candidates manually and approve the list with hosts and slaves.
    On second step program create several realization of multx arrays and run all cases in eclipse . After all runs program create sensitivity analysis for each candidate and i use this data to make final multx array

  8.    Spons.


  9. #30

    Join Date
    Nov 2008
    Location
    Australia
    Posts
    508
    Sounds interesting - I can't quite picture the full workflow, but interesting nonetheless....

    If you are inclined, can you further elaborate on what logic/algorithm you are using to drive your MULTX array realizations? Do they converge as you get towards a solution, or is it a stochastic exercise that through experimental design type approach you eventually find something that works?

  10. #31

    Join Date
    Jan 2009
    Location
    ALA, KZ
    Posts
    435
    The program uses fuzzy logic algos to select candidates - and that is hardest part to implement
    ones i found candidate and reason
    the program "draw" the multx path from the candidate to the source
    using box keyword in schedule section
    and the realizations are quite simple min max and average value for the sake of sensitivity
    like this.....

  11. #32

    Join Date
    Jan 2009
    Location
    ALA, KZ
    Posts
    435

  12. #33

    Join Date
    Jan 2009
    Location
    ALA, KZ
    Posts
    435

  13. #34

    Join Date
    Nov 2008
    Location
    Australia
    Posts
    508
    Ok, understand better what you are doing now.

    I think that you'll find though that while it may deliver a solution, it will in all likelihood not be very predictive.... Unless you actually think there are channels between inj/prj pairs (in which case you should probably stochastically incorporate others as well) the solution will not represent the geology of what is driving the observations.

    When people take the model - nodding at how well it matches history - their next step is to say "ok, now where shall we drill the next well" or "when will this field go uneconomic".... By tuning the pathways between wells only, you will not address the other 80% of the reservoir, so it will be poor at adressing these questions... In the image above - without knowing the geology - I would say there is a reasonable likelihood that instead of a unrealistic looking high perm channel between inj/prd pairs you may instead have a thin high perm layer in the area... the two scenarios will result in significantly different outcomes should you drill a well (for example) between well 611 and well 620... Your channel scenario will show reasonable water free production for quite a while, while a regional high-perm streak will result in early breakthrough again..... This just underscores the importance of having appropriate geology representation.

    As an interim step, perhaps you could try using your multiple MULTX paths as the framework to calculate a MULTX map, that is at least continous through the field? This wont address the layering issue, but if you find that you have problems matching with continuous mult changes, then it may point the finger toward requiring instead vertical heterogeneity changes (high perm streaks etc)

    This should be fairly easy (I've done similar in the past) - In Petrel you'd create a number of straight line polygons between your inj/prd pairs at different depths (where the depth = MULTX Value), then use the 'make simple surface' operation. You may have to append all the polygons into one group or convert them to a points set first though. Ensure you appropriately define your boundary condition limits (ie setup some phantom MULTX values = 1.0 around the structure perimeter, or limit max/min values ion the surface creation step) so that it does not extrapolate silly numbers away from control points. You can then create a MULTX property grid = surface values (one of the operations available on each model property)

    Good luck!
    Last edited by vinomarky; 08-18-2010 at 05:51 AM.

  14. #35

    Join Date
    Jan 2009
    Location
    ALA, KZ
    Posts
    435
    Thankx for comments, but
    Do not agree with you for several reasons.
    Firstly the reservoir you looking for is low permeability ( 1- 10 md) and it carbonates ( my program choose methods dependently on type of rock and permeability , for example there is no sense to make channel in sand reservoir with permeability of 10 Darcy - solution in this case would be good upscaling with unregular cell thickness)
    Secondly this well is in oil zone - there is no water to come unless its come from bad completion or injector - that is why i use fuzzy logic algos when program prepare draft analysis also program pays attention to BPH , for example in this injector model BHP more than existing fracturing BHP so this case is subject to fracture , and i take decision two "draw" the path only if i have certain triggers active
    As for approach of forecasting , using standard approaches you will get oil between this to wells and will put the project well to the existing water zone, my task is to set up the reservoir to the last day of history matching with maximum reliability, but if i will drill 10 more wells i will remake the model , because in geological model only FWL most in any cases is the same everything else is subject to change
    As for multx , i do not use petrel as workflows they are extremely slow - i use my program which makes multx paths to any well and prepares schedule section
    Regards

  15.    Spons.


  16. #36

    Join Date
    Nov 2008
    Location
    Australia
    Posts
    508
    My point was simply if you arrive at a matched model that is consistent with geological understanding, then you stand a far better chance of being reasonably predictive. Conversely, if you match through making non geologically realistic changes, then your model is far less likely to be useful as a predictive tool.

  •   

Similar Threads

  1. Replies: 0
    Last Post: 01-05-2012, 10:12 PM
  2. Replies: 0
    Last Post: 11-15-2010, 03:00 AM
  3. Replies: 2
    Last Post: 11-14-2010, 11:17 AM
  4. Replies: 4
    Last Post: 09-08-2010, 11:55 PM
  5. Article: Reservoir Simulation and History Matching
    By Shakespear in forum Home Articles
    Replies: 1
    Last Post: 06-19-2010, 04:08 PM

Tags for this Thread

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •