0

Commingled injection well modeling

Hi,

I am trying to model a commingled injection well which is a well completed through multilayers (5 layers in my model). Every layer in my model is represented by a single grid block in the Z direction. I want to model the injection well such that it is completed through 5 grid blocks. I would like to know if the following input is correct or not.  I have increased the increment between the code numbers the successive elements with identical sink/source to 5.

GENER----1----*----2----*----3----*----4----*----5----*----6----*----7----*----8
A1  1inj 1    5                    COM3      1.00

9 replies

null
    • Keurfon_Luu
    • 4 yrs ago
    • Reported - view

    It's correct, depending on what you are trying to do. In your case, elements "A1  1", "A1  2", "A1  3", "A1  4" and "A1  5" will inject COM3 at a rate of 1 kg/s each. Make sure that these elements are actually the ones in which you want to inject.

    I usually don't really like using the parameter NAD and prefer explicitly listing all the sources.

      • Refaat_G_Hashish
      • 4 yrs ago
      • Reported - view

      Keurfon Luu I am trying to inject in "A1 1", "A2 1", "A3 1", "A4 1", and ''A5 1". In this case, how to modify the input?

      • Keurfon_Luu
      • 4 yrs ago
      • Reported - view

      Refaat G Hashish As I said, you can explicity list all these elements instead.

      GENER----1----*----2----*----3----*----4----*----5----*----6----*----7----*----8
      A1  1inj 1                         COM3       1.00
      A2  1inj 2                         COM3       1.00
      A3  1inj 3                         COM3       1.00
      A4  1inj 4                         COM3       1.00
      A5  1inj 5                         COM3       1.00
      
      • Refaat_G_Hashish
      • 4 yrs ago
      • Reported - view

      Keurfon Luu Thanks!

    • liu_qi
    • 3 mths ago
    • Reported - view

    Hello, I would like to ask about something. For simulating injection into a 5-layer model, is it possible to specify the lithology of the well for the five source/sink grid blocks in the Rock module to achieve well injection simulation? Additionally, the lithology of the well has permeability and porosity several orders of magnitude higher than those of the reservoir caprock. Can this be implemented? I've been having trouble with convergence while trying to achieve this, which makes me question the practicality of this approach.

    • Reservoir Engineer
    • Alfredo_b
    • 3 mths ago
    • Reported - view

    I think it is definetely possible, as it has already been done. An example is given by Giorgis et al., Energy Conversion and Management 48 (2007) 1816–1826. 

    You need to assign a specific domain for the wellbore elements, basically with porosity equal to 1, very high permeability (1.E-6 m2), rock density negligible, rock specific heat negligible, rock conductivity negligible. The distance of wellbore nodes from the wellbore surface should be nearly zero.

    The wellbore elements should have no PCAP. About RELP functions, for sure you should have Kr=1 when the wellbore is filled by the injected phase (I guess a CO2 rich-phase according to COM3). 

    Then you should inject only in the uppermost wellbore element, letting the code to evaluate which is the fluid flow rate injected into each layer. 

    What could be a bit cumbersome is the initial step of CO2 injection: if you start with the well filled with brine, is it likely that you can encounter some convergence problem because TOUGH2 is using Darcy's law even for the the flow across the wellbore elements, when you have two-phase conditions as the CO2 rich-phase is displacing the brine. If you are not interested into short term transients (otherwise you should use T2Well-ECO2N), you may consider to start with the wellbore already filled by CO2. In this way you neglect the initial brine displacement in the wellbore.

    By the way, using the properties of the injected CO2, you can also estimate the value of wellbore vertical permeability to simulate the correct friction losses along the wellbore. This may be important only if you simulate a long wellbore section, otherwise you can neglect the pressure drops due to friction.

    Regards,

    Alfredo

      • liu_qi
      • 3 mths ago
      • Reported - view

       

      Thank you very much for your detailed response, which has been very enlightening for me. I have the following concerns regarding your explanation:

      1、The reservoir permeability is less than 10 mD. If the wellbore permeability is set to 1.E-6 m² and the injection rate is nearly 10 kg/s, what are some good suggestions to help with convergence during the simulation, apart from refining the grid? I’ve tried increasing RE1 and reducing the time step, but convergence remains very difficult.

      2、Why is the fluid only injected into the topmost wellbore element? In reality, the injection port is usually located at the bottom of the reservoir. Is this approach used to evaluate the injection rate? (This is my guess.)

      I would be immensely grateful if you could address my concerns!

      Best regards;

      qi

      • Reservoir Engineer
      • Alfredo_b
      • 3 mths ago
      • Reported - view

       

      1) it's not possible to say why you experience a slow convergence. It may be due to many different reasons. You have to look at the output, identify which is the element and the equation responsible for the non convergence, understand if phase transitions are related to time step reductions, or if some thermodynamic parameter goes out of the admissible range, etc. If necessary, you have to increase the diagnostic output from TOUGH.

      Have you  ran succesfully at least the injection in a 1D radial model without the wellbore? This would help to understand if the input parameters are OK. If you have already convergence issues, first solve them in the 1D problem before move to a complex one. This is also helpuful to understand the processes going on in the near wellbore zone.

      After that, you can run a 2D problem and then the one with the wellbore included.

      A step by step approach is usually helpful.

      2) during CO2 injection the P gradient along the wellbore is lower than in the brine saturated formation. Thus  the P difference is higher at the top of the interval and lower at the bottom. More CO2 will be injected inside the top layers than in the bottom ones.

      Including the wellbore allows to capture this effect having a more realistic evaluation of the actual  injection profile.

      2. Injecting all the CO2 into the top wellbore element allows the code to simulate the right flow rate injected into each layer depending on the local P difference, which changes with depth. If you do not do that,  I do not see reasons to include the wellbore.

      Regards,

      Alfredo

      • liu_qi
      • 3 mths ago
      • Reported - view

       Thank you for your detailed explanation and guidance.i will try follow your help

Content aside

  • 3 mths agoLast active
  • 9Replies
  • 104Views
  • 4 Following