UofB/CIPR and CMG (Computer Modeling Group, Canada) are collaboration partners, and as such CIPR and the University have academic licenses to all CMG software.This (pseudo)Blog page contains my own experiences when converting an existing ECLIPSE model to IMEX and STARS.Challenge: Convert an ECLIPSE dead oil model to IMEX, and then to STARS.
Original ECLIPSE data file [download]IMEX conversion (data file) [download]STARS 2-component model (data file) [download]STARS 3-component model (data file) [download]IMEX-STARS Match, IMEX data file [download]IMEX-STARS Match, STARS data file [download]Thermal effect, example STARS data file [download]Grid refinement, example STARS data file [download]
Comparison ECLIPSE and IMEX(click pict. to enlarge)Comparison IMEX and STARS initial refcase.Comparison IMEX and STARS final accepted data set, with revised Bo-tables and Rs.Temperature effect:Oil rateTemperature effect:Reservoir pressureTemperature effect:Temperature front at approx center of reservoir
Relative permeability interpolation procedures in STARS have been a challenge -- especially identifying the correct input syntax for the “double interpolation” option, and understanding how it works. This blog entry resolves “everything” concerning interpolation.
This requires an understanding of component mixing that I didn’t have when I started (and probably still don’t have…). The challenge is actually to convert the simple dead-oil description to a component mixture model.In a live oil model more and more gas is liberated from the oil as pressure decreases, so intuitively you could set up a (linear / almost linear) relationship between pressure and degree of mixing.A dead oil model is characterized by a two-regime situation: Reservoir oil is always above bubble point pressure, which means no free gas, amount of dissolved gas in the oil is constant, and doesn’t matter at all until the oil is taken to the surface, and then the liberated gas is constant (per volume). I didn’t and don’t see a straightforward and simple mixing model for this.
After discovering that it was “impossible” to match the results from IMEX/ECLIPSE black oil simulations in STARS it was time for some reflection. Knowing that STARS can simulate “any” physical situation, the question naturally arouse whether the combination of PVT data (Bo vs. pressure) and Rs was possible (valid / legal) at all. What if we attack the problem the other way around – generate Bo and Rs from the STARS data?
Using the matched IMEX-STARS data sets I was now in a position to test if thermal effects are important: Typical reservoir temperatures are in the range 70 - 140 degrees, while injection (sea) water will be 4 - 15 degrees, and hence cool down the oil when they contact.Setting up this case was actually easy in STARS. Enthalpy models and all water behavior was taken from STARS default library.To construct temperature dependent K-values and viscosity parameters I simply used the “Selected Components”-tables from the “Tables” section at the end of the Stars User Guide, plugged in the formula in a spreadsheet, and tested the parameters from the tables until I found a set which matched the IMEX model at 72 degrees. Final tables are found in the data files. Three cases were tested: Reservoir temperatures 72 and 140 degrees, and injection water temperature 8 and 15 degrees.Results are shown to the left, noticeable but not significant differences.Which means we can continue to run these cases isothermally without too large errors.
A significant objection versus the models I’ve studied is that pressure was kept (fairly) constant, so I never got to exercise the interaction between component mixing and pressure variation. This is studied in this blog.
Grid refinement in CMG IMEX/GEM/STARS is relatively simple to define, and several levels of refinement can be defined in a logical manner. The main difference from the ECLIPSE way of doing it, is that in ECLIPSE you define the total number of refined grid blocks in the coarse box, while in CMG you define the refinement for each block.(I.e. to refine blocks i=1-3, j=1, k=1 such that each coarse block is divided into 3 x 3 x 3 blocks, in ECLIPSE you’d set the total number of refined blocks (x-direction 9), while in CMG you set it to 3.The data file includes refinement for the thermal example, with syntax for one and two levels of refinement.The only changes are found in the GRID section and Well perforations.The data files were run in STARS, and results (for both one and two levels of refinement) were identical to the reference case.
In my first attempt at converting a black oil model to compositional (STARS) I was a little too faithful to the STARS manual. E.g., STARS wants to define K-values by polynomial coefficients -- now I’ve found that it’s much easier to use K-value tables. Further, in stead of computing component compressibility and viscosity from the basic formulas I treated them as free variables, determined as matching parameters in the IMEX to STARS conversion.This approach worked much better.