Eclipse Convergence problems

[trans_I]

I am running a carbonate model single porosity single permeability with 775000 cells. My porosity value ranges from 0 to 42 %. Since i don't have permeability values for my model i used a Timur type transform to populate permeability. But the populated permeability is not sufficient to make fluid flow to the wells and a lot more oil rate cut results. As soon as i put some multiplier to the permeability (Global) to make fluid flow across, convergence problem arises and make my model to run very slow (47 hours). I have nearly 100 wells with 10 multilateral and 28 years of history. I am looking for a way out to overcome the problem and make the model run faster.

Moreover when i see my wells, deviation data and perforations in PETREL it seems nice and logical. But after running when i see the SMSPEC in flow viz my wells seem weird/coiled zigzag particularly the multilaterals. Can these be a cause to the problem?

Pl suggest.

[DAH7542]

Dear trans_I, some observations:

1. Timur's correlation holds for sandstone reservoirs. I wouldn't use it for carbonates. Look for an appropiate correlation for carbonates.
2. Insufficient flow rate with lower than observed water cut might be due to several things, including flow through fracture network (not considered in a single porosity single perm model like yours). If there is indeed flow to high perm (low pore vol) features such as fractures, you will be forced to use large numbers for perm or transmissibility multipliers.
Are you sure that single porosity single perm model is ok in your case?
3. As for the convergence problems, I suggest you first identify the source of the problem. To do it, use NEWTON=2 in RPTSCHED. Read carefully the report at each time step while the simulation is still running and identify the problematic cells (those cells with the highest errors). My first impression is that when you increase the global perm by means of the global multiplier the throughput stability condition is not met, this happens when fluid flows throuhg high perm low pore volume cells. If this is the case, you will drastically reduce the run time by increasing MINPV. My suggestion: increase MINPV to the maximum allowable value in your case, run the simulation and compare the run time with the original run.
4. As for the weird trajectories in FloViz, don't worry, this is normal and it's not the cause of your convergence problems. FloViz displays the wells as straight segments between connected cells, that's why you see ziz-zag trajectories. If you want FloViz to display the actual well deviation (as seen in Petrel) you have to load the well deviations by loading the *.dev and *.cnt files into FloViz.

Good luck.

[trans_I]

Dear trans_I, some observations:

1. Timur's correlation holds for sandstone reservoirs. I wouldn't use it for carbonates. Look for an appropiate correlation for carbonates.
2. Insufficient flow rate with lower than observed water cut might be due to several things, including flow through fracture network (not considered in a single porosity single perm model like yours). If there is indeed flow to high perm (low pore vol) features such as fractures, you will be forced to use large numbers for perm or transmissibility multipliers.
Are you sure that single porosity single perm model is ok in your case?
3. As for the convergence problems, I suggest you first identify the source of the problem. To do it, use NEWTON=2 in RPTSCHED. Read carefully the report at each time step while the simulation is still running and identify the problematic cells (those cells with the highest errors). My first impression is that when you increase the global perm by means of the global multiplier the throughput stability condition is not met, this happens when fluid flows throuhg high perm low pore volume cells. If this is the case, you will drastically reduce the run time by increasing MINPV. My suggestion: increase MINPV to the maximum allowable value in your case, run the simulation and compare the run time with the original run.
4. As for the weird trajectories in FloViz, don't worry, this is normal and it's not the cause of your convergence problems. FloViz displays the wells as straight segments between connected cells, that's why you see ziz-zag trajectories. If you want FloViz to display the actual well deviation (as seen in Petrel) you have to load the well deviations by loading the *.dev and *.cnt files into FloViz.

Good luck.

Thank you very much for the reply DAH7542. I am sorry i did not checked my mail for long. As you mentioned it was throughput related problem only. The NTG in the static model was 0 and 1 only but while up-scalling it gave me a range of NTG starting from 0 to 1. May be that was causing me problem. I put some cut-off value for NTG and increased MINPV also. I used porosity beaning and than populated permeability with some logical transform that finally worked a bit good to my situation.

[temr]

The rule is simple the higher the permeability the more problems you will have. So try to make the perm range in normal way, the perm for card is in the range to 40-60 mD , also try to smooth you perm array to avaid rapid changes in perm

[saber.sassi]

I am looking for these document:

1.G. W. Davies, I. J. A. Gamble, John Heaviside:"Field-Wide Variations in Residual Oil Saturation in a North Sea Sandstone Reservoir" SPE 19851, 1989
2.Land, C.S. "Comparison of Calculated with Experimental Imbibition Relative Permeability." SPE 3360

thank you in advance