[v2026.04] The attached models represent a Geothermal EGS system, which is based on published data for the recent Fervo doublet. The model includes 2 of 20 stages with 6 fracs per stage, which is 1/10th of the total system. The case is single phase (WATER) in the black oil thermal model (i.e. E100 TEMP). The case does have high pore volume and permeability contrasts between the fracture flow conduits and a very low perm background, so this is a demanding system for the solver.
Attached:
- Flow model EGS_doublet1-F.7z
- E100 model EGS_doublet1-E1.7z
EGS_doublet1-F.zip
EGS_doublet1-E1.zip
Comparison Flow and E100
The good news is that these models give essentially identical results in terms of the well head temperature profiles and power output, which also ties roughly to Fervo's published test output of around 5 MW(e), i.e. scaling the system x 10 for total flow rate ~ 50 L/s and subject to reasonable assumptions ~ 15% ORC conversion efficiency https://fervoenergy.com/technology/?playlist=69931c4&video=b664b21.
The Flow and Eclipse models give very close results
However there are some issues for feedback to yourselves that may lead to a few improvements:
Issue 1: Flow is significantly slower run time than E100.
I found that parallel 4 cores gave some speedup, but this reduced with 12 cores (max for my laptop). Below the TCPU comparison for Flow (4 cores) total run time 75 minutes and E100 (single core) 18 minutes.
This is the best performance I have achieved after a considerable time trialling run parameter settings. Below are the non-default settings in this case .param file, from these the weakening of the tolerance-cnv-energy settings seems to be the most influential.
time-step-control-target-iterations=12
time-step-control-target-newton-iterations=5
linear-solver="ilu0"
linear-solver-max-iter=400
edge-weights-method=logtrans
nonlinear-solver="newton"
newton-max-iterations=30
newton-max-relax=0.4
dbhp-max-rel=0.1
tolerance-energy-balance=1e-6
tolerance-energy-balance-relaxed=1e-5
tolerance-cnv-energy=1e-1
tolerance-cnv-energy-relaxed=1e2
max-temperature-change=100
tolerance-wells=1e-2
tolerance-well-control=1e-5
tolerance-pressure-ms-wells=1.0
matrix-add-well-contributions=true
Feedback: can you suggest what types of solver parameters can improve run time in this type of model, are there any recommendations that can be documented to help users with this type of system?
Issue 2. Early time anomalous temperature
In the first few timesteps the producer well temperature starts at around 250 deg C, even though the system is initialised at 190 deg C and there is no reason for a higher flow temperature.
Issue 3. EQUALS expression on MULTNUM not working properly
In the attached models it was necessary to import the MULTNUM result as an array into the Flow case. The EQUALS expression assigns MULTNUM regions to the fractures, cells chosen to give a slightly rounded effect, such that these can be assigned randomised permeabilities representing a range of resulting fracture conductivity.
The EQUALS expression on the MULTNUM array did not work properly in Flow.
EQUALS
MULTNUM 1 6* /
MULTNUM 2 10 10 7 17 5 16 /
. . .
MULTNUM 2 10 10 5 22 6 15 /
. . .
MULTNUM 2 10 10 4 24 7 13 /
. . .
/
This is replaced by using the correct result (i.e. the array generated by E100)
INCLUDE
'DOUBLET1_MULTNUM.GRDECL' /
Issue 4. Compatibility of TEMP keyword
The TEMP keyword is accepted syntax in the RUNSPEC section, but does not replicate the E100 TEMP black oil thermal model, instead results in the error below. It is documented that in Flow the THERMAL keyword should be used, but it is not clear why the distinction has been made and why not just allow the compatible use of TEMP.
No suitable configuration found, valid are Twophase (oilwater, oilgas and gaswater), polymer, solvent, or blackoil.
Issue 5. Additional SUMMARY keywords
This is a request. Please can you add the following SUMMARY keywords for the network (group and field) aggregated flow temperatures.
- FTPCHEA, GTPCHEA
- FTICHEA, GTICHEA
[v2026.04] The attached models represent a Geothermal EGS system, which is based on published data for the recent Fervo doublet. The model includes 2 of 20 stages with 6 fracs per stage, which is 1/10th of the total system. The case is single phase (WATER) in the black oil thermal model (i.e. E100 TEMP). The case does have high pore volume and permeability contrasts between the fracture flow conduits and a very low perm background, so this is a demanding system for the solver.
Attached:
EGS_doublet1-F.zip
EGS_doublet1-E1.zip
Comparison Flow and E100
The good news is that these models give essentially identical results in terms of the well head temperature profiles and power output, which also ties roughly to Fervo's published test output of around 5 MW(e), i.e. scaling the system x 10 for total flow rate ~ 50 L/s and subject to reasonable assumptions ~ 15% ORC conversion efficiency https://fervoenergy.com/technology/?playlist=69931c4&video=b664b21.
The Flow and Eclipse models give very close results
However there are some issues for feedback to yourselves that may lead to a few improvements:
Issue 1: Flow is significantly slower run time than E100.
I found that parallel 4 cores gave some speedup, but this reduced with 12 cores (max for my laptop). Below the TCPU comparison for Flow (4 cores) total run time 75 minutes and E100 (single core) 18 minutes.
This is the best performance I have achieved after a considerable time trialling run parameter settings. Below are the non-default settings in this case .param file, from these the weakening of the tolerance-cnv-energy settings seems to be the most influential.
Feedback: can you suggest what types of solver parameters can improve run time in this type of model, are there any recommendations that can be documented to help users with this type of system?
Issue 2. Early time anomalous temperature
In the first few timesteps the producer well temperature starts at around 250 deg C, even though the system is initialised at 190 deg C and there is no reason for a higher flow temperature.
Issue 3. EQUALS expression on MULTNUM not working properly
In the attached models it was necessary to import the MULTNUM result as an array into the Flow case. The EQUALS expression assigns MULTNUM regions to the fractures, cells chosen to give a slightly rounded effect, such that these can be assigned randomised permeabilities representing a range of resulting fracture conductivity.
The EQUALS expression on the MULTNUM array did not work properly in Flow.
This is replaced by using the correct result (i.e. the array generated by E100)
Issue 4. Compatibility of TEMP keyword
The TEMP keyword is accepted syntax in the RUNSPEC section, but does not replicate the E100 TEMP black oil thermal model, instead results in the error below. It is documented that in Flow the THERMAL keyword should be used, but it is not clear why the distinction has been made and why not just allow the compatible use of TEMP.
Issue 5. Additional SUMMARY keywords
This is a request. Please can you add the following SUMMARY keywords for the network (group and field) aggregated flow temperatures.