sandialabs / snl-swan Goto Github PK

Hello

I'm testing the SNL-SWAN for some WEC farms. In this case they are farms of point absorbers.

From my initial test is possible to see that the correct definition of the obstacle in SNL-SWAN is the most important aspect. In the case of a point absorber, the test results show that there is about 50% less absorption from a wave heading of 330º than a wave heading from 270º .

This certainly as to do with the design of the WEC, is this case I followed the tutorials and the WEC is just a straight line.

Is there a better way to define a point absorber in SNL-SWAN?

Best regards
João Ribeiro

Dear SNL-SWAN users and developers,
I was running SNL-SWAN in a coastal area of Portugal in the presence of a wave farm and I realised that OBCASE 4 the devices were not absorbing energy (the POWER_ABS.OUT was zero for all of them).
I tried to run the model with the input files from your tutorial and the same is happening, OBCASE 4 is not working. Can anyone help me understanding why?
I would really appreciate any help you can provide.
Best Regards,
Valeria Fanti

Hi

I found that de DEALLOCATE(TMP) in line 1220 of SNL-SWAN is give a error when compiling, the workaround for me was:
IF (ASSOCIATED(TMP)) DEALLOCATE(TMP).

Also, I'm trying to run a large area and for that I need MPI. Those SNL-SWAN work with MPI?

Best regards
João

This is a binary WEC directional term which only extracts energy from certain incident wave directions.

This code mod is similar to obcase 3, but instead of taking the overall Hs from the entire spectrum, exacts energy content from the waves based on the binned frequency and wave height.

Element zero is being accessed in this array because of indexing to k-1.

SwanBpntlist.ftn90: allocate(blistot(nbptot))

I believe that this might be the solution, but it might require special code instead:

SwanBpntlist.ftn90: allocate(blistot(0:nbptot))

—————————————————

I believe that this line needs added:

SwanCompUnstruc.f90: !$omp private(kteta) &

Because KTETA is written and read by different threads and is specific to the work being done.

—————————————————

This !$omp master section in SwanCompUnstruc.ftn90 needs an !$omp barrier before it

   inocnt = 0
   !
   !$omp master
   inocnv = 0
   !
   islmin = 9999

Because shared variables are reset in it.

I tried the example file found in the repository. Obcase 1 works fine however when I tried Obcase 3 the obstacle vanishes. Not able to understand why the varying transmission coefficient with power file is not working. Should I try to run the simulation in a different way?

Hello everyone,

When I try to run the Tutorial or any other INPUT with any option of obcase I get this error:

SWAN is preparing computation

At line 1220 of file swanpre1.f
Fortran runtime error: Attempt to DEALLOCATE unallocated 'tmp'

Error termination. Backtrace:
#0 0x7f08dcead32a
#1 0x7f08dceaded5
#2 0x7f08dceae2a7
#3 0x5623c287e2de
#4 0x5623c286ce9d
#5 0x5623c286ebef
#6 0x7f08dc4f5b96
#7 0x5623c281eff9
#8 0xffffffffffffffff_

I can run the INPUT if it does not have the line OBSTACLE. But if the INPUT has the line OBSTACLE it crashes and I get the error.

Does anyone has the same problem?
I would appreciate any help.

Thank you and regards,
Melissa Jaramillo

This code:
integer, dimension(10) :: nacvt ! number of active vertices for i-th thread
in
SwanThreadBounds.ftn90
means that all results for more than 10 OpenMP threads are wrong. Please change to:

integer, dimension(:), allocatable :: nacvt ! number of active vertices for i-th thread
.........
allocate(nacvt(nth))
.........
deallocate(nacvt)

Thank you.

This is a feature which allows the WEC module to reflect and extract energy as a function of the incident wave frequency.