Julia crashing when more than one storage is added about pras

[GordStephen]

There's no theoretical reason multiple devices should take substantially longer to run than a single device, so this is probably a bug.

When adding the second storage device, does the run finish, just slowly, or does it run forever and never return? If the latter, an error is probably getting thrown silently on one (or more likely all) of the worker threads and causing the finalizer to hang, waiting for results that will never be sent.

[GordStephen]

Thanks for the examples, this will help a lot with debugging. I'll try to reproduce the issue and see if I can figure out what's causing it.

[GordStephen]

The singlenode_stor test system includes two storage devices as of last week and runs normally, so the issue may be related to specific parameters in your storage devices? Would you be able to post a minimal example of a SystemModel that experiences the problem?

[juliebuard]

The run runs forever and never return. See below the struct that I am using for storage before creating the StorageAsset. Happy to set up a call too if that could help

@with_kw struct StorageParameters
      name::String
      type::String = "Battery"
      charge_capacity_MW::Int64
      discharge_capacity_MW::Int64
      energy_capacity_MWh::Int64
      charge_efficiency::Float64
      discharge_efficiency::Float64
      carryover_efficiency::Float64 = 1.0
      forced_outage_rate::Float64 = 0.00
      repair_rate::Float64 = 1.0
end

I have 32 generators, 10 being VG as well as 8 storages with capacity of 40MWh on average. Since having 2 generators like that never returns, I tried having a bigger generator, and same here, it is struggling. The only result I got was for 1_000 samples and even with that, I could not replicate.
I am using the Temporal() and Modern()
assess(Modern(samples=1_000), Temporal(), singlenode)

Here are the params for my "aggregated Storages()"

name: String "aggregated_storages"
  type: String "Battery"
  charge_capacity_MW: Int64 1421
  discharge_capacity_MW: Int64 1421
  energy_capacity_MWh: Int64 1180
  charge_efficiency: Float64 0.9486832980505138
  discharge_efficiency: Float64 0.9486832980505138
  carryover_efficiency: Float64 1.0
  forced_outage_rate: Float64 0.0
  repair_rate: Float64 1.0

That I then fit into this asset type: PRASBase.Storages{8760,1,Dates.Hour,PRASBase.MW,PRASBase.MWh}

and here is my system:

ResourceAdequacy.SystemModel{load_params.n_timestamps,1,Hour,ResourceAdequacy.MW,ResourceAdequacy.MWh}(
        assets, storages, emptygenstors,
        DateTime(load_params.datetimes[1]):Hour(1):DateTime(load_params.datetimes[end]), load_params.load_MW)

Thanks.

[juliebuard]

What is interesting is that sometimes with the aggregated generator, I will get a results really fast and without getting stuck but only 1/15 times ...

 4.483385 seconds (5.04 M allocations: 236.421 MiB, 2.25% gc time)
LOLE(result_modern) = LOLE = 10.16±0.10 h/8760h
EUE(result_modern) = EUE = 18300±200 MWh/8760h

[juliebuard]

What is interesting to notice is that the Storage capacityMWh and PowerMW seem to have an impact:
with 4 storages of size (powerMW, capacityMW), it is running but never returns (hanging on some error probably as you suggested above)

B1: (10, 40)
B2: (409, 889)
B3: (30, 120)
B4: (30, 120)

B1, B3, B4 are working individually. B2 is "crashing" / not returning.
It is also working fast if we take all of them but B2. With the below 3 storages, we are getting:

B1: (10, 40)
B2 removed
B3: (30, 120)
B4: (30, 120)

LOLE(result_modern) = LOLE = 60.7±0.2 h/8760h
EUE(result_modern) = EUE = 128900±700 MWh/8760h

Which is coherent with the same system without storage results:

Modern()
EUE = 133000±200 MWh/8760h
LOLE = 62.56±0.07 h/8760h