Why are there no solutions to the model if at least one generator has no max capacity?

[ivyseidel]

I have a pretty complex model based on a real system that I'm running. At least one generator, but in this case as many as 30, have their max capacity set to -1 so the model should theoretically be able to deploy as much of these generators as it needs. This has been working fine, except I just went in and removed some generators that were providing a lot of the load and ran it. This resulted in a math opt interface error.

"LoadError: Result index of attribute MathOptInterface.ObjectiveValue(1) out of bounds. There are currently 0 solution(s) in the model."

This is not the first time this has happened, I've taken it to mean that there is not enough generation available to fulfill the demand set in Load_data. But I don't understand how this is possible if it could just provide from other generation sources. I know this would likely be very inefficient and/ or expensive and not the optimal solution, but it is the best solution for the generation capacity I have provided. So what gives?

I'm not sure if people really use these discussion boards but I thought I'd try. Thanks.

[cfe316]

Hi Ivyseidel (Ivy?),

Can you tell us what version of GenX you're running? That could help track down precisely what's happening.
Also would be helpful if you could say what version of Julia and what solver you're using, and perhaps the settings passed to the solver and GenX settings.

If there's a finite Value of Lost Load in Load_data.csv/Demand_data.csv the model should not fail if it can't meet the demand; it'll just have a high objective value.

I'm not 100% confident about this error message, but it looks like you might have an out-of-bounds objective value. That might be what happens when the objective can become arbitrary low, like if an infinite amount of some subsidized generator operation could be performed. What if you set a finite but very high Max_Cap_MW instead of -1?
Are generators, or their operation, subsidized? In general, this error might occur if some cost value which should be positive is negative.
Can I ask which sort of generators you're modeling? (Especially the ones with -1 for Max_Cap_MW).

I could also be wrong about the error message and there could be some policy constraint which is causing the model to be infeasible. I recently troubleshooted a situation like that; it turns out I was missing input values for some classes generators for the Capacity Reserve Margin policy, so this policy constraint could not be satisfied. If this is the case you might try turning off policies and seeing which combinations of on/off give the error.

[ivyseidel]

Thanks for your response!

I'm running GenX version 3.6 on julia 1.9.3 using Gurobi.

There is a finite value for Voll set in load_data.csv. I don't believe anything is subsidized, though i will try setting something other than -1 for max_cap_mw

I've added a picture of the output I get, as you can see it is "infeasible or unbounded". Also attached are my solver and genx settings files.

Thanks again for your help

settings :

[cfe316]

Version 0.3.6 is good.

Nothing jumps out right away, but you should know that the settings OperationWrapping and LongDurationStorage are not used as of v0.3.5, so you can delete those.

It might be worth turning off the CapacityReserveMargin constraint to see if that does anything.
What categories of generators do you have? i.e. which of
THERM
STOR
VRE
HYDRO
FLEX
MUST_RUN
?

This is a shot in the dark, but it might be possible to add BarHomogeneous: 1 to Gurobi settings.

Can I ask also about the time basis for your simulations? Is it a full year of 8760 hours or something different? I see you're not using TimeDomainReduction.

[cfe316]

Looking at my notes, I once got that Gurobi error ("Model is infeasible or unbounded") when I had a problem which was infeasible.

You said that you had a problem when you removed the generators with unbounded supply, right?
You might check that the remaining resources include firm generators that can supply each Capacity Reserve Margin zone.

It's also worth checking for problems like a resource which can be built but not retired but Existing_Cap_MW exceeds Max_Cap_MW. There can also be problems with storage with Min_Duration/Max_Duration constraints which can't be fulfilled by the New_Build settings and combinations of Min_Cap_MW/Max_Cap_MW/Min_Cap_MWh/Max_Cap_MWh.

[cfe316]

@ivyseidel Were you able to figure out the problem?

[ivyseidel]

Turning off CapacityReserveMargin constraint did not fix the error. I am running 8760 time steps with no TDR.

I have THERM, VRE, STOR, and HYDRO.

I get the error when I remove all the new_build wind turbine options. These were unbounded but there are still as much as 30 generators that have unbounded supply. So my understanding is I've removed the most ideal option but the model won't run the non-ideal option that should be feasible.

Max_duration is only set for my systems battery (STOR) rows which have not been edited between the errored version and the functional one.

[cfe316]

I'm a bit stumped. Would it be possible to share the system via email?

Another thing to check: In the output or log file, do you get messages like [ Info: assuming availability of 1.0 for resource MISO_LRZ8910_conventional_steam_coal_1 ? Are there any generators (VRE or Hydro) that have that message which shouldn't? This would indicate a problem with the names of generators not matching between Generators_data.csv and Generators_variability.csv. This can in some cases lead to numerical issues.

[ivyseidel]

I've figured this out, I mistakenly had a value in larger in one min_cap column that in the max_cap column. Silly me, but thank you for your helpful responses!