"Terminated by solver/No Solution Returned"

[kramea]

Hello,

I installed Veda-FE in a new system and tried to run our model in there. It runs Demo model well with no errors, but with our model, it stops abruptly and gives an error message "Terminated by Solver / No solution returned".

The same model runs without error in the other system (both computers have the same VEDA-FE license).

Is this an issue with GAMS? I have attached the LST file of the run with this.  uploads/64/RES-BAU-INSF.txt

Thank you!

Kalai

[kramea]

Also I keep getting "Missing Element" pop-up error message for certain scenarios (when I start the Case Manager), though they are already present in the model and 'synced'. 

[Gargiulo]

Hi Kalai,

based on the LST file this is the problem:
"LP status(4): unbounded or infeasible Presolve found the problem infeasible or unbounded. Rerunning with presolve turned off. LP status(4): unbounded or infeasible Unable to retrieve Cplex solution 1217.=2"

Please try to reimport all the templates from Scratch in both the computer and run again the model.

Thanks,
Mauri

[kramea]

Thanks for the reply, Mauri.

I tried importing from scratch, even uninstalled and installed it again. The thing is, the PC (the new machine) runs "small models" perfectly. Gives error only when it comes to 'big models'. I ran our base case scenario that we know definitely gives feasible solutions. All these models work fine on our server (that also has the same VEDA version).

Would it have anything to do with disk space/RAM speed? (The PC is 8 GB 64-bit 2.2 GHz).

Thanks,

Kalai

[Gargiulo]

Can we have a GTM? I would like to check the model on both computers

[Gargiulo]

Kalai,

as discussed during the GTM session the problem is infeasible in both the computer if you are running the model with the same CPLEX option (Solution Algorithm: Barrier).

Thanks,
Mauri

[Antti-L]

Kalai:
Thank you for providing me with the DD and RUN files for my investigation. It turned out that the model is indeed infeasible. The infeasibilities are caused by inconsistent "lower bound" availability factors (NCAP_AFA(r,y,p,'LO') and NCAP_AFS(r,y,p,s,'LO')).

However, the model can still be solved to optimum when using an older TIMES version, because the lower bound availability factors became supported only in TIMES v3.1.0, and were simply ignored in older versions.

The model can thus be made again feasible also with the latest TIMES versions by removing all these lower bound availability factors. Then you can start inspecting all these bounds and adding them gradually back if they seem to be correct.  But if the model becomes infeasible again, you will know that the most recent AF(LO) parameters that you added back are being inconsistent.

Hope this helps,
Antti