Solution is available, but not proved optimal

[saleh]

Hi,

When I run the model I get this error:

"Solution is available, but not proved optimal, due to numeric difficulties during optimization", I wonder about the cause of this problem.

Thanks,

Saleh

[Antti-L]

I don't think this has anything to do with VEDA. As far as I know, VEDA does not give any messages like this, and so I am rather confident that VEDA is not guilty of this problem.

I have never seen this message myself, but searching a bit on the Internet,  I found some discussions about Cplex producing exactly that message when returning the status code CPX_STAT_NUM_BEST.

So it seems the message may be produced by Cplex. If so, maybe consider using Gurobi instead?

[Antti-L]

If you want to know the exact reason for the numerical problems of the solver you are using, I am afraid there is no general answer. Numerical problems can be caused by a wide variety of conditions which may occur more or less randomly during the solution algorithm.

See a general discussion about handling numerical difficulties here (with focus on Cplex):
https://www.ibm.com/support/knowledgecen...culty.html

A few additional recommendations for avoiding numerical problems can be given specifically in relation to TIMES:

• Check the QA_check.log for errors and warnings, and at least eliminate any free output flows if such are listed there;


• Disable any “dummy import” processes if such have been used in your model (use e.g. PRC_NOFF to remove them from the model);


• If you have been using very high costs for any other processes in order to prevent them entering the solution, remove such processes altogether e.g. by using PRC_NOFF;
• Check for any very high upper bounds specified and consider reducing them or removing them altogether if they are not necessary;


• Check all your market share and user constraints for mutually redundant equations, and remove if such found;


• Check for any zero-cost commodity flows that could become arbitrarily large, and add either small costs or reasonable bounds for them if such are found.