Not getting lowest discounted system cost

[cyang]

I am seeing some very strange behavior from my small H2 model.

In a specific scenario, I know that H2 pipelines are cheaper than H2 trucks, but when I run without any constraints, the model results show only truck delivery.  If I include a user constraint that prohibits the use of trucks, then I see the model build pipelines, but the strange thing is that the total system discounted cost (ObjZ) is higher in the unconstrained case and lower when prohibiting the use of truck.  In the screenshot posted here - "NT" refers to the constraint with No Trucks. 



My model includes discrete investment and I'm guessing that the problem could have something to do with the CPLEX settings.  (I think Mauri once mentioned to me that some settings might speed up solution time but the answer might be more approximate).  In a previous post I had mentioned that I wanted to speed things up, but since I'm running a highly simplified model right now (just analyzing the H2 infrastructure) the model only takes 1-2 seconds to run, so I'm happy to have it take longer but obviously be more accurate.  Which settings might help make the solution more accurate?  Or could there be some other issue with the model?



Any help would be greatly appreciate!
thanks!

[Antti-L]

If you have not yet done so, check at least the OPTCR setting in the control panel. It controls the MIP optimality tolerance. See the GAMS documentation for details. 

[cyang]

Thanks Antti.  I checked the OPTCR settings and it was blank.  I believe that means a tolerance of .1 or 10%.  I set it to be .01 (i.e. 1%) and got pipelines!  It made the solution time over 6 sec, but the answer seems to be more accurate and result in lower costs. 

thanks again!