I am using GAMS for solving a MILP problem which includes binary variables. However there is a problem in the solution. Surprisingly, I have seen that one of the binary variables in the solution has a found value of "-1" another one "2". That is not acceptable and logical. I do not know what happened. GAMS gives me the message primal infeasible."resolved with fixed discrete variables"
code has attached.
Dear Iman
I guess that you get that inconsistency because the problem is not feasible, as the software is telling you
Try by changing some values and try also the dual of the same problem
@dear Nolberto
actually when I use BENDERS DECOMPOSITION,problem is solved because it is solved by dual of subproblem,however recently I faced this problem. moreover,the problem is solved for small case only when I increase size of case ,this happened.Before that I solved it without any problem,something is wrong is in GAMS.
1. Compare in detail if your mathematical model in GAMS is identical to what you have written down "on paper".
2. Further - could it be that the subproblem is not always feasible, depending on how you choose the values of the binary variables? One way of solving that problem is to add additional (and VERY expensive) variables in the right-hand side of the "less-than" constraints, so that you know that there is a feasible solution.
Your results of course depend on the solver you used. There is nothing wrong with GAMS as far as I see - have a look at the attached files, where I solved your model with BARON as the MINLP solver framework, giving it an options file to have a smaller relative gap than the default 0.01.
You'll see in the lst file that the binary variables take binary values in the solution.
So what solver did you use? Maybe the solver is buggy...these are done by humans, too. To err is human ;-)
@Ralf
I used BONMIN solver MINLP,GAMS distributor version 24.1,is it possible depond to version of GAMS or maybe because I have overload many times GAMS ?
I am just trying to solve with BONMIN, which takes much longer than with BARON - and it produces some infeasible problems on the way even with GAMS 24.7.4
NLP0014I 230 OPT 3905.6402 113 8.673162
NLP0014I 231 OPT 3905.7 49 1.310505
NLP0014I 232 OPT 3905.6402 78 2.198324
NLP0014I 233 INFEAS 3.6958136e-05 168 4.13332
NLP0014I 234 OPT 3905.6408 78 2.433217
NLP0014I 235 INFEAS 3.6958136e-05 90 2.247263
NLP0014I 236 OPT 3905.6405 57 1.64453
....
Cbc0020I Exiting on maximum time
Cbc0005I Partial search - best objective 1e+50 (best possible -1.7976931e+308), took 12131 iterations and 28 nodes (1006.55 seconds)
Cbc0032I Strong branching done 108 times (14791 iterations), fathomed 0 nodes and fixed 5 variables
Cbc0035I Maximum depth 14, 0 variables fixed on reduced cost
NLP0014I 249 INFEAS 96 141 4.464136
Bonmin finished. No feasible solution found.
A lot of the GAMS solvers are academic code...
At least it seems that BONMIN is not the best solver choice for this MINLP. Do you have other licensed MINLP solvers? (your model is too big for demo licenses)
Try those...
for instance
gams Untitled_48.gms MINLP=SCIP optfile=1
with the attached files. It solves the problem - and is MUCH faster than BONMIN with its fruitless run.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Control Theory
- Control Systems