How to use GMP::Row::Deactivate in a MIP Mathematical Program?

Question

I Want to solve a MIP Mathematical Program without some of the constraints by use GMP::Row::Deactivate, but the Program Status return Integer infeasible. However, when I Replace

GMP::Row::Deactivate with GMP::Row::Delete, the Program can be solved. I could’t understand the difference between GMP::Row::Deactivate and GMP::Row::Delete.

My code is as follows:

SCUC_Instance := GMP::Instance::generate(MarketDA_MP);
GMP::Solution::DeleteAll(SCUC_Instance);
SCUC_InstancePre := GMP::Instance::CreatePresolved(SCUC_Instance,"SCUC");
GMP::Solution::DeleteAll(SCUC_InstancePre);
NFGConstraintsSetRowNum := GMP::Instance::GetRowNumbers(SCUC_InstancePre, NFGConstraintsSet);!NFGConstraintsSetRowNum is a Subset of Integer with index as nk
   for nk do
       GMP::Row::Deactivate(SCUC_InstancePre,nk);
   endfor;
endif;
NFGVariablesSetColumnNum := GMP::Instance::GetColumnNumbers(SCUC_InstancePre, NFGVariablesSet);!NFGVariablesSetColumnNum is a Subset of Integer with index as nv
   for nv do
       GMP::Column::Freeze(SCUC_InstancePre,nv,0);
   endfor;
endif;
UCSolver := GMP::Instance::CreateSolverSession(SCUC_InstancePre , "UCSolver" , 'CPLEX 12.3');
GMP::SolverSession::Execute(UCSolver);

Do you know how can I solve a Mathematical Program when the constraints do not take effect but remains in the instance?

Marcel Hunting · Accepted Answer

Hi @denghaiying.GMP::Row::Deactivate basically changes a constraintinto a constraint with type <= and a right-hand-side value of 1e20. So theconstraint x + y >= 20would become x + y <= 1e20after deactivating it.Deactivated constraintsare not printed in the constraint listing but they will be printed in the LP file by CPLEX if the CPLEX option 'LP file’ is set. You could print the LP file once whileyou useGMP::Row::Deactivate, and once when you comment outGMP::Row::Deactivate, and then compare the LP files using, e.g., WinDiff or WinMerge.Normally theright-hand-side value of 1e20 should be high enough but I can create an(extreme) example for which deactivating a constraint will make the model infeasible: x + 1e10 * y >= 1000 y >= 1e11 x >= 0Deactivating the first constraint will turn it into x + 1e10 * y <= 1e20And becausey >= 1e11 and x >= 0, the left-hand-side must be at least 1e21 and therefore cannot satisfy <= 1e20. Note that this example has severe scaling issues.Note that onecan rewrite the first constraint to -x -1e10 * y <= -1000and then deactivating it works fine. So, if the constraint that you are deactivating is a >= constraint then you could reformulate it as a <= constraint to see whether that resolves the issue. (And if this works then itis a strong indication that your model has scaling issues.)

Reply

Sign up

Login to the community

Scanning file for viruses.

This file cannot be downloaded