Originally posted by: NicoGDFSUEZ
I am using Opl/Cplex in a simulation context using the opl .NET api.
I am trying to do the following thing :
-
Generate and solve a first LP problem
-
Saving the basis of the optimal solution
-
Generating a second LP problem where the only difference with the first problem is the coefficients of the objective function
-
Solving the second problem using the basis of the precedent optimization
The performance of the second solve are very bad : it would be more efficient to run a new complete optimization not using warm start !
I tried different sets of parameters (dual and primal simplex, AdvInd = 1 or 2) but I always obtain the same problem.
Which is very surprising is that previously, I did the same kind of thing after modifying the bounds of some constraints and adding/deleting some (few) constraints : in this case, the warm start worked well !
If I remember correctly my old OR lessons, the warm start would work perfectly if I am only modifying the objective function or/and the bounds of the constraints.
Is there a better set of parameter for Opl/Cplex to increase the performance ? Has someone a better idea to improve the computation time ?
#DecisionOptimization#OPLusingCPLEXOptimizer