Hi. I am solving the wind turbine placement and array cable route design problem which is formulated as a MILP problem.

I am trying to solve the problem using GAMS CPLEX, but the result shows that problem is integer infeasible or unbounded.

I would appreciate it if anyone can help to identify the problem of my model.

Thanks

SC_Jejuwind.gms (3.78 KB)

input_jejuwind.xlsx (409 KB)

If you deal with infeasibilities and MIP then first see if the underlying LP is already infeasible (change model type to rmip). You need to probably solve the model many times with different experiments, so working with LP is much faster (in your case it takes minutes until Cplex says infeasible for the mip while it takes ~1 sec for the rmip). Now you can analyze the actual infeasibility. There are two methods people use: relaxation and iis. I also recommend reading Bruce McCarl’s blog on this: https://www.gams.com/blog/2017/07/misbehaving-model-infeasible/. The iis in your case is large and therefore not so useful (to me). Nothing sticks out (I recommend better equations names then things are often clearer). The relaxation method (I used Cplex option “feasopt 1”) brings it down to just two infeasibilities in equation eq1 for i=138 and i=142.

Good luck.

-Michael

Dear Bussieck,

Could you please give me an example for the term you mentioned " The relaxation method (I used Cplex option “feasopt 1”) ". It would be very helpful for me if you could reproduce it in existing code above.

Thanks !!!

I would study https://www.gams.com/latest/docs/S_CPLEX.html#CPLEX_USING_FEASIBILITY_RELAXATION and perhaps look at example https://www.gams.com/latest/gamslib_ml/libhtml/gamslib_feasopt1.html.

-Michael