Mixed-Integer

[malin33]

Hi!

Kindly asking for some advise or tips on how to implement a mixed-integer problem in OpEn.
The approach I'm trying is to make the decision variable u an integer with the help of the constraint FiniteSet by:

set=[[1]*N,[2]*N,[3]*N,[4]*N,[5]*N]

bounds=og.constraints.FiniteSet(set)

where N is the length of the prediction.
The problem then is that the solution only consists of a sequence with the same decision variable (for example [1,1,1,1,...]), and does not toggle between values (for example [1,1,1,1,2,2,2,2,...]) which is the behavior I'm looking for.

My question is if the FiniteSet method is the way to go to make the decision variable u an integer or is there another way?

Kind regards
/Marcus

[alphaville]

Hi Marcus. If your decision variable is $(u_0, u_1, ..., u_{N-1})$, then you can use a Cartesian product of Finite sets. You can find an example here.

[malin33]

Hi,
Thanks for the reply, I think I've got it to work using your suggestion!
Including a example if anyone have a similar problem, probably not the best implementation but might be a start

finiteSet=og.constraints.FiniteSet([[2],[3],[4],[5]]) 
#expand to horizon N 
sets=[finiteSet]*self.N
#Cartesan product to get combinations of finite sets
finalSetsConstraints=og.constraints.CartesianProduct(range(0,len(sets)),sets)