0

I have a MINLP problem. I can solve it quite fast (seconds) for small instances using SCIP. However, in big instances, it takes a long long time (hours) in getting to a feasible solution. I wanted to know if anyone here has some tips on how to reach to a fast solution faster.

I can't know beforehand any initial solution, and setting the heuristic to be more aggressive or fast or null has not helped.

orpanter
  • 115
  • 6
  • This question is too general; the answer depends upon the specific MINLP problem. You will need to find a better formulation or use a custom propagator or brancher. – Brannon Aug 25 '22 at 17:18

1 Answers1

1

It may help to make the model more elastic. I.e. allow (some) constraints to be violated at a cost. Often that makes economic sense (i.e. renting extra capacity).

Erwin Kalvelagen
  • 15,677
  • 2
  • 14
  • 39