Number of iterations increases with successive optimizations

Dear all,

I am running an optimization problem (multi-period DC-optimal power flow). And when using set_normalized_rhs function to modify the constant term on the right-hand-side of the concerned constraints and re-optimizing, the number of iterations (and the computational time to solve the optimization) increases directly starting from the second optimization. From the second optimization, the number of iterations is relatively constant but much greater than the first one, i.e. the first optimization is solved in 7 seconds (+/- 30 iterations) and all the following optimizations are solved in 200 seconds approximately (+/- 150 000 iterations).
I cannot explain this, if anyone has faced the same problem or has an idea to solve this problem, please let me know.

Do you have a reproducible example?

A first guess: when you modify the RHS, you probably perturb primal feasibility, that is your solution may become infeasible from one solve to the next. Depending on the method, it may interfere with the optimization process. What happens if you solve all problems starting from the same initial point?
Usually, you want to use a dual algorithm (e.g. the dual simplex for linear programming) because perturbing primal feasibility doesn’t impact dual feasibility, so the solution to the previous solve is still dual feasible when the RHS changes.

1 Like