Different objective function value given by ipopt and YALMIP

[KosMy]

I have a problem when I am trying to solve an optimization problem with IPOPT. The output of IPOPT shows a value for the objective function, but when I try to extract this value from YALMIP I am getting another value. Any ideas of why is this happening?

[Johan Löfberg]

impossible to answer without reproducible code, or at least more information

[KosMy]

What kind of information would be of help to find out why this is happening?

[Johan Löfberg]

The code?

Or at least a definition of what "another value" means

[Johan Löfberg]

I can reproduce it, but this takes forever to run making it horrible to debug. Can you supply a smaller reproducible example

[Johan Löfberg]

THeproblem is the use of the convexity-aware operator abs which will map to a norm expression, which will be modelled using an epigraph, but ipopt does not have any support for the non-smooth norm operator, and thus yalmip has some internal hacks to approximate it, and all those small approximation errors escalates to a large difference in the actual objective compared to the epigraph-modelled objective.

Better to derive the quadratic model

z = Model.Iline{index};

z = real(z).^2 + imag(z).^2;

Model.losses= Model.losses + 1*Parameters.Ccurt*sum(z.*Network.branch(:,3)/Network.baseMVA);