I am looking for a way to optimise electric machinery control parameters in order to minimise the electrical grid losses.
The steps involve solving a Newton-Raphson problem and quite some matrix dot products in the way. So it is convenient to treat the problem like a black-box.
The dificulty lies in that in a real life grid there are thousands of generators. In the end I am not looking for a mathematical optimal value but for an order of magnitude optimal value.
Exploring DYCORS and other general purpose genetic algorithms has proven to not to be very efficient, therefore I'd like to know if there are better methods for black-box, high-dimensional optimisation.
It depends... how many dimensions are we talking about? I assume you have no gradient information on your objective function. Are you willing to set for a local optimum or do you need to hunt for multiple local optima (or a global optimum)? Are you considering commercial tools or just open source ones? How long does it take to evaluate the objective function?
My answer is of course incomplete as I don’t know the answers to the questions above. That said, see here for some examples in Python:
http://infinity77.net/global_optimization/
In the past, I also had relatively good experiences with the Matlab code for MCS (https://www.mat.univie.ac.at/~neum/software/mcs/)
There are many, many possible optimization algorithms that may be used for your problem, depending on your needs and answers to the questions I posted.