Use genetic algorithm for transport solver

[orso82]

No description provided.

[TimSlendebroek]

I like that idea! What about on the stationary plasma actor as I think we want it to be self-consistent?

[orso82]

It's worth a try, though I think in general the stationary plasma actors takes so few iterations to converge that may not be a good target. For sure I'd start with the fluxmatcher.

[TimSlendebroek]

Will it not also be based around the HCD and q/j profile?

[orso82]

Possibly, as an option. But not where we would start. At the beginning, just plain FluxMatcher on the Z's would be enough to see if it makes sense.

The benefit of doing GA would be that the solution will be independent of the initial condition, we'd be able to define bounds and constraints. We can even frame the problem as a multi-objective optimization problem, with one objective for each channel. This will have the benefit of completely sidestepping the issue of normalizing the different transport channels and we could study the tradeoff of matching one channel versus another.