How to save conversion time and not loose PP performance

[DavidEnergie]

Dear pandapower developers and community,

which is the best way (fast and robust) to repetitively run power flows on large (mostly radial) distribution networks, while changing network parameters (branch impedance values and nodal load/injection)?

In Thurner et al. 2018, it is mentioned that the internal BBM (Bus-Branch-Model) and admittance matrix is kept for the repetitive (time-series) power flows, while only nodal injections are changed. Though, I additionally want to change the branch impedances (and maybe later connectivity/switch state), while saving on the conversion time and keeping as much of the good performance and convenience of pandapower (e.g. numba acceleration).

I mostly need the repetitive runs for feasibility checks, the exact results and the difference between them is not so relevant (at least so far).

After converting from net to ppc via pp._pd2ppc, I am thinking to manipulate the admittance matrix in the ppci (or ppc?) data structures directly and for successful pf executions, convert them back (ppci + results) via _ppci_to_net(result, net).

Is there an approach you could suggest? And what considerations should be taken, in your opinion?

I would be happy for any pointers.

Best wishes