Issue with PV-PQ-change in combination with voltage dependent loads

[midrantos]

Hello everybody,

I encountered a problem with the current implementation of the PV-PQ-change of generators when using voltage-dependent loads.

In the original runpf(), in this case a generator is fixed to the reactive power limit by converting its [Pg,Qg_lim] to a negative load, i.e. -[Pg,Qg_lim] are stored in mpc.bus. That works fine, if no voltage dependency is considered. However, when using this, the reactive power injection is changed accordingly to a load, which is not correct.
I created a little test case to demonstrate this issue, see test_pvpqgens_loads() in the attached zip-file. I modified the case9-example, limiting gen 2 and 3 to Qmax=1 and Qmin=-1. If you have a look at the results output you will notice, that the printed Q-injections are interesting. Limited gens 2 and 3 should have Q_g = 1Mvar (gen2) resp. Q_g = -1Mvar (gen3) according to the limits set, and this is also shown in the bus and generator part of the results. However, if you look in the branch section of the results, you will see that the according branch injections are 1.02 (gen2) and -1.04 (gen3) respectively.
The reason for that is the different way, how these results are calculated. While the branch flows are computed in pfsoln() with regards to the solved bus voltages as well as the line admittances, the generator injections Qg are only set back to the limited values before the PV-PQ-change. However, with makeSbus() considering wrong loads at the busses (i.e. the PV-PQ-changed generator injections) and manipulating them according to the bus voltages in the previous step, wrong bus voltages are calculated in the next step of the Newton-Raphson.

I created a workaround for my case (makeSbus_rev()), which is however only applicable, if there is only one generator at a bus without further loads or other gens. I hope that this was understandable, feel free to ask otherwise!

Have a nice weekend and kind regards,
Florian
Matpower_Test_PVPQGens_Loads.zip