load model simplification

[sanderclaeys]

A simplified version of a test feeder is obtained by
PMDTestFeeders.simplify_feeder!(data)

Directly creating a simplified dss file is not feasible, because the impedances have to be rescaled to obtain a network with a similar strength.

First of all, the transformers are removed from the feeders; this is equivalent to setting their taps to 1 (for wye-wye transformers). This lowers the voltage throughout the feeders, since most test cases include some boosting along the feeder.

For the loads, there are two options.

1. Simply change the load model to constant power, and convert delta to wye by assuming balanced voltage phasor. This is what I implemented now. The problem is that this leads to even higher voltage drops, because the consumption no longer decreases when the voltage drops (constant power). This can be quite drastic, see ieee34 in the table.
2. Run a power flow with the full component detail, and obtain the power drawn. Convert the loads to wye-connected constant power, and update their setpoints with the results from the initial power flow. This will lead to the same voltage drop.

@frederikgeth @ccoffrin @pseudocubic thoughts on which procedure is preferred for the loads?

case	vmin	vmin when simplified
ieee13	0.975163	0.887828
ieee34	0.916653	0.66764
ieee123	0.985809	0.919146
lvtestcase t=1000	1.02323	1.02323

[frederikgeth]

0.66 pu voltage is way to low indeed. The other cases are defensible though. What about say that for IEEE 34, the load set points are all scaled by a factor x after delta->wye and ZI->P conversion? And then target a vmin around 0.9 pu voltage?

[ccoffrin]

I agree with @frederikgeth, 0.66 is too low. Would another option would be to add some extra reactive support into this specific system?