Power networks are usually almost planar, with a few exceptions in which
lines do cross each other without connection. The obvious place to
connect a network to another is in peripheral nodes, just a few of them
should be common. But if you don not have the geographical location of
nodes this can be hard to do. On the other hand, it might be possible
that the person who put the data together listed the buses and lines in
some kind of systematic manner and it might perhaps be possible to
figure out a planar map of the system.
carlos.
Shri wrote:
On Aug 15, 2012, at 9:03 PM, Carlos E Murillo-Sanchez wrote:
I can envision that this method of randomly connecting two systems might easily
create unsolvable systems, with wild angular differences even if solvable.
That is not how transmission networks are designed.
Thanks for your comments Carlos. Any tips on how to go about connecting test
cases in an engineered way rather than just naively putting in random lines?
If anyone has larger test cases (> 3375 buses) (not neccessarily MatPower
format) and willing to share it, I would really love that :)
Thanks,
Shri
Shri wrote:
Hi,
I am trying to create a bigger test case by duplicating a MatPower test case
(>= 2383 buses) multiple times. To have connectivity between each of the bigger
areas, I've added randomly chosen tie lines. However, I am not able to get power
flow to converge.
Monitoring the residual norm shows that Newton convergence is linear in the
first few steps and then it diverges. I suspect that this might be due to the
initial guess given to Newton. My premise is based on observing that if a flat
start is chosen for some (or all?) of the MatPower test cases (>= 2383 buses),
Newton doesn't converge and has a similar divergence pattern.
I was wondering whether the bus(:,VM:VA), which is part of the initial
guess to Newton, data was computed using some algorithm or provided by the test
case data provider.
Thanks and appreciate your help as always!
Shri
