Wacky Weekend: NERC to relax power grid frequency strictures
pete at altadena.net
Sun Jun 26 22:46:09 UTC 2011
On 06/26/2011 05:43 PM, Ingo Flaschberger wrote:
>> 2) Allowing transformer fields to collapse. Even in phase, without a
>> delayed transition ATS you can end up with a partially collapsed
>> transformer field with a new field being created at non-ground state.
>> This can cause a transient back wave that can snap circuit breakers.
>> Yep, this one happened to us a few times before we switched to a
>> delayed ATS, was a PITA to debug and resolve.
The collapse can take more than one cycle, especially in a 3-phase
transformer... As can the startup transient. In our case, if a big
thunderstorm came, we started one generator ahead of time and synced it
to the line, then did a 0-delay (overlap) switch. Avoids this problem,
though the generator had better be a bunch bigger than the load to bring
this off. Short delay ATS (less than 1sec) is a disaster, though. If
you always wait till the power actually fails before starting the gen,
you should wait 5-10 secs (or more) before putting the load on it, at
least if there are any motors involved. (our generators were 1mw
brushless ones that took 10-15secs for the voltage to come up
anyhow...) And HVAC compressors have their own problems; once fully
stopped you have to wait for the liquid to clear the compressor before
restarting, or have LOTS of torque (like a car unit) available (and a
supply of new belts :-)
> a transformer should be switched to the network when phase is at
> highest/lowest point, not at zero.
> zero: highist current
> highest/lowest point: lowest current
> because it's a coil.
It isn't this simple... Switching on involves transients that overwhelm
the sinusoidal waveform for a few cycles. Also the above is only
strictly true for an unloaded transformer; if it has a matched resistive
load the current and voltage are (mostly) in phase. (leakage inductance
notwithstanding, though it isn't very high for high-power
transformers.) Unfortunately most larger feed transformers are
integrated 3-phase units where the current-voltage curves are shared
between windings and get much more complicated, and there is no time (in
normal operation) when all the voltages or currents are 0. Also for
high enough ratios the interwinding capacitance is important (480 to
120/208 isn't high for this purpose; 8 or 16kv to 120/208/240 is). Big
data centers (most bigger buildings) have 2 stages of transformers, one
from "distribution" (8, 16, 20, 34kv) to 480 and a transformer per floor
from 480 to whatever. Sometimes the generator is between these two and
the UPSs are after the floor transformer(s). Big enough UPSs run at 480
so are between also.
To complicate things even more, (modern) computers look kind-of
capacitive, modern fluorescent lighting (electronic ballasts) looks
kind-of capacitive, and motors and most other loads, and older lighting
look inductive. That means it is hard to predict how to switch loads;
either no-delay (need to sync the generator, though) or lots-of-delay
operation is safer; short delay isn't. (and synchronous motors can
actually look capacitive if they have enough of a flywheel on them, but
the startup transient for a non-VF drive can be a killer.) BTW - in
reply to a misconception long before in this thread, 3-phase sync motors
self-start easily, and most older single-phase clock motors had enough
of a shaded-pole to start in induction-motor mode then transition to
sync once close to speed. The means to do this were subtle; sometimes
it involved clever multilayer plating on the rotor and/or very clever
shaping of the holes in the rotor.
What I mean by kind-of capacitive is a bit odd; it looks capacitive on
the voltage rise but resistive and/or a little inductive on the peak and
fall, and the current is 0 when the voltage is below some threshold.
Actually these days the feds (and I believe EC also) spec power-factor
correction for switching power supplies so this effect is less. The
main way this is arranged is to make sure the input-rectifier filter
capacitors are SMALL enough; then have the switcher waveform compensate
for the voltage droop. Bigger VF motor controllers do this also.
> Kind regards,
> Ingo Flaschberger
More information about the NANOG