In message <4FBDD81DDF04FC46870DB1B9A747269202916B42@mbx032-e1-va-8.exch032.ser
verpod.net>, "Thomas D. Erb" writes:

The interesting thing about this is that all research and experiments
(for instance on the danish island Bornholm) indicates that the only
way we stand any chance of keeping future AC grids under control in the
medium term is to lock the frequency hard to UTC.

Its a very interesting topic.

In the traditional AC grid power is produced by big heavy lumps of
rotating iron.  This couples the grid frequency tightly to the
power-balance of the grid:  If the load increases, the generators
magnetic field drags harder slowing the rotor, lowering the frequency
and vice versa.

This makes the grid frequency a "proxy signal" for the power balance,
and very usefully so, because it travels well and noiselessly through
the entire AC grid.

The only other possible "balance signal" is the voltage, and it
suffers from a host of noise mechanisms, from bad contacts and
lightning strikes to temperature, but worst of all, it takes double
hit when you start big induction motors, thus oversignalling the
power deficit.

Where the frequency as "proxy" for grid balance reacts and can
be used to steering on a 100msec timescale, you need to average
a voltage "proxy" signal for upwards of 20 seconds to get the
noise down to level where you don't introduce instability.

The big picture problem is that we are rapidly retiring the rotating
iron, replacing it with switch-mode converters which do not "couple"
the frequency to power balance.

For instance HVDC/AC converters, solar panel farms, and increasingly
wind generators, do not try to drag down the frequency when they
cannot produce more or drag the frequency up when they can produce
more power, they just faithfully track whatever frequency all the
rotating lumps of iron have agreed on.

As more and more rotating iron gets retired, the grid frequency
eventually becomes useless as a "proxy-signal" for grid balance.

Informal and usually undocumented experiments have already shown
that areas of grids which previously were able to run in "island"
mode, are no longer able to do so, due to shortage of rotating iron.

One way we have found to make the voltage a usable fast-reacting
proxy for grid power-balance, is to lock the frequency to GNSS at
1e-5 s level at all major producers, which is trivial for all the
switch-mode kit, and incredibly hard and energy-inefficient for the
rotating iron producers.

The other way is to cut the big grids into smaller grids with HVDC
connections to decouple the frequencies, which allows us to relax
the frequency tolerance for each of these subgrids substantially.

This solution gets even better if you load the HVDC up with capacitance
to act as a short time buffers, but the consequences in terms of
short circuit energy are ... spectacular?

(It is already bad enough with cable capacitance in long HVDC
connections, do the math on 15nF/Km and 100.000 kV yourself.)

All these issues are compounded by the fact that the "50/60Hz or
bust" mentality has been tatooed on the nose of five generations
of HV engineers, to such an extent that many of them are totally
incapable of even imagining anything else, and they all just "know"
that DC is "impossible".

In the long term, HVDC is going to take over, because it beats HVAC
big time on long connections, and it is only a matter of getting
semiconductors into shape before that happens.  That however,
is by no means a trivial task:  It's all about silicon purity.

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

Isn't this "hard" lock to UTC creating a single point of failure? A
solar burst, an EMP, or
a software error could leave us all in the dark.  After all, smart
inverters could be
programmed to act like big lumps of rotating iron and be compatible with
the current
system.

Pete.

On 2/9/2017 4:31 PM, Poul-Henning Kamp wrote:

On 9 February 2017 at 21:31, Poul-Henning Kamp phk@phk.freebsd.dk wrote:

Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20

I'm not sure what you mean by "balance signal" here. But I was chatting to
a friend, who used to control generators at two power stations in the UK -
one coal, the other nuclear. It is fairly obvious that the power put into
the grid must equal the power consumed, plus losses. What is also true, and
less obvious, is that the V*A must balance too. He said that he would
receive a call from the CEGB, saying they wanted X Watts, and a power
factor of Y.

Dave

On 2/9/17 1:31 PM, Poul-Henning Kamp wrote:

I think also of the issues from distributed generation - consider a
rooftop solar installation with 20 or so MicroInverters, all "slaved" to
the line.  Just from manufacturing variations, I suspect each
microinverter is a little bit different than the others.

yeah, but that's a "solvable" problem in terms of circuit breaker
design. We've all seen the Lugo substation video (not DC, but big AC
with the suppression disabled so it can "pull an arc" for test)

The Pacific Intertie is 1MV at 3000A for 1360 km.

2 bundled conductors 10m apart is about 10pF/m, or 10nF/km  (vague
recollection from somewhere)

The pacific dc intertie is a lot more than 10m apart, so it's probably
lower, but still..  14 uF @ 1MV is a bunch o'Joules. (about 14 Mj)
Fortunately, there's a fairly large series L also to slow down the
transient.

In comparison, the fairly large Marx at the Deutsches Museum is 33 nF at
1.2 MV..

underground cables have substantially higher capacitance.. I think your
number is a "suspended in air" value.

I'm not sure that's as big a deal..  the stored energy in cables (as
opposed to overhead lines) is also a big problem with AC distribution..
transient settling times are enormous and it leads to big stability
issues.  Pretty much, you can use long AC cables only for unidirectional
power transfer "source to load" not for "source interconnect" because of
the stability problems (with that rotating iron)

Umm… you left out the 25Hz power system that runs across the
road just a few miles from here :) Some of these ideas take a long time
to die. It’s been happily doing it’s thing for over a century.

Except we’ve had HVDC distribution running around for many decades
and it seems to work quite well. Indeed that’s your point here.

… or silicon carbide purity or something even more weird and (so far)
poorly understood.

Bob

In message ff3ad224-243b-f87f-6d2f-d2103b547593@comcast.net, Peter Reilley wr
ites:

Well, to be honest, all of those things would wreck havoc with
any big grid...

The bigger concern is what happens when the three GPS receivers
on your 1GW nuclear block disagree 10 milliseconds...

It is harder than it sounds.

Small solar inverters are the best, they an regulate down at milliseconds
notice, and many jurisdictions impose asymetric frequency bands on
them to exploit this.

Big inverters, no matter what you put behind them, get quite a bit
more expensive if they are designed to provide "non-VA" power,
because you suddenly have to run the current both ways in the same
half-cycle.

Nobody wants to pay for that voluntarily, and nobody are particular
keen to cause the first explosion/fire while they get the control-law
debugged.

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

Hi

If you are talking about big power gizmos, putting a GPSDO on them is pretty simple
cost and system wise. Given the fact that 10 ns sync is not required, the actual implementation
might be pretty cheap.

Bob

In message CANX10hC=ayCN_hs8EdPCt0tK=SZmRS51pm2JXj+Aj_o5W390Eg@mail.gmail.com
, "Dr. David Kirkby (Kirkby Microwave Ltd)" writes:

By "balance signal" I mean "which meter tells you if you need more
or less power in the grid".

Exactly.

Back when it was all rotating iron, they would only have
asked for the "X Watts" and they would do so because the frequency
was sagging, because that was the "balance signal" telling them
that more power was getting used than produced (or vice versa).

These days it has gotten much tricker, and I think getting into
all the details may be stretching the patience here on time-nuts,
but let me just give you two examples of how the consumption side
has also made the job harder:

It used to be that pretty much anything which drew power from the
grid would be (give and take at bit of powerfactor) an ohmic loads.

That means that if you sag the voltage, consumption drops (motors
run slower, lamps are dimmer etc, and vice versa, high voltage would
make consumption increase.  This was a beneficial feedback mechanism
trying to keep the grid stable.

These days almost anything, including computers, cars, washing
machines and lightbulbs, have a switch-mode PSU which makes it a
constant-power load.

This means that if the grid voltage increases, current drops,
reducing transport losses, which increases the voltage further.
And vice versa.  This can make voltage regulation much harder.

The other factor is batteries.  (This was first noticed during the
rolling blackouts in California caused by Enrons market manipulations.)

A city block would drop out at X kW, and usually when you cut it
in again it would be Y% higher because all fridges and aircons would
want to start.

Thesedays when you cut in a cityblock it comes in at the same
+Y%, and then about five seconds later all the chargers,
in UPS, laptops, mobile phones and whats not, cuts in, and
that can more than double the Y% and in some cases takes
the grid right back out.

Regulations have been proposed that it would make it illegal to
change any battery if the frequency is below some set limit,
in order to ensure that the grid can be relit faster and with
less energy.

So far no such regulation has been enacted, but everybody expects
it to happen after the next big urban blackout.

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

In message 63beea7a-f9fc-6e1d-b855-2c7056de3cc7@earthlink.net, jimlux writes:

Surprising there is almost no variation, because it hurts badly on
both your nameplate efficiency and thermal design.

Yes, that's the easy case, and good old dynamite based emergency
switches work too.

But shorts are another matter:  They are so much easier to deal
with when you have regular zero-crossings a hundred times a second:
HVDC shorts are explosions which doesn't end until you've used up
all the energy.

Yes, it slows down the front flank, on the other hand, once you get
the arc going, with DC that L really keeps it going for a long time.

The difference in practice is that a HVAC short leaves a burn-mark
on the metal parts, HVDC solidly welds them together if it doesn't
outright melt them.

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.