60Hz US Power Grid and precision timekeeping

Factual Questions
21

Okay, I know something about this, and I can tell you that 58.73 Hz on a large 60 Hz synchronous system is very unlikely, and would represent a dire emergency for the system operator.

I don’t know exactly how it’s managed in North America, but in Europe it works something like this:
Primary regulation (of frequency) is provided automatically by generating units, which modulate their power output up and down slightly to compensate for small variations (+/-0.2 Hz) in the system frequency. The timescale is of the order of 15 seconds.

Secondary regulation involves generating units being instructed to increase or decrease their power output more substantially, to compensate for a larger frequency excursion. The timescale here is approximately 1.5 minutes.

If this did not happen, a sudden drop in frequency (caused, for example, by the loss of a major generating unit) can cause other generating units in turn to trip, leading to a runaway effect that in the worst case can leave country-sized areas blacked out for long periods. To prevent this, decisive action in the first few seconds is critical.

About the time service, every hour or so, the system operator will adjust the system frequency to try to achieve exactly 50 Hz on average over the day.

And yes, it is still possible to synchronise a generator of any size using light-bulbs without making the earth move, and shift managers still know how to do that.

22

I will second this. It takes perhaps 5 minutes and 2-3 repetitions to teach someone how to do this. The indicator I have used was just two light bulbs. You adjust the speed off line until the pulsing is very slow, and throw the switch when both are off. I think one would work, but two provide redundancy.

It may be that gas turbines, etc do not throttle so nice, and utilities may want them online as quickly as possible, so modern practice may be to “catch” the at-speed, in-phase condition as they ramp up under rather high power input…so computer control may indeed be preferred.

But it is rather trivial to monitor the extant phase of the grid if you want to automate the process: there is never a need to predict the phasing based on absolute timekeeping. That is to say that you have the mother of all phase references connected to any generating station.

As for the Glenwood Canyon plant providing a phase reference for an entire grid, I can’t really see how that could work. The frequency/phase of the grid depends on the net mechanical power input at the aggregate of all generators, and the net load on the grid. A big generating station can push the frequency/phase around a little bit, but the Glenwood Canyon plant is fairly small. Small enough that I have seen Kayakers “playing” in the discharge. They are also pretty limited on their ability to control mechanical power input, as that dam contains very little water as such things go. If they tried to run at 2X the inflow rate, the level would probably drop noticeably within 10-20 minutes or so. (total WAG, but I have an engineer’s eye, and probably looked at this place at least 100 times over the last 50 years)

23

On a three phase generator you want three bulbs, one for each phase. If they blink out of order you’ve got your phases out of sequence and you need to switch two of your lines around or very bad things will happen when you throw the switch.

Technically one bulb would work for a single phase or a split single phase generator, but two bulbs gives you a visual indication that you don’t have one of your lines and neutral reversed. It’s a bit of an idiot check, but it’s nice to have (IMHO).

24

I suspect that what they really mean is that Glenwood Canyon is the official time keeper for the system and sends out the commands to all of the other systems to coordinate their generators and get the cycle counts back where they want them.

I know kevbo understands this, but for the benefit of others who might not, here’s how generators work.

If you take two generators and tie them together, they hold each other in sync. If you add power to one, it ends up trying to spin the other generator like a motor to get it to catch up to the same frequency, so it gets much more difficult to speed the two generators up. Similarly, if you remove power from one generator, it starts to act like a motor and gets spun by the other generator, making it more difficult to slow the pair down. With more than two generators it gets even worse. If you have ten generators on a line and you remove power from one, the other nine will spin it like a motor, and with that many generators adding power or removing power from one generator has very little effect on the overall line. So, it gets very difficult to speed up or slow down the entire line. You have to coordinate with all of the generators to slightly add or remove power simultaneously.

So speeding up or slowing down the entire “grid” ends up being a coordinated effort between all of the generators on the system. One generator can’t directly control the line. Anything that “controls” the overall grid is really sending control commands to each generator to coordinate their actions on the grid.

Power systems in general are loaded much more heavily during the day, especially on hot days. In the U.S., the northeast and southwest are both close to being overloaded on hot summer days when everyone is running their air conditioning. This is why one little failure in either location can lead to massive blackouts (overloading makes them vulnerable to what is known as a “cascade failure” where since everything is tied together, a fault on one means that the one next to it can’t supply enough power to cover the fault and it also trips, causing the one next to it to also trip, and so on and so on in a long cascade of failures).

Since everything is overloaded during the day, this tends to slow down all of the generators at once. So they make up for it by speeding things up a bit at night when the load decreases.

It may read low during the day and higher at night.

I also thought I read somewhere that they weren’t going to maintain the long term frequency like they used to since synchronous clocks aren’t used much any more.

25

In theory, it would be because power is cheaper and more plentiful at night, as loads are lower. Keeping loads constant, frequency is a function of generation levels. During the day, we have what we call our “peak” hours that require us to ramp up our generation resources, commit less efficient units, etc. At night, we often run into minimum generation constraints, where it is difficult to ramp our generation levels down to the low load levels that occur at night. This often causes utilities to need to “dump” energy at night (selling at a loss). Obviously, you could reduce gen by decommitting units, but if you need those units for the peak of the next day then can be problematic - either expensive (incurring start costs) or sometimes impossible, due to minimum down times or ramp rate constraints.

So, all in all, you have plenty (or more) power at night, and are often just squeaking by during the day, which means that if you wanted to fiddle with frequency a little bit, then running a little short during the day, and a little long during the night would make sense. The reality is that this isn’t really done, as the regulations from the NERC are pretty tight, and it’s just easier to play it straight. What we do is to continually over or under generate by small amounts, trying to swing back and forth several times per hour (this is easier than trying to nail it, and more efficient).

FWIW, the actual thing we monitor on our trading floor / dispatch desk is ACE, or Area Control Error - we see frequency, but ACE is what we dispatch to.

26

I(Am)APE

My vote is on an error at the Fluke meter or something in the distribution system may be causing a false reading. That frequency is well below where automatic relay systems would shed load in order to bring the frequency back up.

Somewhere around 59.3 the load aggregators are required to shed around 6% of the system load. Below 59.1 Hz another 6%.

Way to much info on this here The table at section 1.1.1 shows frequencies and load percentages.

Basically at 58.73hz about 20% of the load has been shed and generator safety systems are quickly timing out and tripping the generators off line in order to avoid damaging them.

27

I thought you(were)Dog.

28

But that last sentence just glosses over the question Napier had that’s still unanswered. Why do they need to make up for it? Why not just aim for the correct frequency? I don’t think Darth Panda’s link on Area Control Error answers this either.

Prior to electric clocks, I have to believe people were used to adjusting their clocks often, or at least living with them off by a minute or five. A minute is 3600 cycles, so nailing that number of cycles every day or so seems like overkill.

I thought of that, but racer72 says his clocks lose time over a month.

(Unless… racer72, these electric clock aren’t battery powered, are they? Just checking…)

29

On rereading ZenBeam’s post, I’m changing my post.

30

2 parts:

  1. We do.
  2. Kind of, we swing back and forth about 15 times per hour.

The day-time night-time thing is kind of explained in my post, if you read it carefully. (But to summarize and be more explicit: sometimes we drag during the day, because load is creeping up and we have to chase it. But, in general, we stay pretty flat most of the time).

I’m not hypothesizing - I’ve spent years working with these guys, the actual dispatchers who run a very, very large US system.

31

Missed edit window again, sorry.

Next Tuesday, I’ll ask our director of dispatch operations if there’s anything else that’s relevant here.

32

So the 5184000 cycles per day thing is a myth? Looking at Napier’s links supporting the “speed up at night to get to 5184000 cycles every day” theory, they do all seem a little weak. Nothing authoritative, really.
On another note:

Alright Smarty-pantses. Where were you back in 2008? (OK, Darth Panda gets a bit of a pass for not having joined yet…) Seriously, if you know or can find out anything about that, I’d be interested.

(In the thread linked to by the thread above, I linked to a bunch of pictures that aren’t there anymore, but I was playing around with Google Sites then, so they’re available here if you’re interested.)

33

Back in the 30’s many rural power systems were basically islands of isolated producers and customers. The frequency could have been different from town to town.

You’d probably have to sell your soul to the devil to get the answer.

34

During a hot day, it’s much more likely that you’re “dragging,” that is your frequency is too low (we’re talking small amounts here - the NERC fines the living shit out of us if we don’t stay pretty on), than running too hard. If that happens, we’ll catch up as soon as we can, which happens to basically be “night.” This doesn’t happen all that often for us though - but every power system is different, so I can’t really speak for other utilities (except for to say that I doubt they like getting fined any more than we do).

So it’s not entirely a myth, more of an exaggeration.

I just started with the Company in 08. Like I said, I’ll talk to some of my buddies, and one guy in particular, on Tuesday. I assure you I’ll find out much, much more than I really want to know.

eta: here is a somewhat comprehensible review of some of the standards: http://www.nerc.com/docs/oc/ps/tutorcps.pdf (warning: pdf)

35

But which is more accurate… the grid frequency, or the time base in your meter?

36

Well, they do. But it will never be *exactly *60 Hz at any instant. It will either be less than 60 Hz (e.g. 59.999234 Hz) or greater than 60 Hz (e.g. 60.0001059 Hz). Furthermore, I would assume any drift in the frequency occurs slowly. Their goal is for the *average *frequency to be 60.0000000 Hz over a given time span, like 24 hours. To do this, they make small adjustments to the frequency so that the long-term average is 60.0000000 Hz or whatever.

Why do this? Because many instruments use the grid as a frequency reference, and the power company doesn’t want any instrument maker to blame them for not providing a good, accurate frequency. It is impossible for the power company to provide a frequency of exactly 60 Hz at every instant, so they do the next best thing: an averge of exactly 60 Hz (or as close as possible to it) over a given time span.

37

From a fundamental level, the answer is because it is impossible. With a varying load you must use some form of feedback system to match the power input (ie the power on the shafts of the generators) to that varying load. You either need to be prescient (which is hard for the average engineer that likes to obey the laws of physics) or you need to apply a correction factor to the system in order to vary the power input to the load, the correction being derived from observing the current behaviour. No physical system can adjust instantly, everything takes some delay, and the moment you have delay you are in the realm of feedback systems and stability analysis. Get it wrong and the system will go unstable. So engineers like like nice well damped feedback system that behave nicely. (Very roughly you must ensure that the speed that the feedback system demands the power input change is slower than the speed at which it actually can vary. The actual margin needed is defined mathematically.) In such a system the frequency will naturally hunt above and below the target. When it does this in a nice smooth manner people are happy. They get given a set of bounds that the hunting behaviour must stay within, but as noted, these - of themselves are reasonably wide. If they were much closer the feedback system would be less easy to get stable, and instability is a very bad word around power engineers.

The power generator managers have a slightly more strict goal than just maintaining the frequency within some defined bounds, they have a target integrated frequency - ie the time. So this guides their operation as described above. Basically it provides an additional set of corrective input to the power input control in addition to just hunting for a target frequency. This additional input is subject to the same stability constraints as the main one. Luckily giving them a 24 hour period in which to get there avoids stability issues by a good margin, but in principle it must be considered.

38

I would think minimizing the RMS frequency error would be “the next best thing”, not minimizing the average frequency error.

Which came first? Instrument makers relying on the number of cycles being correct, or electric power maintaining the correct total number of cycles?

That’s if they even do maintain the correct number of cycles, which I’m not sure is the case. Being off by five cycles per day is one part in a million, which is pretty accurate. And if you needed better accuracy, there are (or were) time-keeping radio stations. The Area Control Error equation that Darth Panda linked to is a mix of power error and frequency error. It’s not obvious to me that minimizing ACE will necessarily lead to the correct total number of cycles over a span of time. As I said, it’s sounding like a myth to me.

Maybe I missed it, but can someone point to an authoritative cite that they do maintain the correct number of cycles over a day?

Sigh. It’s certainly possible to aim for the correct frequency.

The question, which is surprisingly difficult to get across for whatever reason, is, when you’re already off in one direction (say, low), why aim for the correct number of cycles for the day, instead of aiming for the correct frequency, and not worrying about making up those lost cycles? Why isn’t the goal to minimize RMS frequency error?

39

It is, but the whole point of what I wrote was to point out that this is not the best solution. You can aim, but you can’t ever achieve. Just specifying an RMS value does not include a time period over which you will accept it varying. Stability of the control system places a hard bound on how fast you can vary it, and that places a bound on how tight you can pull in the error. The tighter the error bounds on frequency the less and less stable the control system becomes.

You can place constraints on the damping of the feedback loop. A highly damped system will slowly run up to the set frequency, a less damped one will overshoot and drop back, an underdamped one will wobble back and forth.

In the end you seem to be asking - Why bother with time at all? That is different. The error in frequency is allowed wide for good reason. Keeping everyone’s clocks in sync a nice to have. The two aren’t in conflict, so it is easy to do. Tight frequency control is very hard through to impossible to do.

40

Years ago I purchased a turntable for my mobile DJ business. (Yes, Virginia, there were such things as “records” back in the dark ages)

The store had just taken delivery of a whole lot of these direct drive turntables. I got to the job and set up my equipment.

Good damn thing I hadn’t bought two of them. The turntable was a 50hz motor and the 60 hz power really screwed up the motor speed. I had to do the whole night on one turntable. Not fun.

I took it back to the store the next day. They couldn’t see the problem until I insisted they plug it in and play a record. Turns out the whole batch they got were 50hz motors.

The timing on the power is for a lot more than clocks.