Wacky Weekend: NERC to relax power grid frequency strictures

Pete Carah pete at altadena.net
Sat Jun 25 21:29:38 UTC 2011

On 06/25/2011 03:52 PM, Jason Roysdon wrote:
> On 06/25/2011 08:06 AM, William Herrin wrote:
>> On Sat, Jun 25, 2011 at 10:49 AM, Jay Ashworth <jra at baylink.com> wrote:
>>> Perhaps I read the piece incorrectly, but it certainly sounded to *me* like
>>> the part that was hard was not hitting 60.00, but *staying in sync with
>>> others*...
If the grid is enough bigger than any one generator, that happens
automatically; once
connected, you can't get out of sync without tripping a circuit
breaker.   There are some
minor exceptions for DC and VF AC generation (e.g. solar and wind
>> Way I read it, when they occasionally run at 59.9hz for a few hours
>> (and according to my UPS monitoring software this is a regular
>> occurance), they're no longer going to run at 60.1 hz for a while so
>> that the average comes out to 60.
Finer gradations than that, but yes.
>> Regards,
>> Bill Herrin
> This paper describes what they currently do to keep clocks accurate with
> Manual Time Error Correction (which is what they are going to suspect
> for a year):
> http://www.naesb.org/pdf2/weq_bklet_011505_tec_mc.pdf
> As I said in my last post, I'm not an EE, but just follow some of topics
> on that side of the house.
> What I gather is that Manual TEC, which is done by purposely running the
> frequency away from 60Hz to correct an average deviation, can actually
> cause more problems.
> "NERC is investigating the possibility of eliminating Time Error
> Corrections. NERC has been collecting data regarding Interconnection
> frequency performance, including the number of clock-­‐ minutes during
> which actual frequency dropped below the low Frequency Trigger Limit
> (FTL) of 59.95 Hertz. During the period of July 2005 through March 2010,
> approximately 44% of the minutes during which clock-­‐minute actual
> frequency dropped below the low FTL occurred during Time Error
> Corrections when scheduled frequency was 59.98 Hertz (1,875 of the 4,234
> total minutes observed below 59.95 Hertz). Upon further investigation,
> it was found that almost all of those minutes (1,819 of the 1,875 total)
> represented frequency deviations that would likely not have dropped
> frequency below 59.95 Hertz if the scheduled frequency had been 60
> Hertz. In other words, approximately 97% of the Low FTLs were of such a
> magnitude that if the Time Error Correction had not been in effect, the
> exceedance of the low FTL would not have occurred.
This is an example of counting an intentional deviation as a fault and
is really only
an accounting problem (even though the triggers are mostly fixed hardware).
On the other hand, if customers can notice, it matters.  (what was the
SLA again?)

Interesting...  Normally slow is not a problem since it happens normally
when the grid is running near
capacity...  The "frequency runs" (when I worked IT/comms with a power
company they were on Thursday nights after midnight...) were almost
always fast (I was involved in this in the mid 70's mostly; our company
rarely reached capacity but could if the weather was hot enough.)  Given
that the slew command has to be done almost simultaneously at most of
the power plants in US and Canada, in order to avoid the throttle
hunting problem that JDA mentioned (yes, even big plants can hunt, just
ask the NE folks (twice!!!) though
both of them were caused by circuit faults (network partitions) and not
time corrections)

Now that most clocks are run by 32khz crystals and not counting cycles,
the corrections don't matter as much.  I guess the experiment is to find
out who complains...

Then again, as a kid I remember So Cal Edison sending out new rollers
for phonographs (I know - what's that?) when they changed from 50 to
60hz.  Talk about inaccurate clocks (and I don't know what they did
about those either.) And most of LA's hydro was still running generators
that were built for 50hz even into the 70's, so lost some efficiency.
> These Frequency Trigger Limits in and of themselves are only indicators
> of system behavior, but the nature of their relationship to Time Error
> Corrections calls into question the potential impact that Time Error
> Corrections can have on frequency behavior overall. While it is
> intuitively obvious that any frequency offset that moves target
> frequency away from the reference point to which all other frequency
> sensitive devices (such as relays) have been indexed will have a
> potential impact on those devices’ performance, the industry has by and
> large regarded Time Error Corrections as harmless and necessary as part
> of the service it provides to its customers. However, in light of this
> data, NERC’s stakeholders are now questioning whether or not the
> intentional movement closer to (or in some cases, further away from) the
> trigger settings of frequency-­‐based protection devices as is evidenced
> during Time Error Correction events is appropriate.
> Accordingly, NERC is planning a Field Trial during which the practice of
> doing Time Error Corrections will be suspended. Because of the
> fundamental nature of this 60Hz signal, NERC is reaching out to various
> industries to get their thoughts on whether they anticipate any problems
> with the elimination of Time Error Corrections. Those industries include
> appliance manufacturers, software companies, chemical manufacturers,
> companies that make automation equipment, computer manufacturers, and
> many others."
> Source:
> http://www.wecc.biz/library/Lists/Announcements/Attachments/9/June%2014%20Time%20Error%20Webinar.pdf
> The main point I gather is that trying to do manual Time Error
> Correction actually makes the power grid less stable at times, and as
> such they want to do away with it (thus making the power grid more stable).
Given that you have to bump up everyone's throttle half the time (or
more) to
accomplish the correction, if things are already near capacity, you are
asking for it :-)
> Think of it the same as patch management risk assessment.  If there are
> no security or bug fixes that directly affect you or even feature
> enhancements that you don't need, do you apply a patch/upgrade to
> critical systems?  Nah, you skip those, because we all know every
> patch/upgrade carries with it risk of an unknown bug or even security flaw.
> That's what they're doing here, opting to skip "patching" the time
> error.  They're not ignoring frequency altogether, but rather only
> minding that aspects that have to do with grid stability, not your alarm
> clock.  This is for the better anyway, and NTP/GPS/WWV/WWVH is the way
> to go to keep clocks accurate and hopefully will be the outcome of any
> consumer complaints.
Already most "modern" consumer clocks are driven at least by
crystals if not externally sync'd...
> I've seen conversation in various forums and lists I read that they are
> going to ignore or not care about the 60Hz standard.  This is incorrect.
>  They just aren't going to purposely deviate from the scheduled
> frequency to perform manual TEC.
> Mind you, that they still care about why the frequency is off, and when
> things are not able to quickly compensate, they want to know and be able
> to pinpoint it and fix it:
> http://www.nerc.com/filez/standards/Frequency_Response.html
I didn't read that, but the short answer is: build more generators... 
(and I don't know the
politics of how the ISO's fit into this; they complicated things a lot
from when I was involved
in the power biz.) 

> Specifically, read this PDF:
> http://www.nerc.com/docs/oc/rfwg/NERC%20Balancing%20and%20Frequency%20Control%20July%205%202009.pdf
> The AP piece was focused on hype and word-spinning (I couldn't find an
> AP.org link, so used one that I could find,
> http://www.huffingtonpost.com/2011/06/24/clock-problems-power-grid-clock-disruptions_n_884259.html
> ):
> "The experiment would allow more frequency variation than it does now —
> without corrections."
> The NERC BAL standards already hold the NERC entities to very high
> frequency standards, and this will be unchanged, except for manual TEC.
>  All it is doing is eliminating the corrections made purely for time's
> sake, which actually eliminates more frequency variation.  This may, or
> may not, create more average frequency variation, and that is part of
> this test.
It probably will eliminate most of the faster slews and some risky manual
> "Officials say they want to try it to make the power supply more
> reliable, save money and reduce what may be needless effort."
> This is the real goal, and should have been the focus of the news story
> - but that doesn't make headlines.
> I'm going to go shop for a new clock.  I had one that used the WWV/WWVH
> stations, but then they messed with DST and it was off for a few weeks
> around each DS change.  
Which is cute since there is a field in the wwvb message that gives a
countdown until dst change, and flags the direction.  One of my "atomic"
clocks gave up and just has a DST button...  The other is like yours.
> That forced me me to pull it off the wall and
> change the TZ at one time of the year to correct it, but the other time
> of the year I could not correct it (as it only had 4 TZ settings), so I
> took it down.  Beware "Automatic Time Set" clocks which don't really
> learn the time from the WWV/WWVH stations (like the Sony ICF-C218, which
> has a preset time and battery, but still uses the frequency from the
> wall to maintain time).
Booo.  I'm surprised at that in this day & age (a 32khz crystal and the
counter it needs is actually cheaper than the analog signal-conditioning you
need to count the 60hz reliably).
>   The best bet is a clock that requires batteries
> as you know it won't get time from the power grid.
> Jason Roysdon

More information about the NANOG mailing list