Speak EV - Electric Car Forums banner
1 - 20 of 64 Posts

·
Registered
Audi e-tron 50
Joined
·
64 Posts
Discussion Starter · #1 ·
I work in the control room of a large power station. Over the weekend I was in over four nights. On each of these nights we saw grid frequency dropping suddenly at exactly 00:30 hrs. This is the first time our shift's noticed this. As all of us in the control room have EVs, we guessed what was happening and called National Grid to see what they knew. They confirmed the drop was due to the tariff change for suppliers such as Octopus Go. The frequency drop happens because of an instantaneous increase in power demand which is typically 500 - 600 MW. Right now, the UK electricity demand is just under 35 GW and, at 00:30 hrs it is around 27 GW (G. B. National Grid status) so the step change is around 2% this may not sound much but it is significant and has to be carefully managed by the Grid.
 

·
Super Moderator
Joined
·
7,508 Posts
Law of unintended consequences - GPS accurate clocks in our cars mean their charge timers, when set the same, all start within seconds of each other. If they were controlled by their own internal clocks that all slipped one way or another over time it would be a bulge not a cliff of demand.
 

·
Registered
Peugeot e-208
Joined
·
659 Posts
I know Octopus offers different slots for Go - presumably for exactly this reason. Maybe in the future they'll have to spread it out even more, starting at random minutes past the hour.

But why would it suddenly happen now? Are there suddenly more customers on Go?
 
  • Like
Reactions: pressurized

·
Registered
Joined
·
17 Posts
I saw a graph of agile prices and noticed they actually jumped up at 1230 before steadily falling again, presumably because of this 1230 spike. I know my charger is set to come on at 1230.

I think there are now many more miles being done in EV's than say a year ago and not just because there are more EV's on the road but because the people who are now buying them are higher mileage drivers. The 0% BIK means business users who are generally higher mileage but also now cars with 200 mile+ ranges are common, serious mileage users are switching to them. For the past few years with smaller ranges I'd guess that they weren't used by people who drove all that much.
If your doing 100 miles a day its well worth your while being on Octopus Go, if its only 100 a week you might not bother.
All guess work on my part.
 

·
Registered
Audi e-tron 50
Joined
·
64 Posts
Discussion Starter · #6 ·
The step change won't have come suddenly. It will have gradually grown as EVs are gradually bought and customers gradually migrate onto variable tariffs. I only noticed it in my last run of nights because it was above a threshold large enough to bring in a fleeting alarm on an uninterruptible power supply.
Grid will have become aware of this 'problem' far sooner than I was and were ready for it - they are very good at what they do.
Changes in tariffs' start-stop times will be driven by the National Grid and they'll do it by influencing the hour-by-hour cost of power. Octopus Agile is a brilliant example of this. Grid will be very happy about the potential here and will exploit it as a win-win for them and us.
Also bear in mind that the chargers many of us use have been part-funded by the government (the wise parts of it). This is not just because they want us to move over to less polluting things but also because, when we inevitably do, the chargers they've got us to use can be remotely throttled to reduce demand when the grid goes through an unexpected transient like the loss of a big generator. While this may annoy you, it shouldn't. Unless the transient is unusually large (like ones you see in the news), the throttling is unlikely to be total and last longer than it takes to bring replacement plant online (typically 10 minutes). And the Grid's ability to do this is likely to avoid power cuts which these days are really annoying.
 

·
Registered
Joined
·
170 Posts

·
Registered
Audi e-tron 50
Joined
·
64 Posts
Discussion Starter · #9 ·
(I urge you not to make it widely known but Geeks do actually run the world...)
National Grid's openness is excellent and I'm sure the data in your trend are correct but they appear to be average frequencies over a minute so very short-term transients are not visible.
This is what the instrumentation visible to me recorded last night:
142129

The averaging here is only around maybe three seconds so this trend's resolution is much higher.
Frequency is purple and is baselined at 50 Hz. The '00:32' bottom right is the time when I froze the trend to copy it and the time base (period?) of the trend is thirty minutes.
The step-change you see at 00:30 hrs is significant to the grid. While it only bothered me where I work because it brought in a fleeting alarm and made me glance at my alarm screen, to the National Grid that change will have required action which they appear to have carried out beforehand.
 

·
Registered
Joined
·
2,792 Posts
I work in the control room of a large power station. Over the weekend I was in over four nights. On each of these nights we saw grid frequency dropping suddenly at exactly 00:30 hrs. This is the first time our shift's noticed this. As all of us in the control room have EVs, we guessed what was happening and called National Grid to see what they knew. They confirmed the drop was due to the tariff change for suppliers such as Octopus Go. The frequency drop happens because of an instantaneous increase in power demand which is typically 500 - 600 MW. Right now, the UK electricity demand is just under 35 GW and, at 00:30 hrs it is around 27 GW (G. B. National Grid status) so the step change is around 2% this may not sound much but it is significant and has to be carefully managed by the Grid.
I was with EDFs electric tariff (now switching to Go but not there yet) and it turned out that users were randomly assigned a start time of between 0 and 11 minutes either side of the theoretical start time. Maybe bigger users like Octopus need to start doing this? Assuming they dont already?
 

·
Registered
Joined
·
570 Posts
The octopus load is quite small compared to storage heating. The problem such as it is, appears to be that the fancy new smart charger switches the whole population at exactly the same time, a population of old fashioned mechanical timeswitch will drift / not be perfectly set so the hard edge is softened over several minutes.
More "modern" teleswitches that switch load based on a radio pulse have a randomisation built in so that the whole population doesn't switch at once.
It shouldn't be difficult for a properly thought through charging solution to implement something similar.
 

·
Registered
Corsa-E Elite
Joined
·
63 Posts
On those graphs the frequency rises in the minutes before 0:30, that's probably them bringing on extra generation in preparation for 0:30
 

·
Registered
Kia e-Niro 4 MY20, Zoe Z.E.50 GT Line
Joined
·
1,363 Posts
Fascinating stuff, thanks for sharing. I wonder if we’ll see charge timers in cars subject to some subtle randomisation to soften the blow, as it were. It’s a shame the smart evse are not smarter to be honest. I think the Ohme approach is potentially best, don’t they have a gps? If so, theoretically, all Ohme cables in an area can be staggered in their switching
 

·
Registered
Audi e-tron 50
Joined
·
64 Posts
Discussion Starter · #14 ·
The extremely precise timing of things is likely to be due to their connection to the Internet and the Network Time Protocol (NTP) rather than the timing of any GPS they might have. GPS clocks around the world contribute to keeping the timing of the NTP accurate.
This is how one of my raspberrypis is tracking time at the moment:
Code:
     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
 0.debian.pool.n .POOL.          16 p    -   64    0    0.000    0.000   0.002
 1.debian.pool.n .POOL.          16 p    -   64    0    0.000    0.000   0.002
 2.debian.pool.n .POOL.          16 p    -   64    0    0.000    0.000   0.002
 3.debian.pool.n .POOL.          16 p    -   64    0    0.000    0.000   0.002
-time.cloudflare 10.21.8.19       3 u  857 1024  377    5.598   -2.727   0.418
*85.199.214.99 ( .GPS.            1 u   27   64  377    7.945   -0.091  14.879
-time.cloudflare 10.21.8.19       3 u  835 1024  377    6.044   -2.533   0.509
+lux.22pf.org    139.162.219.252  3 u  834 1024  327    5.943   -1.081   0.583
+ns3.turbodns.co 85.199.214.99    2 u  271 1024  377   11.418    0.059   0.374
-ntp0.sotaconnec 114.199.6.79     2 u  243 1024  377    8.028    0.139   0.493
The starred line is its currently preferred Internet clock which appears to use GPS.
 

·
Registered
MY18 R110
Joined
·
389 Posts
Real data from real people at the coal face (so to speak :) ) great stuff.
For the love of Bob nobody tell the Daily Mail, the headlines would read that we are killing the grid.....and that Diana is still the greatest !

Interesting to see the switch to EVs is making an impact. I guess we will see more things like project FRED from MyEngergi if you have a Zappi going forward.
 

·
Registered
Joined
·
4,716 Posts
There are probably a lot of people like me, who during lockdown are going pretty much nowhere. I only need to charge a couple of times a month at the moment. I expect this effect will be a lot greater once lockdown ends.
 

·
Registered
Joined
·
5,099 Posts
Law of unintended consequences - GPS accurate clocks in our cars mean their charge timers, when set the same, all start within seconds of each other. If they were controlled by their own internal clocks that all slipped one way or another over time it would be a bulge not a cliff of demand.
Sounds like a good reason to use a charge end timer instead of a charge start timer...

I use a charge end timer (set to 7am as I don't have a dual rate tariff) and the start time will automatically fluctuate depending on how low the battery is the night before since the car calculates when to start charging to meet the deadline.

So setting a charge end timer to the end of the cheap period instead of a charge start timer to the start of the cheap period would certainly alleviate the problem of everyone starting to charge at the same time on dual rate tariffs.
 

·
Registered
MG ZS EV
Joined
·
542 Posts
The extremely precise timing of things is likely to be due to their connection to the Internet and the Network Time Protocol (NTP) rather than the timing of any GPS they might have. GPS clocks around the world contribute to keeping the timing of the NTP accurate.
This is how one of my raspberrypis is tracking time at the moment:
Code:
     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
0.debian.pool.n .POOL.          16 p    -   64    0    0.000    0.000   0.002
1.debian.pool.n .POOL.          16 p    -   64    0    0.000    0.000   0.002
2.debian.pool.n .POOL.          16 p    -   64    0    0.000    0.000   0.002
3.debian.pool.n .POOL.          16 p    -   64    0    0.000    0.000   0.002
-time.cloudflare 10.21.8.19       3 u  857 1024  377    5.598   -2.727   0.418
*85.199.214.99 ( .GPS.            1 u   27   64  377    7.945   -0.091  14.879
-time.cloudflare 10.21.8.19       3 u  835 1024  377    6.044   -2.533   0.509
+lux.22pf.org    139.162.219.252  3 u  834 1024  327    5.943   -1.081   0.583
+ns3.turbodns.co 85.199.214.99    2 u  271 1024  377   11.418    0.059   0.374
-ntp0.sotaconnec 114.199.6.79     2 u  243 1024  377    8.028    0.139   0.493
The starred line is its currently preferred Internet clock which appears to use GPS.
Oh this is wonderful stuff, I’m already the centre of piss taking because of my love of wind farms , more stuff to talk about when the pubs open!!
 

·
Registered
Joined
·
4,795 Posts
@pressurized , here's an idea for you. EVSEs all have a cpu in them, which signal the available current to the EV via a PWM modulated signal down the CP wire. The EV has to respond to an indicated change of current within 5 seconds. It would not be difficult to add a frequency measuring feature, an A2D input monitoring a voltage divider will provide the crossing point info no problem. The EVSE could then simply reduce the current on offer itself, while the mains frequency is looking a bit low. Would need a bit of tuning, as you don't want an unstable feedback-loop with millions of EVSE's all going into a panic, and reacting far faster than the grid can! So whatever's the response-time of the grid to "instantaneous" frequency load changes, the EVSE's would need to be slower than that to keep the behaviour stable & well damped.

If the grid prepares in advance by firing up some power stations just before East Enders finishes, then the freq will be >50 Hz, and all EVSEs would be operating at expected current. The moment the kettles go on and Hz < 50, let's say <49.9 to give it a deadband, then you could have any EVSEs that are charging in "gentle" overnight-slowish-mode could slowly reduce their offered current, maybe reduce it by 10% at most (some to-be-decided sensible amount that will help the grid but not inconvenience the EV) until the freq >= 49.9 & normal service is resumed as the EVSEs crank up their current over say a 10 minute period. You could of course make the %age reduction proportional to the frequency-error, if you wanted. Whaddya think? I could knock this up on my EVSE if you're interested to see how it might behave? I'm thinking of making a CP-interceptor unit for solar-control reasons anyway, and I've already coded my own version of mains-monitoring library for power-factor voltage & current measurement, so it's not hard to do tbh.
 
1 - 20 of 64 Posts
Top