andy.g4jnt@gmail.com said: > I can probably measure the grid frequency to a few micro Hz over a period > of tens of seconds. So I make a continuous recording of this, averaged over > say 10 second periods. Now take a 7kW load (the maximum reasonably possible > on a domestic circuit) and switch it on and off at intervals of perhaps 10 > minutes, precisely timed so it can be correlated with the frequency log. > That 7kW load will be about 0.2ppm of the average for that for the whole of > the UK. By post processing, and some deep correlation covering days worth > of cycles of load on-off with the frequency, I wonder if it would be possible > to see the loading, the mean frequency changing by a few uHz. Not sure what > the time constant of the grid control is, but for* small signals* I doubt it > can be faster than a few minutes. I'd expect an ADEV type pattern. For long tau, the wander in the system will dominate. We have lots of long term data so should be able to plot that part of the graph. For short tau, the noise will dominate. I don't have any data below 10 seconds. But I do have a TICC. I wonder if the middle is the classic V or has some wiggles. How much energy is in the rotating turbine and generator? That should provide a lower limit on the time constant of the control loop. If the ADEV pattern is a clean V and the signal is 2 levels, I'd expect the optimal modulation pattern is a homework problem for the right course. We can synchronize the transmit and receive clocks with GPS. This feels like I have to use something like 8B/10B with lots of transitions so the receiver can track the system frequency? -- These are my opinions. I hate spam.

-------- Hal Murray writes: > I'd expect an ADEV type pattern. For long tau, the wander in the system will > dominate. We have lots of long term data so should be able to plot that part > of the graph. I did that some years ago: The Nordic grid bottomed out around 1e-10, with a frequency error less than 2e-11 measured over 45 years. > How much energy is in the rotating turbine and generator? That should provide > a lower limit on the time constant of the control loop. Answer A: Surprisningly much (do the math on 50 tons of iron, 1m diameter at 3000 or 3600 RPM) Answer B: Not enough to keep modern grids stable (Google term: "Low inertia conditions") -- 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, On 2021-02-08 12:31, Hal Murray wrote: > andy.g4jnt@gmail.com said: >> I can probably measure the grid frequency to a few micro Hz over a period >> of tens of seconds. So I make a continuous recording of this, averaged over >> say 10 second periods. Now take a 7kW load (the maximum reasonably possible >> on a domestic circuit) and switch it on and off at intervals of perhaps 10 >> minutes, precisely timed so it can be correlated with the frequency log. >> That 7kW load will be about 0.2ppm of the average for that for the whole of >> the UK. By post processing, and some deep correlation covering days worth >> of cycles of load on-off with the frequency, I wonder if it would be possible >> to see the loading, the mean frequency changing by a few uHz. Not sure what >> the time constant of the grid control is, but for* small signals* I doubt it >> can be faster than a few minutes. > I'd expect an ADEV type pattern. For long tau, the wander in the system will > dominate. We have lots of long term data so should be able to plot that part > of the graph. For short tau, the noise will dominate. I don't have any data > below 10 seconds. But I do have a TICC. I wonder if the middle is the > classic V or has some wiggles. ADEV is a poort estimator for this. ADEV was meant of noise power estimation, but power-grid is mainly systematics (pseudo-random/systematic) and feedback-loops on to counter that. There is other tools in the toolbox for that. There is a wealth of tools to cover that. > How much energy is in the rotating turbine and generator? That should provide > a lower limit on the time constant of the control loop. Only some loops are automatic, and often those are local. That is, protection switches break lines on over-voltage, over-current, over-frequency-deviation, over-ROCOF (Rate Of Change Of Frequency) etc. Regular read-out of the state is done and risk calculation is done. Then the overall state is monitored by these protein computers supervising it all. The overall feedback loop is thus fairly slow but kind of works. The read-out rate has classically been an issue, especially as you want to see two or three readouts being consistent, and you get an update very 5 s or so. > > If the ADEV pattern is a clean V and the signal is 2 levels, I'd expect the > optimal modulation pattern is a homework problem for the right course. We can > synchronize the transmit and receive clocks with GPS. Modulating over long distances on powergrid? Can be done, but not on my electical bill. > > This feels like I have to use something like 8B/10B with lots of transitions > so the receiver can track the system frequency? > > WSPR is more like it. Cheers, Magnus