Stability of grids with a high percentage of asynchronous sourcing

[AussieBruce]

I live in a country that appears to be going for broke in terms of moving to renewable generation – mostly solar and wind. Having a long background in industrial controls, I’ve tried to get hold of as much information as I can on stability of large, widely extended grids as the proportion of inverter sourcing grows. There’s a lot on the web from Texas (that state has its own grid), UTex puts a lot of informative technical stuff in the public domain. There are also a couple of local sources.

My best assessment from all this is that there is still no proven configuration defined, segments of a grid can run OK on quite high percentage of asynchronous input, but they are always solidly tied to substantial amounts of proven, stable kit in the form of very large rotating generation. I’m also not aware of any instance in the world where asynchronous supply strongly dominates on its own.

What’s peoples view on the current status? Is this likely to be a crunch issue down the track?

[Someone]

but they are always solidly tied to substantial amounts of proven, stable kit in the form of very large rotating generation.

That's anti-renewables propaganda, zero rotating mass/inertia/flywheel is required.

Pick through a recent sealioning of the topic:
https://www.eevblog.com/forum/chat/will-ac-power-distribution-become-obsolete/msg5551181/#msg5551181
100% inverter grid is practical and possible today, small grids already do this and the methods/techniques perfectly scale up.

[woodchips]

Zero mass/inertia might not be needed, but something has to define the base frequency and voltage.

If a sole inverter grid, does each inverter sync to the existing frequency and voltage, or does it just generate at what it thinks the grid should be? What stops a drift? If each inverter runs at its own defined frequency and voltage then are there any circulating currents with a tiny discrepancy with the grid? I suppose the ultimate test is to run a synchronous clock off the grid, if it stays dead accurate then it is stable.

What if a sudden load appears, lightning strike or similar, do the inverters just skip over the transients and just carry on?

The grid is described as an infinite bus bar, nothing that occurs overloads it or upsets it. Is there a minimum size of inverter only grid? Obviously a rotating generator will just continue rotating and self sync.

[zilp]

If a sole inverter grid, does each inverter sync to the existing frequency and voltage, or does it just generate at what it thinks the grid should be? What stops a drift?

The same thing that stops drift in any other grid.

I mean, I have no particular clue as far as grid management is concerned. But obviously, the equations that govern the behavior of rotating mass generators are pretty well understood. And you obviously can build a solid state inverter that just evaluates that equation to derive its control signals, and thus electrically behaves identically to a rotating mass generator, so you obviously can apply the same strategies.

[Siwastaja]

You need to separate differences between fundamental issues (aka showstoppers) and control algorithm details. And as explained by others, grid stability relates to the latter.

It's important to realize whether showstoppers exist or not, because if they don't, then you can pretty much start transitioning, and any small issues get fixed along the way. The limited rate of change, limited by manufacturing capabilities of inverter/cell/panel manufacturing, installation/electrician resources, and people's general willingness, helps in that any such issue is caught in time and rectified. Example of such "small issue" could be an adjustment needed in default parameters that define the inverter's control curves. Or taking some small percentage of inverters (say, a few large PV parks, probably not any consumer customers) under closed loop control.

[Someone]

If a sole inverter grid, does each inverter sync to the existing frequency and voltage, or does it just generate at what it thinks the grid should be? What stops a drift?

Why does drift matter?
Smells like more sealioning, as this was covered in the above linked thread:
https://www.eevblog.com/forum/chat/will-ac-power-distribution-become-obsolete/msg5551223/#msg5551223
Come back when you actually find some problem.

[trobbins]

The control and simulation of stability has been front of mind of grid operators for well over a decade, and the hunt for larger pools of stability analysts is still on-going.  Tools and techniques continue to advance.  Any new grid connect inverter has to come with an approved control simulation model.  Battery grid connect systems like the large Tesla systems are approved for advanced grid support, and I'd anticipate for grid forming.  South Oz often leads the way.  As yet, I don't see widespread commanded disconnects of small generators, and likely for a number of reasons, even though inverters have to include that ability. It's a very large mixing pot, but mixing seems to be slow wrt electric vehicle battery integration.

[nctnico]

If a sole inverter grid, does each inverter sync to the existing frequency and voltage, or does it just generate at what it thinks the grid should be? What stops a drift?

The same thing that stops drift in any other grid.

I mean, I have no particular clue as far as grid management is concerned. But obviously, the equations that govern the behavior of rotating mass generators are pretty well understood. And you obviously can build a solid state inverter that just evaluates that equation to derive its control signals, and thus electrically behaves identically to a rotating mass generator, so you obviously can apply the same strategies.

AFAIK the generators of large (hundreds of megawatts per generator unit) power plants are steered using an atomic clock to make sure that the frequency is stable long term. It is not like these generators are freerunning. The large power plants form a low impedance to which inverters can synchronise. With the large power plants gone at some point, there needs to be a way to keep the grid synchronised to an atomic clock. You can't have a large number of independent inverters pump power into the grid at any phase / frequency they like.

[BarrowBoy]

Some incorrect statements, and some correct in the above. Of course a grid needs to be balanced, in time. That's why we now have negative feed in tariffs.  In the UK, the are trying a novel approach. During peak generation ONLY, EV's get a free re-charge.

Need to know more and stop guessing - https://us.flukecal.com/literature/electrical-calibration/testing-and-calibration-phasor-measurement-units

[Phil1977]

In fact we are talking here about two -partially overlapping- parts of net regulation.

On a time scale of single seconds, you need a certain power reserve that traditionally was given by the rotational energy of the big turbines and generators. 20 years ago a lot of academic studies did research on how it is possible to transfer this behaviour into small inverters, so that also a big mass of small producers is phase-stable without a big "master generator". As far as I know these studies have been widely successful, and these technologies are nowadays obligatory for big wind- and solar farms. They practically simulate the rotating mass of generators with clever circuits and algorithms.

On the next time scale you do regulation by switching loads and controlling production power where possible. This is established by higher-level communication, e.g. by internet or radio control. This is where you can catch load peaks by offering free EV-charging or ramping down a gas turbine.

[Geoff-AU]

https://www.aemo.com.au/Electricity/National-Electricity-Market-NEM/Security-and-reliability/-/media/80004354B88B4FEF96EAB673DDC99820.ashx

AFAIK the generators of large (hundreds of megawatts per generator unit) power plants are steered using an atomic clock to make sure that the frequency is stable long term. It is not like these generators are freerunning. The large power plants form a low impedance to which inverters can synchronise. With the large power plants gone at some point, there needs to be a way to keep the grid synchronised to an atomic clock. You can't have a large number of independent inverters pump power into the grid at any phase / frequency they like.

From the above PDF, the frequency's normal range control band is +/- 0.3%.  Back when people wore watches, the watch crystals are accurate to 20ppm.  Atomic clock not required.  Over a short timescale, big power plants are free-running.  But they have such large inertia that load changes take seconds to have an impact.  Over a long term (several days) you can try and keep time with an external reference but it's absolutely not necessary in a world where nobody uses synchronous clocks any more.

Grid-tie inverters cannot just "willy nilly" export at whatever phase they want, the circulating currents of an inverter trying to do that would let magic smoke out.  Therefore, because they stay synchronised, they cannot "drift" away from the grid.  The grid as a whole can drift, but individual generators have to stay synchronised.

Grid-tie inverters synchronise with the existing waveform, and then start exporting in-phase (by locally raising the voltage ever so slightly, thus managing to push current back out).  If voltage goes too high, they back off (Volt-Watt control). 

Modern grid-tie inverters also have Volt-Var mode.  This absorbs or generates reactive power to limit voltage excursions.  It also produces a leading or lagging effect on the grid, that is to change the frequency by a tiny amount.  The inertia of the rest of the system limits the impact.

The next step is to have some kind of frequency-reactive control.  AEMO already has FCAS which will turn generators on or off, or loads on or off, if the frequency goes outside bounds.  We would need something similar in distributed generation to ramp up generation when frequency is low, and ramp it down when frequency is high.  Then load-shedding sort of disruptions can be avoided.

As others have said, it's just a control system issue.  Very easily solvable (especially since it's someone else's problem  ).  There will likely be economic incentives for EVs to export to the grid when the grid needs energy.

Zero mass/inertia might not be needed, but something has to define the base frequency and voltage.

Yes, then you need a grid forming inverter.  A single inverter can do what it likes (if it has the capability).  As soon as you have more than 1, they need to synchronise. 

What stops a drift? If each inverter runs at its own defined frequency and voltage then are there any circulating currents with a tiny discrepancy with the grid?

As above, you cannot be substantially out of phase because the circulating currents will make a rather exciting light and noise show.  But yes, on a smaller scale you circulate a lot of current trying to fight the grid.

I suppose the ultimate test is to run a synchronous clock off the grid, if it stays dead accurate then it is stable.

Doesn't really matter.  The grid can wander up and down in frequency all the time and still be stable.  The point is the rate at which it wanders, and you won't pick that up with clock-watching.

What if a sudden load appears, lightning strike or similar, do the inverters just skip over the transients and just carry on?

An inverter "just" skips a direct strike in the same way a human "just" keeps playing golf after being hit.

Depends on the size of the disturbance.  Small blips will sail through, bigger blips will cause an inverter to disconnect, and it might cause catastrophic failure of the nearest few inverters.

The grid is described as an infinite bus bar, nothing that occurs overloads it or upsets it.

By whom?  Very simplified view, and that's useful for some analogies but important to know the limitations of such a simplistic understanding.

[Siwastaja]

I suppose the ultimate test is to run a synchronous clock off the grid, if it stays dead accurate then it is stable.

It never was, and will never be. Clocks syncing onto mains were a thing of the last century, but they happily drifted by, say +/- 30 seconds, maybe even more. There was (probably still is) effort to keep grid very close to nominal frequency long term, that is, in annual time scale, so that those clocks never need adjustments, if you are happy with significant back-and-forth drift. On the other hand, for people who like their clocks to be in correct time within a second or two, never were able to use those synchronous clocks because grid never was accurate enough on the short scale.

AFAIK those retro clocks (maybe such circuitry is still used in microwave oven clocks or something, I don't know) are the only reason for maintaining long-term accuracy of the grid. Probably not many people would complain at all even if we changed grid frequency to say 50.1Hz overnight.

[Phil1977]

Clock syncing seems something aesthetic - if you have resources to do it then you may do it because it´s beautiful to know your grid frequency is that perfect. And old clocks run precisely.

There´s not fundamental reason why you need it for power regulation or stability.

[nctnico]

I suppose the ultimate test is to run a synchronous clock off the grid, if it stays dead accurate then it is stable.

It never was, and will never be. Clocks syncing onto mains were a thing of the last century, but they happily drifted by, say +/- 30 seconds, maybe even more. There was (probably still is) effort to keep grid very close to nominal frequency long term, that is, in annual time scale, so that those clocks never need adjustments, if you are happy with significant back-and-forth drift. On the other hand, for people who like their clocks to be in correct time within a second or two, never were able to use those synchronous clocks because grid never was accurate enough on the short scale.

AFAIK those retro clocks (maybe such circuitry is still used in microwave oven clocks or something, I don't know) are the only reason for maintaining long-term accuracy of the grid. Probably not many people would complain at all even if we changed grid frequency to say 50.1Hz overnight.

You'll be unpleasantly surprised. Every mains powered alarm clock and every other household appliance with a clock you can find in the shop is synchronised to the mains frequency because it is much more stable compared to a crystal. A couple of years ago 'our' street was running from a diesel generator for over a week because a distribution transformer blew up; the clocks where off by many minutes within two or three days.

With a 50ppm crystal oscillator, a clock can be off by over 2 minutes a month. There is a good reason equipment relies on external time sources which are referenced to UTC instead of internal crystals when possible. Computers use NTP by default, mobile phones use the time from the wireless network.

Where it comes to the grid frequency stability: there has to be some common ground on where to synchronise the grid to where it comes to large scale grid sources which drive the grid frequency for smaller sources to follow / track. Whether it is an NPP, gas/coal plant, storage facility, large solar park or wind farm, synchronising it to a UTC derived time source is the obvious choice. Long term stability comes with it for free.

People can keep going on about long solved control algorithms but in the end they all rely on a feedback loop and thus are prone to running into stability problems when the phase margin of the control loop gets thin.

[Andy Chee]

Here's some data collected on grid frequency stability by some time nut:

http://www.leapsecond.com/pages/mains/

In general, grid frequency is extremely accurate over 24 hours, but can and will drift up and down during the day as industrial and commercial machinery is turned on and off.

As mentioned above, the Australian grid has a defined specification of frequency tolerance range (as well as the more familiar voltage range 230V +10%/-6%).  Your local grid will also have a defined performance specification, I encourage you to look it up.

Do not assume your grid is synchronised to an atomic standard.  Look up the specification.

https://www.aemc.gov.au/sites/default/files/2024-01/Frequency%20Operating%20Standard.pdf

[Geoff-AU]

Do not assume your grid is synchronised to an atomic standard.  Look up the specification.

https://www.aemc.gov.au/sites/default/files/2024-01/Frequency%20Operating%20Standard.pdf

Precisely.

REQUIREMENT                                MAINLAND   TASMANIA
1 Accumulated time error limit…            no limit   no limit.

Our grid is not required to keep time.

[Siwastaja]

Timekeeping is a convenience thing, and mains-operated alarm clock radios, while nice retro thing from 1980's, are dying out as the people who own them die. For microwave oven clocks, I would prefer them not to exist at all and I'm sure 97.5% would agree. In any case the grid never was accurate for anything else than very crude minute-level time keeping (which is why those mains-operated clocks almost never show seconds at all).

Anyway, this timekeeping synchronization thing already existed in 1960's, probably even before, and AFAIK was done manually by the operators steering some of the larger plants. Like "oh, today it's 30 seconds off, let's adjust it this way..."

This behavior is trivially easy to replicate on our imaginary inverter-only grid. I'm not sure if it's 1% or 10% of capacity which needs synchronization, but definitely nowhere near 100%. And just like in the 1960's, this does not involve super fast communication; no, it's more like a PI controller (or a super slow PLL) where adjustments are made in hourly/daily timeframe.

All this can be implemented with a $10 GPS module and less than 100 lines of code, and only small part of producers need to implement it anyway (and it's not like the grid would collapse even if no one did that), so it's hilarious how some people paint this as some kind of showstopper.

[nctnico]

GPS is the worst choice ever in today's world. An attacker can destabilise the grid using an SDR which can be bought for $100 from Aliexpress. Grid operators are actively moving away from using GPS for grid synchronisation because it is too vulnerable.

[Phil1977]

GPS is still a good choice for reference. I´m very sure a grip operator does a plausibility check of the time data and if it´s out of tolerance falls back to some TXCO or whatever.

In that case your Aliexpress powered super villain can only delay the street lighting by several microseconds. And in 99.99% of time while humanity is good you have a really precise reference clock.

(PS: Probably the different levels of grid synchronization have been confusing again. Using a GPS-disciplined-XCO for phase control of the generators may seem feasible, but really would be dangerous. Using it as the time master in the grid operations center together with some backup should be easy going.)

[Siwastaja]

destabilise the grid

Note to others: nctnico's "destabilising the grid" means making some already inaccurate retro clocks slowly drift, timekeeping of which is not even currently being guaranteed in many (most?) countries.

[Phil1977]

destabilise the grid

Note to others: nctnico's "destabilising the grid" means making some already inaccurate retro clocks slowly drift, timekeeping of which is not even currently being guaranteed in many (most?) countries.

That´s exactly the confusion I meant:

- Destabilising the phase synchronization of single generators --> chaos and destruction
- Destabilising the long-time-frequency regulation --> deviations at old clocks

I think that nctnico thought of (old) plans to stabilize a grid by giving each inverter a GPS disciplined oscillator. The accuracy would be fair enough to ensure phase synchronization, but in that case a "jitter" of only a few milliseconds could indeed break down the whole grid.
Nowadays the inverters are syncing themself with the grid without a frequency master. So GPS deviations wouldn't make any transformer burn.

[nctnico]

destabilise the grid

Note to others: nctnico's "destabilising the grid" means making some already inaccurate retro clocks slowly drift, timekeeping of which is not even currently being guaranteed in many (most?) countries.

No, not at all! As usual you are not reading /understanding correctly at all and are just gaslighting.

You are confusing two things here as the discussion moves back & forth a bit. But I'll clarify for you:
1) big plants which drive the mains frequency need some reference to synchronise themselves to as they have no other reference for guidance. Small inverters can track the big plants but these big plants will need to form a low enough impedance. This is simple control theory 101.

2) Mains frequency is maintained to a certain frequency stability making it suitable for use as a frequency reference for clocks. As I wrote before, it is not retro clocks but modern digital clocks use mains frequency as their reference source as well. So maintaining a degree of frequency stability of the grid will be handy otherwise hundreds of millions of clocks will need replacing. Ofcourse this depends on the region where you are in. Ofcourse you are free to find it important or not but that is just your opinion. Facts are facts.

[Someone]

I think that nctnico thought of (old) plans to stabilize a grid by giving each inverter a GPS disciplined oscillator. The accuracy would be fair enough to ensure phase synchronization, but in that case a "jitter" of only a few milliseconds could indeed break down the whole grid.
Nowadays the inverters are syncing themself with the grid without a frequency master. So GPS deviations wouldn't make any transformer burn.

Frequency accuracy vs integral time error at some non specific period are nonsense arguments. Grid synchronisation and stability do not require anything more than a few cycles of the mains (100ms or so) and a 100ppm or so absolute accuracy.

What GPS disciplined oscillator fails to maintain 100ppm accuracy during loss of satellite communications ? Laughable.

No-one is trying to blindly synchronise generation without looking at the local phase.

[Phil1977]

You are confusing two things here as the discussion moves back & forth a bit. But I'll clarify for you:
1) big plants which drive the mains frequency need some reference to synchronise themselves to as they have no other reference for guidance. Small inverters can track the big plants but these big plants will need to form a low enough impedance. This is simple control theory 101.

This simply does not get true by repetition. "Master-less" synchronized networks need no central guidance. They can be and are stable just by solving the underlying differential equations the right way and implementing their solution into the individual control loops.

It´s like waves on the ocean. There´s no "master-wave" that dictates the water molecules to go up or go down. But the system of inertia and dynamics enables them to stay in time where ever they are.

No-one is trying to blindly synchronise generation without looking at the local phase.

Yes, but that´s exactly what would be required if a grid should break down by one failing oscillator.

[Someone]

You are confusing two things here as the discussion moves back & forth a bit. But I'll clarify for you:
1) big plants which drive the mains frequency need some reference to synchronise themselves to as they have no other reference for guidance. Small inverters can track the big plants but these big plants will need to form a low enough impedance. This is simple control theory 101.

How are big or small plants inherently any different? Either could produce accurate frequency control. Modern electronics make that practical for even small residential inverters.

2) Mains frequency is maintained to a certain frequency stability making it suitable for use as a frequency reference for clocks. As I wrote before, it is not retro clocks but modern digital clocks use mains frequency as their reference source as well. So maintaining a degree of frequency stability of the grid will be handy otherwise hundreds of millions of clocks will need replacing. Ofcourse this depends on the region where you are in. Ofcourse you are free to find it important or not but that is just your opinion. Facts are facts.

Some countries choose to do this. It is not mandatory or necessary, or anything to do with grid/supply stability. One way to do this is to have generators change their operating setpoint (power output vs frequency line), which can be done slowly in some centrally co-ordinated matter if required (as it is currently done in Australia). Again residential (or larger scale) inverters could participate in this if there was economic reason to do so, it is not something preventing the share of inverter sources being increased, even to 100% of the grid.