The following article was written by a political reporter but contains technical aspects of frequency control that Smurf may like to comment on.
Translating it to layman's terms:
An AC power system (eg the grid) runs at a specified frequency which in Australia is 50 Hz. So that's 50 complete cycles of the AC sine wave each second.
A key concept is that anything sychronised to the grid will be at the same frequency. So the frequency at Loy Yang power station (Vic) will the same as someone could measure in their home in Brisbane since it's all one big interconnected AC system.
Now, the frequency needs to be controlled pretty tightly. A 1% variation is pushing the limits of what's acceptable. If it's 2% then that's an incident and warrants a proper engineering investigation and report. If we're talking about 5% or more away from where it should be then at that point a total system collapse is a very definite possibility since neither power stations nor connected loads will be happy at that level.
So it's like saying that you're driving a car at 50 km/h. Slow below 49.5 km/h and that's a problem. Below 49 and you'll find yourself being thoroughly investigated as to what went wrong. Below 47.5 and there's a risk that you've just wrecked the car and at the very least it's now coming to a complete standstill, getting it going again will take hours or days.
So it's all pretty unforgiving and there's not much room for error.
If generator output exceeds load then frequency will rise across the entire system. The opposite will occur if generator output is less than load. This all happens pretty quickly - seconds or at the most minutes depending on the severity of the mismatch.
All of which brings about two key requirements for generation:
1. It must match the combined load of all consumers in real time. Therein lies the problem with wind and solar with their uncontrolled and constantly varying outputs versus the directly controllable output from coal, gas or hydro.
2. The more inertia you've got, the better.
Inertia? In simple terms it's just physical mass. You've got a great big steam, gas or hydro turbine with a lot of metal in it. That physical attribute means there's energy stored simply by virtue of it turning and it takes quite a bit of effort to slow it down. Lots of big rotating machines thus naturally tend to stabilise system frequency.
As a concept that's very similar to a fully loaded freight train. It slows down only gradually even if someone were to completely cut power (from whatever source eg diesel) to the engines. The train will travel quite some distance before it completely stops. Even applying the brakes it still takes a while because all that energy, inertia, has to go somewhere (dissipated as heat in this case).
In contrast a solar panel and most wind farms contribute absolutely zero inertia to the grid. Zero. Even though wind does involve big rotating machines, they're not synchronised to the grid directly - and it's hard to do that with wind because if they were synchronous machines then they'd always be turning at the same speed.
There are some partial workarounds to create "synthetic" intertia at wind farms using fancy electronics. That helps but (1) no amount of electronics is at present a full substitute for the sheer physical mass of a great big rotating steam, gas or hydro turbine and alternator and (2) most wind farms don't have that at all since it costs more $ to install, means more things to maintain, and produces no extra revenue.
So a grid being powered predominantly by solar or wind becomes electrically weak. It can still be delivering the exact same quantity of power as if it were supplied from coal, gas or hydro but without those big rotating machines there just isn't the inertia to stabilise frequency. It becomes like a motorbike rather than a freight train - speed increases or decreases very easily.
If there's low intertia then the room for error in matching generation to load decreases accordingly and generation now needs to respond far more quickly to changes in load. Problem is, the very same circumstances which create that low inertia, high use of wind and solar, are the same generation sources which aren't much good at varying their output in response to changes in load.
This creates a situation where the remaining synchronous generators, that's coal, gas, hydro and any oil-fired plant that happens to be running, are now in a situation of generating (say) 30% of the power but having to provide 100% of the response to changing load
and they've got to respond to changing output from wind and solar as well. If load goes up at the same time as wind and solar output falls, or vice versa, then that's a pretty tough challenge for the few synchronous machines to cope with.
Overall it's a bit like any situation where only a small % of the team are really playing the game and everyone else is along for the ride. Works only as long as it doesn't get too difficult but it's pretty easy to end up in a situation where those seriously playing are "swamped' and just can't cope, after which the whole thing collapses.
As analogy, suppose that you were given the task (with full backing of the law etc so all legit) of stabilising the ASX20. Your job is to keep the index precisely flat overall, with daily variations no greater than 0.5%. You can buy or sell any stock in the index you like, you've got as much cash as you'll need and you can order any company to issue new shares at any time of your choosing so as to dilute their value. Sounds pretty easy doesn't it? Just buy or sell all 20 stocks as needed to keep the price flat. So easy you could do it sitting on the beach just using a smart phone.
Now suppose I change the rules of that game and that you must still keep the ASX20 flat but the only stock you can trade is BHP. Nothing else. You can buy as many BHP shares as you like but you can't increase their price to anything above the historic all time high and you can't decrease it below 20% of its all time high. And you can't do anything at all about the other 19 stocks in the index.
That's going to work just fine so long as the other 19 stocks only move by fractions of a % but you're completely screwed if there's a decent move. Doubling the price of BHP, whilst the other 19 all fall 20%, isn't going to keep the index stable. It might work this week or even this year but pretty clearly it's going to fail at some point given that you have no control over most of what you're trying to stabilise.
That latter scenario is comparable to the dilemma faced by synchronous generators in a system with heavy use of wind and solar. They can keep frequency stable only as long as nothing major happens. The moment we get a surge in load combined with the wind rapidly diminishing they're pretty much stuffed and the whole system falls in a heap.
Under normal circumstances only the SA and Tasmanian grids are ever electrically "weak".
SA - because at times wind can supply more than 100% of the system load.
Tas - because under certain circumstances there's a single very large generation source (Basslink supply from Vic) which generally occurs at the same time (middle of the night when power is cheap in Vic so Hydro Tas is a keen buyer) as overall system load is low and just 4 factories are using 70% of all power consumed in the state.
In Tas the workaround is an elaborate and unique (custom designed and built) system which dumps specific industrial loads in the event of a non-synchronous generation (Basslink) failure before the rest of the grid even "sees" that anything happened. That effectively contains any sudden failure to only affecting specific industrial loads with neither the hydro system nor other consumers "seeing" any impact at all.
In Tas there's also a similar scheme to make the GGCT (gas) plant seem smaller than it is in the event that it suddenly fails. So if the CCGT fails then the hydro system takes about 69% of the "shock" and the rest is dissipated via a specific industrial load. That was done to eliminate risk to the rest of the system if the worst does happen.
It wouldn't be impossible to have the hydro system absorb those shocks, failure of either Basslink or the CCGT, by the way and technically it can certainly do that. Trouble is that doing so means running a lot of machines at low output for extended periods when 99.9% of the time there's no need (knowable only in hindsight). That costs money through loss of efficiency and unnecessary wear on machines so the workarounds were done for economic reasons with the benefits being shared between Hydro and those industries forming part of the scheme.
The underlying issue with Tas is simply that Basslink is the only link to other states. Any failure, even just a random trip for no real reason (and that's an inherent risk with DC interconnectors especially those not running parallel to an AC link), thus results in the Tas system becoming completely independent of the other states. That's not a problem in itself, we just did 5 days running separately due to a planned outage of Basslink, but it creates technical difficulties if such a failure occurs suddenly without warning when Basslink is operating at a high level of transfer. Hence the workarounds to keep the system stable if that occurs (and there have been far more incidents than those reported publicly although most are simply a trip - best analogy would be having to restart a computer after it locked up but after restarting it then works perfectly and there's nothing actually wrong with it as such).
In SA the present "workaround" involves large scale blackouts. They don't have any elaborate control systems, just the normal UFLS (Under Frequency Load Shedding) that every grid has but nothing else. Beyond that, all they can do is "export" the problem to Victoria via drawing heavily through the interconnector but that has limits and with the system collapse last year those limits were exceeded to the point that the interconnector shut down to protect itself from damage due to being heavily overloaded. There was another very close "near miss" incident more recently when the explosion happened at Torrens Island - the interconnector was certainly overloaded for a brief period and came pretty close to the "throw my hands in the air and give up" point.
End result is that if something goes wrong when the grid is in an electrically weak position in SA then the lights go out in a big way if the problem is too large to be stabilised via Victoria or whatever synchronous (gas or diesel in SA's case) generation happens to be running at the time.
Qld, NSW, ACT, Vic don't presently have these issues since:
1. At no time is wind, solar or other non-synchronous and controllable generation dominant. The majority of supply being from dispatchable synchronous generation - coal, gas, hydro and on occasion minor amounts from oil. They don't presently have the issue with very high levels of wind and solar, as a % of the total, as occurs in SA.
2. Generally no issues with loss of interconnection. Vic and NSW are strongly interconnected at AC via the Snowy scheme and that applies even if all Snowy power stations are idle at the time (as they commonly are overnight). Qld - NSW are connected by two AC and one DC circuit which are pretty reliable in practice. ACT is simply part of the NSW system, any distinction between the two being purely administrative rather than technical.
3. No individual loads or generation sources account for a large portion of the total as occurs in Tas.
WA (south west) system is completely separate to anything else and also reasonably robust. There is wind and solar but it's not dominant to the extent that it is in SA. That said, as a smaller system not connected to anywhere else it does have some vulnerabilities, a problem can't be shared with another state but must be dealt with locally, and that has caused some issues in the past but overall it's fairly robust.
NT has multiple small systems, each completely independent of each other and not electrically connected. Darwin and surrounds is the most significant and does have the vulnerability of high reliance on Channel Island power station and the gas supply to that but its overall track record isn't bad. No real issues with non-synchronous or non-controllable generation.