By Phil Kreveld
Australia’s energy transition has been thrown a curveball in the form of a rapidly expanding data centre sector—a behemoth that has its sights set on deployment Down Under. How we feed this power-hungry beast, and how it will behave in the grid, has become a subject of great debate.
Australia is an attractive prospect for data centre developers, given our stable regulatory environment, abundant land, renewables potential, and proximity to the Asia-Pacific market. But an unfettered rollout of data centres and AI factories represent a real risk to both the grid, energy consumers, and Australia’s net zero ambitions.
With this in mind, the Commonwealth Government is in the process of developing a set of nationally mandated standards for data centres, with Energy and Climate Change Minister Chris Bowen recently announcing the government would “legislate binding national standards for data centres across the country without exception, which will require 100% renewables backed by firming”.
This has been a sticking point at recent National Cabinet meetings, with Queensland Premier David Crisafulli and Northern Territory Chief Minister Lia Finnochiaro gunning for the use of gas and coal to power data centres proposed for their patches.
But the National Cabinet’s concern over how electricity for data centres comes about is misplaced. We know that wind and solar are here for keeps—and a bit of coal and gas to help things along doesn’t destroy carbon emission reduction. A far more pressing concern is the weak state of our transmission grids.
Related article: Anthropic inks lease deal for Australia’s largest data centre
Our Achilles heel
Weak grids cause the Australian Energy Market Operator (AEMO) conniptions. It is the reason why connecting large data centres seems to take forever. This is the real kicker—never mind where the electrons came from.
As to more electrons being needed for data centres—sure, that is reasonable concern. However, a glance at AEMO’s integrated systems plan indicates oodles of over-capacity by 2050 (15 times today’s annualised yearly average power flow. Therefore, we ought to put out the welcome mat and let data centres soak up some of that capacity.
Grid weakness is a two-way street. In one direction you have connection of a large, power-gobbling data centre, which makes control of frequency and voltage difficult. In the other direction you have the data centre’s own uninterruptible power supply (UPS), which is required to manage voltage and frequency ‘wobbles’.
The UPS then takes over—and turns the wobbles into much bigger ones for AEMO to control by disconnecting the data centre from the grid. This is why connection approvals can take years.
We cannot do much about weak grids—it is the result of having long transmission lines—and our data centre connects to transmission or sub-transmission.
Large changes in power demand or generation cause voltage and frequency instability. This is highly disruptive to the operation of rooftop solar in distribution grids and thus must be avoided—if not all costs, than certainly at costs sheeted home to large data centre developers.
They are required to ‘insure’ their grid connection by stabilising the voltage and frequency at their connection point—in addition to their UPS.
Synchronous generation
The best stabilisation is achieved with synchronous generators—big spinning machines that are far superior to battery storage grid-forming inverters.
Meanwhile, data centres will be required to provide their own battery storage system and grid-forming inverter, housed in their grid-connecting substations—all of which comes at a significant cost.
Energy arbitrage is made tough because that battery energy system must have a low state of charge so that it can absorb a huge dose of power if a grid wobble forces its disconnection.
In other words, AEMO wants data centres to ‘ride through’ times of voltage or frequency instability—but if they can’t, then the BESS-inverter must step in and provide the stability at the point of connection.
But if we toss out the above and instead of a battery, we store heat—i.e., a ‘battery’ storing kilo-calories—then by means of a heat exchanger, a turbine-generator could be energised. The heat, from electrical heating elements, stored in graphite blocks, would raise steam to run the turbine-generator.
This would provide secure power for the data centre (the stored heat is ‘off-grid’, unaffected by grid wobbles). But there’s more, because the synchronous generator could stabilise the connection point in a beautiful, very tightly controlled way, to make AEMO’s day.
In this scenario, during normal operation the synchronous generator runs as a synchronous condenser, stabilising the grid at its connection point. Everyone is happy.
Uh oh! A grid problem occurs, and the synchronous condenser becomes the generator and supplies the data centre—and maybe a suburb or two. To make the climate-conscious government happy, the electricity required for heating the graphite blocks could come from wind and solar plants via power purchase agreements.
But wait, there’s more!
The exhaust heat from the generating plant can be used to vaporise refrigerants required to cool data centres, thus lowering or possibly even avoiding cooling water needs. So, in this scenario, the thermal plant supplies the data centre 24/7 and, depending on its rating, other loads connected to the same bus. This provides the basis for a meshed grid and would therefore begin to strengthen national networks utilising synchronous generators.
To recap, we have weak grids—data centres have no patience with weak grids—AEMO has no patience with skittish data centres and wants them to ‘ride through’, no matter what. AEMO loads the ‘fix’ onto data centres because it is essentially powerless to spend money to fix grid strength, so it makes life difficult for anyone with a large load connection.
Data centres are not allowed to run their own show. They must connect to the grid, and they have to invest in grid-strengthening plant as a condition for connection. So, they buy into large BESS with double-bridge converter-inverter topology. And guess what? Unless there is synchronous condenser nearby, grid stability isn’t really fixed nearly as well as with old-fashioned synchronous generation because that can save the bacon when there are large transient disturbances.
To make this point clear, think about the distinctions between the two forms.
• Inverters: over current limited to maximum of 1.5 times for milliseconds
• Synchronous generators: 3 to 5 times for seconds
• Inverters: loss of synchronism because of current limitations
• Synchronous generators: extremely strong synchronising torque.
We are focussing on data centres—not on electrical engineering of transmission networks. Yet, deployment of more ‘green’ synchronous generation would do a lot for our transmission network—the more synchronous generation we can stick into the 3000 or so buses in the south-eastern grid, the happier AEMO will be. It knows only too well that those very long transmission lines between renewable energy zones and load centres constitute a stability nightmare. So thermal storage can soak up over-capacity in wind and solar, support grid stability hugely, take the pressure off going to the syncon store, and run the ‘flavour de jour’—data centres.
A look under the bonnet
The thermal storage-synchronous technology described in outline is plenty-sophisticated but highly doable. A bit of intestinal fortitude to commission a pilot plant here in Australia is needed. Why? Because we are the focal point for international renewables studies. And we have serious problems—take South Australia, the wunderkind of renewables. It simply cannot withstand an islanding event because it is short of synchronous generation.
We now have an opportunity to be a standout with a pilot plant for data centres. A pilot plant would involve in addition to the thermal storage, steam or other thermal fluid for example CO2, driving a turbine-SSC (synchronous, self-shifting clutch)-connected flywheel-synchronous generator. Under no-load the turbine would be spinning a smidgin under synchronous speed and through the clutch, energise the synchronous condenser to now function as generator. The flywheel would ‘keep the show on the road’ while the change-over is occurring.
Related article: How will data centres impact Australia’s environment? The irony is, we need more data
Thinking outside the box
To dismiss the thermal storage-synchronous generation idea as some harebrained scheme would be an enormous disservice. To claim such a thing can be easily countered because a 100% inverter-run grid is a theoretical concept. Succinctly then; if you populate the grid with synchronous generation as described, you:
(a) Improve voltage and frequency stability, no end
(b) Keep all your protective relaying in place
(c) Reliably operate your existing circuit breakers (that’s the over-current capability)
(d) Run all your inverters on phase-locked loop basis and not have to worry about transient response (a la Adelaide, a la Spain).
The reason this technology has not yet been trialled is in part because of the ‘NIH syndrome’ best explained by two scientists walking out of a meeting, with one saying to his colleague, “Thank God, we heard nothing new”. The ‘Not Invented Here’ syndrome had ARENA turn down funding because CSIRO already has a thermal project—solar reflectors beaming on to a molten salt reactor, capable of energising a steam turbine generator.
This is a shame, as the technology described above is likely to be any amount more compact, and therefore rapidly spread throughout the strategic buses in the grid—leading to ‘meshing’, and enhancing reliability. Information on the technology was obtained from Quantum Graphite Limited and The Sunlands Co (they formulated a project for ARENA’s consideration). As far as can be ascertained, there is no similar project afoot in Australia.
Thoughts? Share your views with us at editor@esdnews.com.au
