South Australia ran for stretches last decade on almost nothing but inverter-based generation, and more than once the frequency wobbled in ways that made engineers in Adelaide and at AEMO’s control room in Sydney distinctly uneasy. That’s not a criticism of the state. It’s the clearest demonstration the NEM has had of a problem that’s about to become everyone’s problem: most of the inverters we’ve bolted onto the grid over the past fifteen years don’t actually know how to run a grid on their own. They know how to follow one.
That distinction sounds pedantic. It isn’t. It’s the whole ball game for what happens after the last big coal units leave the system, and it’s why the phrase grid-forming inverters has gone from a line item in obscure AEMO technical papers to something project developers are now asked about in investment committee meetings.
Grid-following versus grid-forming, properly explained #
Almost every solar farm, wind farm and battery connected to the NEM today uses what’s called a grid-following inverter. Think of it as a very capable follower, not a leader. It measures the voltage and frequency the grid is already running at: set, historically, by the spinning mass of coal, gas and hydro turbines, and matches its output to that reference. It’s good at this. It’s fast, cheap, and it’s why rooftop solar and utility-scale renewables have scaled as quickly as they have.
The problem is that a grid-following inverter has nothing to follow if there’s no strong reference signal to lock onto. Spinning turbines don’t just produce power; their physical mass resists sudden changes in frequency, a property engineers call inertia. Take enough of that mass off the system: through coal retirements, through periods where the grid is running on majority renewables, and the frequency becomes twitchier, more prone to drifting, and less forgiving of a sudden trip or fault.
Grid-forming inverters are built to behave more like that retiring fleet. Using sophisticated control software and, in most designs, a battery or a modest buffer of stored energy to draw on in the first fractions of a second, they can set a voltage and frequency reference themselves rather than waiting for one to exist. They can ride through a fault without tripping offline, and crucially, they can keep a local pocket of grid alive even if the wider network around it has gone dark – something a grid-following inverter cannot do at all.
Why this stopped being theoretical #
AEMO’s Engineering Framework and its Integrated System Plan have both flagged the same structural issue: as synchronous generation exits, system strength and inertia fall in specific pockets of the grid faster than new synchronous plant or synchronous condensers can be built to replace them. I’d point anyone wanting the primary source to AEMO’s own published engineering roadmap, which lays out minimum system strength requirements region by region rather than as one national number, because the NEM doesn’t have one grid strength, it has dozens of local ones, and some corners of western Victoria and far north Queensland are thinner than others.
Western Victoria is the sharpest illustration. It’s a region stacked with wind and solar, thin on synchronous plant, and increasingly reliant on new transmission to move power out. The same logic, on a smaller and more isolated scale, applies to WA’s SWIS grid, which doesn’t get to lean on an interconnector when things go wrong. There’s no neighbour to call, which makes local inertia and grid strength an even more immediate design constraint rather than a long-term planning line item.
Who’s actually building these things #
The Hornsdale Power Reserve, Neoen’s original big battery in South Australia, has been progressively retrofitted and expanded with grid-forming capability: a point Neoen has referenced in its own project updates, and worth noting given how much of the “big battery” conversation still treats all batteries as functionally identical. They aren’t. A battery with grid-forming inverters does a materially different job to one that’s simply following the grid and arbitraging price spreads, and I’d argue the market hasn’t fully priced that distinction yet. We’ve covered the broader question of whether big batteries are being built for the wrong job before, and inverter mode is a big part of that answer.
The Waratah Super Battery in NSW is probably the most visible local example of a battery being built explicitly, from the ground up, as a system-strength and stability asset rather than a pure arbitrage play – see our rundown on Waratah Super Battery finally earning its keep. Akaysha Energy, which built it, has described the project in its own materials as providing essential system services alongside energy trading. That’s the template I expect more large batteries to follow, particularly anything sited near REZ transmission bottlenecks like those discussed in our piece on the New England REZ build.
What it actually buys the grid, and what it doesn’t #
Let’s be careful with the numbers here, because it’s easy to conflate power and energy in a sentence about batteries and inertia. A grid-forming inverter’s contribution to system strength and its ride-through capability are about power electronics and control response measured in milliseconds – not about how many megawatt-hours the battery can store. A 100 MW battery with a one-hour duration and a 100 MW battery with a four-hour duration can offer near-identical grid-forming services in the first second of a fault; the stored energy only starts to matter if the disturbance runs long or if the inverter needs to sustain an island for an extended period. People conflate the two constantly, and it leads to overstated claims in both directions – overstating what a short-duration battery can do for long-duration firming, and understating what it can do for system strength.
A worked comparison helps:
| Attribute | Grid-following inverter | Grid-forming inverter |
|---|---|---|
| Reference signal | Must track an existing grid frequency | Can establish its own voltage/frequency reference |
| Fault ride-through | Often trips offline under severe disturbance | Designed to ride through and stabilise |
| Black-start / islanding | Not capable alone | Capable, within its power rating |
| Dependence on stored energy | Low – follows regardless of duration | Moderate – needs a buffer, but duration isn’t the main lever |
| Cost premium | Baseline | Higher control hardware/software cost, improving with scale |
What grid-forming inverters don’t do is replace the need for firming capacity across an evening peak. That’s a separate, energy-duration problem, the one we’ve gone through in detail in what firming actually means and in the ongoing argument over gas peakers versus big batteries. Grid strength and firming get lumped together in public commentary constantly, and they’re genuinely different engineering problems with different solutions.
The cost question nobody loves answering #
Retrofitting or specifying grid-forming control adds cost to a battery project, not a trivial amount, though the premium has been falling as manufacturers like Tesla, Fluence and Hitachi Energy build the control software into standard product lines rather than treating it as bespoke engineering. The AEMC has been working through rule changes on how grid services like inertia and system strength get procured and paid for, because right now the market doesn’t have a clean, nationally consistent mechanism to reward a battery owner for providing grid-forming services the way it rewards them for supplying energy. Developers are essentially making a bet that the services market catches up to the hardware capability. Some will win that bet. Some will have sunk capital into a capability the market underpays for several years running, and that’s a real commercial risk, not a footnote.
A system strength parallel from somewhere unexpected #
I play a fair bit of chess, badly but stubbornly, and the clearest analogy I’ve found for system strength is the difference between a position with a strong, centralised structure and one that’s technically fine on the clock but structurally brittle. It looks fine until one piece gets traded off and the whole thing collapses two moves later. A grid with plenty of synchronous inertia is the first position. A grid thinning out its synchronous generation without replacing the underlying strength is the second. Grid-forming inverters are one way of rebuilding that structure without putting a spinning turbine back in the ground. Synchronous condensers are another, and AEMO has been procuring those too in some of the weaker pockets of the network.
Where this leaves planning for the next five years #
AEMO’s Integrated System Plan already assumes a meaningful share of future battery and inverter-based capacity will need grid-forming functionality, particularly in regions losing coal units on the timeline set out in our coverage of Yallourn’s 2028 closure and the broader pace mismatch we’ve tracked in NSW’s coal exit outrunning its REZ build. The honest engineering position is that this isn’t an optional upgrade bolted on for environmental credibility. It’s closer to a structural requirement for running a grid that’s increasingly built on power electronics rather than spinning steel, and the regions furthest along the coal-exit timeline are the ones that will find out first, in real time, whether the replacement capability actually arrived on schedule.
Worth remembering too that the AER’s published network performance data gives a sense of how often faults and disturbances already test the margins in thinner parts of the grid. It’s not an abstract risk being planned against, it’s one the system already lives with weekly, just currently covered by a fleet of coal and gas units that won’t all be there in ten years.
– Anjali Rao, Grid & Storage Correspondent
Photo by Raisa Milova on Unsplash