AEMO marks its own homework
Each December, AEMO publishes a Forecast Accuracy Report. The National Electricity Rules require it, and it grades the Electricity Statement of Opportunities (ESOO) published the year before against the outturn. Most of its verdicts are qualitative. AEMO says so directly: "quantitative metrics are not always feasible," and its summary table uses three coloured indicators rather than scores.
Rooftop PV is one place where the report does put numbers side by side. Table 5 in each edition sets AEMO's Central scenario forecast of installed rooftop PV capacity at 30 June beside the estimated actual, region by region, with the error between them. Because the quantity and the basis are the same in every edition, three editions can be read as one record.
Short, shorter, then over
The 2022 ESOO fell 2.2 per cent short of the rooftop PV installed by 30 June 2023. The 2023 ESOO fell further short, by 6.5 per cent, against 30 June 2024. AEMO then revised its forecasts upward, and the 2024 ESOO overshot by 5.5 per cent against 30 June 2025.
| Forecast | Graded at | Estimated actual | Central forecast |
|---|---|---|---|
| 2022 ESOO | 30 June 2023 | 18,003 | 17,604 |
| 2023 ESOO | 30 June 2024 | 20,775 | 19,418 |
| 2024 ESOO | 30 June 2025 | 23,180 | 24,457 |
The regional detail shows the reversal was not confined to one region. Under the 2023 ESOO, four of the five regions were under-forecast, and Tasmania and New South Wales by 12.0 and 11.0 per cent. Under the 2024 ESOO, all five were over-forecast, South Australia by the most at 9.8 per cent.
| Forecast | Graded at | NSW | QLD | SA | TAS | VIC | NEM |
|---|---|---|---|---|---|---|---|
| 2022 ESOO | 30 June 2023 | -3.6% | -2.9% | -0.4% | -1.5% | -0.3% | -2.2% |
| 2023 ESOO | 30 June 2024 | -11.0% | -7.1% | -5.6% | -12.0% | +0.8% | -6.5% |
| 2024 ESOO | 30 June 2025 | +2.0% | +5.8% | +9.8% | +1.2% | +8.6% | +5.5% |
12.0 points
the swing in AEMO's rooftop PV forecast error between two consecutive ESOOs
One assumption, both directions
AEMO's own explanation for both errors points to the same input: the size of the average system being installed.
- Grading the 2023 ESOO, AEMO wrote that rooftop PV "forecasts were under-forecast, with growth in average system size not slowing down as expected," and that forecasts in the 2024 ESOO "have been subject to an upward revision for all scenarios."
- Grading the 2024 ESOO, AEMO wrote that "these errors were due to an over-estimation of future growth in average PV system size," and that "this assumption has been revised down in subsequent forecasts."
The 2025 ESOO records the correction from the other side. It describes its minimum demand forecasts as higher "due primarily to weaker growth in distributed PV than forecast in the 2024 ESOO."
This is the pattern the record makes visible and a single edition cannot. Each report, read alone, describes an error and a fix. Read in sequence, the fix to one year's under-forecast is the cause of the next year's over-forecast.
What these numbers are not
Several limits apply, and they matter more than the headline.
- The actuals are estimates. AEMO states that rooftop PV actuals "are not known precisely at any point in time, and are subject to revision because PV installers have up to one year to submit applications" for certificates. Each actual above was read on a stated date, and may since have moved.
- The horizon is short. An ESOO published in August forecasts capacity at the following 30 June, less than a year ahead. These are errors on a near-term stock, not on a ten-year outlook.
- A revision is not an error, and one parameter is not a model. Three readings of one input, moving in opposite directions, are evidence of how hard that input is to call. They are not a verdict on AEMO's forecasting as a whole, and they are not a trend.
- Rooftop capacity is not the whole distributed PV picture. Smaller commercial systems of 100 kW to 30 MW, which AEMO forecasts separately, are excluded here.
What the record does support is narrower and still useful: on installed rooftop capacity, the error a year out has run to more than six per cent in both directions, and it has changed sign after a method revision.
What settles the next reading
AEMO's December 2026 Forecast Accuracy Report should grade the 2025 ESOO against installed capacity at 30 June 2026, on the same basis, in the same table. That edition will show whether revising the system size assumption down brought the error back towards zero, or carried it past zero again.
That forecast is the natural claim for this publication to put on its Forecast Ledger, so the grading can be marked when it lands. It is not ledgered here. The figure sits in AEMO's Forecasting Data Portal, which refused automated access when this report was written, and the ledger records only figures verified word for word against their source.