As Australia's Grid Leans on Self-Forecasting, a Vendor Glitch Already Caused a Frequency Excursion

As Australia's Grid Leans on Self-Forecasting, a Vendor Glitch Already Caused a Frequency Excursion AEMO is onboarding participants to submit their own dispatch forecasts to manage a renewable-heavy NEM — but a newly released incident report shows how a single self-forecasting error can move system frequency. The Australian Energy Market Operator has released a Reviewable Operating Incident Report titled "Self-forecasting errors and frequency excursion on 19 August 2025," documenting a frequency spike caused by a self-forecast vendor glitch.² The report matters because it puts a concrete operational failure behind a structural shift in how AEMO runs the National Electricity Market — a growing reliance on forecasts submitted by market participants rather than produced solely in-house.³ The shift is deliberate and broad. AEMO's operational forecasting work includes integrating demand-response units into its forecasting systems and onboarding self-forecasting providers so participants can submit their own dispatch forecasts.³ Devolving forecasting to the parties closest to each asset can improve accuracy for a grid filling with distributed and variable generation, but it also distributes the points at which an error can enter the dispatch process.³ The August incident is the cautionary case. When a self-forecasting vendor's output glitched, the error propagated into the dispatch and forecasting systems far enough to produce a frequency excursion — a real-time deviation in the balance between supply and demand that system operators work hard to prevent.² A forecasting input error becoming a frequency event shows how tightly self-forecasts are now wired into operational control.³ The stakes are systemic, not administrative. Forecasting and planning information supports the electricity industry and government in decision-making and guides investment in Australia's electricity markets and infrastructure.¹ When the same forecasting pipeline feeds both long-term investment signals and real-time dispatch, the quality of participant-submitted forecasts becomes a question of system security, not just market efficiency.¹ The governance challenge is verifying inputs AEMO does not fully control. Onboarding third-party self-forecasting providers means trusting external vendors' models and data feeds, and the August excursion shows that a single vendor's malfunction can reach the frequency of the entire system.² The reporting obligations AEMO maintains — including providing forecast information for power system incident reports — are part of how it traces such failures after the fact.³ The trade-off is the heart of the matter. Self-forecasting promises better granularity for a NEM managing more rooftop solar, batteries and variable wind, but it widens the surface area for error, and the operator must build validation and fallback around inputs it did not generate.³ The August incident is the evidence that this surface area is already being tested in real time.² The signal to watch is how AEMO tightens validation of self-forecasting inputs in response to the incident report, and whether the accuracy gains from participant forecasting outweigh the new operational risks as adoption widens.^2,3 If the controls hold, self-forecasting helps a renewable-heavy grid dispatch more accurately; if vendor errors keep reaching system frequency, the model that distributes forecasting also distributes the risk of the next excursion.¹