AEMO frequency monitoring report provides some details of frequency event (due to self-forecast glitch) on 19th August 2025

On or before 13th November 2025, the AEMO published its ‘Frequency Monitoring – Quarter 3 2025’ report.

I’d just like to call out the content that AEMO has included specifically with respect to the frequency spike on Tuesday 19th August 2025 (which was caused by a glitch at one particular vendor of self-forecasts).  On p17/35, the AEMO wrote:

‘The mainland had one event where frequency was outside the NOFB for more than five minutes this quarter without an identified contingency event. This event occurred on 19 August 2025, due to a self-forecasting error with a single forecasting provider that affected several semi-scheduled generators. The self-forecasting error event led to affected units to be dispatched with a 0 MW target. However, many of the affected units continued to generate at levels close to their target in the previous dispatch interval. This led to an oversupply of generation causing system frequency to increase and exceed the NOFB as shown in Figure 9. Once the provider detected the issue, the self-forecasts for all identified impacted semi-scheduled generators were suppressed by the provider resulting in the dispatch forecasts being sourced from AEMO’s wind and solar forecasting system (AWEFS and ASEFS), mitigating the issue. ’

… and then included Figure 9 on the following page:

Then on p24/35, the AEMO wrote the following:

On 19 August 2025 at 1152 hrs the frequency exited the NOFB. The maximum mainland frequency reached 50.21 Hz at 1206 hrs. As detailed in Section 2.2, the problem arose from a self-forecasting error involving multiple semi-scheduled generators across the NEM. An overview of the sequence of events is provided below.

– The incident started when AEMO received availability forecasts of 0 MW for various renewable sites leading to their dispatch at 0 MW. However, many of these sites continued to generate nearly at their target in the previous dispatch interval, resulting in a total MW error from the target of approximately 1,700 MW at one point.

– The self-forecast provider, having identified the error, suppressed all self-forecasts, which allowed AEMO’s systems to automatically revert to its internal forecasts ensuring accurate inputs for dispatch.

– At the same time AEMO increased the mainland lower regulation FCAS requirements.

– Following AEMO actions and the self-forecast provider suppressing all self-forecasts, the frequency stabilised within the NOFB after 1,139.2 seconds.

– Once this issue was identified following the incident, these sites were transitioned by AEMO to AEMO forecasts as a corrective action and the self-forecasts remained suppressed until AEMO was satisfied that there was no longer enhanced risk to the dispatch self-forecasts.

– Additionally, the self-forecast provider responsible for the error implemented immediate changes to its deployment process to mitigate the risk of recurrence or similar incidents, along with a plan for continuous improvements to all processes.

– An analysis is currently being conducted regarding PFR performance of all the relevant plants in NEM and for all participants enabled for L1, L6, L60 and L5 services. After AEMO finalises the performance assessment, it will provide further insights in the next quarterly report.

This event differs from traditional contingency events recorded in the NEM, as it entails a system-level failure beyond AEMO’s control. At present, there are no explicit specifications for frequency performance during such an occurrence; however, AEMO has proposed that the Reliability Panel consider establishing one, in the Technical Review of the NEM Frequency Control Landscape report

We thought readers here might find that of interest.