Self-forecasting refers to the process of estimating (and providing to the market operator) a unit’s Unconstrained Intermittent Generation Forecast (UIGF) for the next dispatch interval.
(A) Purpose of Self-Forecasting
As at 13th May 2024, the AEMO says here that:
‘It is anticipated that the use of self-forecasting will deliver system wide benefits by reducing generation forecast error and providing greater autonomy to existing semi-scheduled generators.’
However whilst some participants might place some value in the intended system-wide benefits of the self-forecasting regime, a much more compelling value has emerged in that Semi-Scheduled generators are increasingly using self-forecasts to achieve low or zero Regulation FCAS costs (note – not Contingency FCAS costs).
… as an illustration of this more direct benefit (and the participant’s pursuit of it), in the article ‘Regulation FCAS Costs in 2023’ published on 3rd May 2024, Linton wrote that
‘A cluster of wind units using self-forecasting and having low Regulation costs is clear in the chart’
We have observed instances where the original supposed holistic benefit has appeared to be in conflict with the more direct ‘zero Causer Pays’ benefit … which prompted us to write Observation 15/22 within GenInsights21: ‘What is the purpose of Self-Forecasting?’.
A1) method for the Apportioning of Regulation FCAS costs is changing
It’s important for readers here to understand that the method of apportioning Regulation FCAS costs recovering is changing:
1) is being calculated by a single-sided ‘Causer Pays’ method up until 7th June 2025;
2) thereafter (from Sunday 8th June 2025) will be calculated by a double-sided method (a refinement of the ’causer pays’ principle) being implemented with Frequency Performance Payments process.
… these methods are described here, and in linked pages.
A2) assessment criteria for Self-Forecasts
There’s many other articles on WattClarity pertaining to self-forecasts and the apportioning of Regulation FCAS Costs, but we wanted to highlight these particular articles
1) on 14th April 2023 we wrote ‘Delving deeper into dispatch availability self-forecasting performance’, and in this article provided an illustration of the AEMO’s Assessment Criteria for whether to allow (or suppress) a participant’s self forecast.
2) On 11th August 2023 we wrote ‘What inputs and processes determine a semi-scheduled unit’s availability’.
3) on 18th March 2024 we wrote ‘How many Semi-Scheduled units were submitting Self-Forecasts through 2023 (and how many of these were actually used in dispatch)?’.
(B) History of Self-Forecasting
We recommend that readers here familiarize themselves with the broader evolution of the Semi-Scheduled category. The following table focuses specifically on the role of self-forecasting:
| Date | Milestone (re the Semi-Scheduled category) |
| 28th March 2018 |
Trials of Self-Forecasting were commenced with the coordination of AEMO and ARENA – with ARENA calling for expressions of interest on 28th March 2018 from various parties. That Media Release notes: ‘In partnership with the Australian Energy Market Operator (AEMO), ARENA is seeking to demonstrate wind and solar farms can provide more accurate forecasts of their output into AEMO’s central dispatch system.’ Further details, including some Knowledge Reports, are provided here on the ARENA website. In particular it’s worth highlighting ARENA’s report from January 2020 ‘Short-Term Forecasting Trial on the NEM: Progress Report (April to October 2019)’ which they said (p5/15) is … ‘to summarise insights and progress from initial reports submitted by the 11 participants of the Short-Term Forecasting (STF) trial that is taking place between March 2019 to mid 2021.’ |
| September 2019 |
From September 2019, suitably approved Self-Forecasts were allowed to substitute* for the AEMO-developed forecasts in AWEFS and ASEFS. Remember that the UIGF was being produced for four different time horizons, so where the self-forecast was able to slot in was as follows: Time Horizon #1) In the ST PASA time horizon (out 8 days into the future): … AEMO’s UIGF from ASEFS or AWEFS the only possibility. Time Horizon #2) In the P30 predispatch time horizon (out until 04:00 tomorrow or the day after); … AEMO’s UIGF from ASEFS or AWEFS the only possibility. Time Horizon #3) In the P5 predispatch time horizon (out 11 x dispatch intervals into the future); and … AEMO’s UIGF from ASEFS or AWEFS the only possibility. Time Horizon #4) In the dispatch interval time horizon (i.e. a forecast, at the start of the dispatch interval, of what was possible for the end of the interval). (a) by default this is the AEMO’s UIGF from ASEFS or AWEFS the only possibility. (b) but a suitable accepted (and unsuppressed) self-forecast can substitute here. |
| 23rd Nov 2022 |
On 23rd November 2022 AEMO began to use its improved ASEFS and AWEFS forecasts. 1) Prior to this point, the AEMO had outsourced the supply of AWEFS and ASEFS forecasts to a European vendor: (a) but (following an intensive internal process) had now developed its own capability (b) which was expected to improve the quality of its forecasts 2) At that time we’d written ‘Enhancement of ASEFS and AWEFS forecasts provided by AEMO – from Wednesday 23th November 2022’ to explain. Coincident with this: 1) the AEMO paused any new suppressions for Semi-Scheduled plant that failed the ongoing self-forecasting ongoing assessment; and 2) notice was given that this ‘Grace Period*’ would end on Tuesday 28th February 2023: (a) ~3 months after the upgrade. (b) which has almost completely elapsed now. * note that the term ‘Grace Period’ is our term, not the AEMO’s. |
| Tuesday 28th Feb 2023 |
On Tue 28th Feb 2023, the AEMO 3-month ‘Grace Period’ following the upgrade to ASEFS and AWEFS (which went live on 23rd November 2022) ended. Approx 6 weeks on from this date (on 14th April 2023) we wrote ‘Delving deeper into dispatch availability self-forecasting performance’,: 1) To see how performance had improved 2) and, in the process, illustrating the AEMO’s Assessment Criteria for whether to allow (or suppress) a participant’s self forecast.
|
| 8th June 2025 |
The method of apportioning FCAS Regulation costs changes, to the new ‘Frequency Performance Payments’ method (as discussed here). |
Will add more here, as time permits.
(C) Discussions about, and Analysis of, Self-Forecasting on WattClarity ®
As the energy transition has progressed, we (at Global-Roam Pty Ltd, the publishers of WattClarity) have been taking an increasing interest in Wind and Solar units and the workings of the Semi-Scheduled category … including the role of self-forecasting.
As such, there’s coverage of Self-Forecasting
C1) via WattClarity Deeper Insights
Via our* WattClarity Deeper Insights publications (both Analytical and Statistical) we have been taking an increasing interest in self-forecasting, as one part of our broader interest in Semi-Scheduled units.
* as with prior iterations of this report, these are a collaborative effort between us at Global-Roam Pty Ltd and the team at Greenview Strategic Consulting. That’s what we mean here by ‘our’ and ‘we’.
C1a) GenInsights21 and GenInsights Quarterly Updates
On 15th December 2021 we released GenInsights21, a detailed 622-page analytical report that analysed in detail many aspects of the energy transition, including self-forecasting.
Beginning with 2022 Q2, we continued this process with a series of GenInsights Quarterly Updates that update and extend these insights, particularly with respect to Semi-Scheduled units (in Appendix 3) and the role of Self-Forecasting.
C1b) Generator Statistical Digest
With the release of the GSD2023 on 15th February 2024 we extended the insights further, including with a focused view of monthly statistics for Semi-Scheduled units as a new feature in the GSD.
C2) via articles on WattClarity
Also via freely-accessible articles on WattClarity, we are progressively exploring the ins and outs of self-forecasting in the NEM. Readers might like to refer to:
(a) This collation of articles pertaining to ‘Self-Forecasting for Semi-Scheduled units’ under this category; and also
(b) Other articles tagged with ‘self-forecasting’.
(D) Other Notes & Resources
The AEMO has its own sub-site focused on Solar and Wind Energy Forecasting, which includes:
1) these resources focused on Participant (self) Forecasting; and
2) these FAQs from August 2023.