Through 2022 and now into 2023 we have published a number of different excerpts from the report via articles here on WattClarity, and have delved into more some aspects in more detail. We also provided briefings to a number of Companies and Industry Organisations (our briefing to the Smart Energy Council on Tuesday 5th April 2022 was recorded, and has been widely viewed since that time). Some pieces of analysis there are now being updated each quarter in GenInsights Quarterly Updates.
For various reasons we’ve chosen to share here the whole of Observation 15/22 in its entirety (with links added, for ease of reference) …
In Appendix 6, we explore the recent rise of self-forecasting, and pose the question about whether the incentives to participants are delivering the most valuable power system security benefits:
- Is the purpose of self-forecasting to help the participant reduce their Causer Pays costs?
- or is the purpose of self-forecasting to provide a better objective forecast of wind or solar capability than AWEFS or ASEFS can provide, to improve power system security, and
- … is this being done?
Semi-scheduled generators (large wind and solar farms) are dispatched based on forecast energy availability, and in addition to their dispatch level, receive a flag (the semi-dispatch cap (SDC)) from AEMO which when set requires the generator’s output to remain at-or-below the dispatch level.
15.1 Self Forecasting introduced as an option
The forecast energy availability was provided exclusively until 2018 by:
- AEMO’s Australian Wind Energy Forecasting System (AWEFS) and
- the corresponding solar ASEFS.
Self-forecasting, as a trial, was introduced in 2018, allowing generators to provide their own five-minute-ahead forecast of energy availability to AEMO.
AEMO’s website currently* says:
“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.”
* Editor’s Note: The AEMO’s statement-of-purpose was there when we published GenInsights21 on 15th December 2021, and is still there today (Monday 27th February 2023) when we publish this article.
Several innovative technology providers stepped up and worked with generators, through an ARENA project and then subsequently through other commercial arrangements.
15.2 Stress testing during difficult times
While there’s an ongoing benefit from slight improvements to everyday forecasts, self-forecasting is most beneficial to power system security in “difficult” times.
One of the most difficult times to forecast a generator’s output is when it is limiting its output due to a semi-dispatch cap, meaning its possible unconstrained future output can’t be guessed from its current output. The generators themselves have access to far more data and information than AEMO on specific conditions at the site and the impact on generation, so should be able to provide improved forecasts at these times.
15.3 Benefits to the participant
For a participant, self-forecasting gives control of the impact of the forecast on the asset’s performance, in reducing under-forecasting during semi-dispatch cap periods (thereby reducing lost revenue), and in improving its causer-pays factor, which determines regulation FCAS costs.
The causer-pays factor is based on the deviation of the plant from an ideal linear trajectory between dispatch targets (which depend on the forecasts) and a metric based on the power system frequency.
The incentive isn’t just to improve forecast accuracy – but to do so considering a metric of power system frequency at the time.
15.4 Increased incidence of usage – but not all the time
Appendix 6 looks at the uptake of self-forecasting, finding:
- more than 50% of solar farms are using self-forecasting, and
- around 40% of wind farms, and
- these percentages are increasing.
However, an unexpected finding was that many of the generators aren’t using the self-forecast all the time, with some around 80-90% of the time.
If the forecast is good, then for power system security it’d be more beneficial to be used all the time – and if it’s not good at some times, then AEMO needs to work more on improving AWEFS and ASEFS as they’re still going to be called upon.
A striking revelation in Appendix 6 is that there are several wind and solar farms that appear to withdraw the self-forecast during times of semi-dispatch cap, exactly when it’s of most benefit to AEMO.
15.5 Is the submission of a self-forecast value tantamount to a rebid?
If participants (through either manual or their self-forecasting vendor’s automated processes) are going to ‘play’ with the value under semi-dispatch cap conditions, and this value is feeding the availability of the asset in the dispatch process, should this be governed by a something similar to a change in Availability for a scheduled unit, and thereby require a rebid reason at the time?
There’s clearly a mismatch of benefits between the participant and AEMO here!