As noted on the weekend, the AEMO preliminary operating incident report touches briefly on the earlier loss of ~500MW of production at Stockyard Hill Wind Farm at around 11:35 (NEM time) on Tuesday 13th February 2024 (we’ve called that Event 1) … but AEMO’s main focus, along with many others, is on the later and larger issue that began at ~13:08 with transmission towers being blown over (we’ve called that Event 2).
Whilst it’s logical that the main focus of attention is on Event 2, at Global-Roam Pty Ltd (publishers of WattClarity) we’d like to investigate all 4 Events that occurred on the day … and Event 1 is a logical place to start (hopefully nowhere near as complex as Event 2 and Event 3).
Assumed pre-reading
In addition to the AEMO’s Preliminary Operating Incident Report, we’re assuming you’ve read through each of these three earlier articles:
1) On Thursday 15th February, we’d posted our ‘sequence of events for what happened in Victoria through the day on Tuesday 13th February 2024 (via Timescale #1)’.
2) Then on Saturday 17th February 2024 we posted this ‘quick first pass through ENERGY bids (for Victorian units) on Tuesday 13th Feb 2024’.
3) Then on Sunday 18th February 2024 we posted this ‘From the 4-second SCADA data, a view of Production by Fuel Type on Tuesday 13th February 2024’.
4) Earlier today (Tue 20th Feb) Allan wrote about ‘Victorian renewable generation on 13 February 2024: One picture and a thousand (sorry, many more) words.
It’s also worth noting that ‘AEMO preliminary report identifies 4 separate Events … including trips* at 4 Generators and 1 Load’ and …
1) with particular reference to Stockyard Hill Wind Farm, the report refers to ‘the loss of … Stockyard Hill’,
2) as distinct from the disconnection of Dundonnell Wind Farm refers to this as being a ‘trip’.
… so I was curious to understand the distinction (i.e. between a ‘trip’ and a ‘loss’), starting with this article about Stockyard Hill Wind Farm – and following later with a similar article focused on Dundonnell Wind Farm.
Finally, also worth noting this discussion that ensued here on LinkedIn earlier on Tue 20th February 2024 in response to Allan’s earlier article:
Grass Fires around Stockyard Hill Wind Farm impact on output
I’d heard whispers and rumours about what might have been the genesis of the issue, so I did a bit of Googling and found this update from CFA Victoria via Twitter published at 11:23 (NEM time) on Tuesday 13th February 2024:
Clicking through on the link shown in the message suggests that ‘the incident may have expired’, but a Google search with respect to Stockyard Hill Wind Farm finds articles putting the grass fire in the vicinity of Stockyard Hill Wind Farm – including:
1) On Tue 13th Feb 2024, Anna Houlahan wrote ‘A grassfire is burning at Stockyard Hill, west of Ballarat, “that is not yet under control” and travelling in a north-easterly direction towards Lake Goldsmith.’ in this article in the Examiner;
2) On Tue 13th Feb 2024, Tess McCracken wrote ‘An out of control grassfire at Stockyard Hill, Beaufort, 183km west of Melbourne has been upgraded to a watch and act alert as the blaze travels in a north-westerly direction towards Crossroads.’ in this article at News.com.au;
3) On Friday 15th Feb 2024, Simon Johanson wrote ‘Origin’s gas-fired plant in Mortlake, in Victoria’s west, is operating at full capacity under direction from the Australian Energy Market Operator. The company cut power it usually supplies to customers from the Stockyard Hill Wind Farm, west of Ballarat, because of two major bushfires.’ in this article in SMH;
4) Giles Parkinson wrote ‘a reported grass fire took out the Stockyard Hill wind farm’ in this article on Sunday 18th Feb 2024;
So it seems clear the grass fire came close to, and impacted the output of, the Stockyard Hill Wind Farm. Though I still have more questions!
Zooming in on the AEMO 4-second Data for Stockyard Hill Wind Farm
In this top-down article on Sunday 18th February 2024, we included a chart containing a 14-hour time-series of output for every single Wind Farm in Victoria from 06:00 to 20:00 on Tuesday 13th February 2024:
It was a very busy chart, but (in part because of its large size) I was able to colour in the STOCKYD1 DUID along with several other large (and/or topical) units. In that article, I noted the following:
2) I have coloured some of the topical (or otherwise larger units) … starting with STOCKYD1, the subject of Event 1:
(a) I’ve not seen any reports yet that note why the unit was ‘lost’ (the word used in AEMO’s Preliminary Report) around 11:35, and am certainly interested to understand this;
(b) This quick view of the 4-second data did not explain why that was the case – other than that it was a rapid loss;
(c) But what was a surprise is that the unit appears to power up rapidly at ~12:26 for unknown reasons, and then disappear just as quickly?
… something to explore in a further article (Observation/Question #21).
So I thought the most logical place to start, in diving in further, would be with the same chart but stripping out all the other units, as such:
(as I hit publish now, I note I used the word ‘tripping’ in relation to Stockyard Hill – hence adding to the confusion – hence my apologies)
On this chart I have noted three ‘events’ with respect to Stockyard Hill Wind Farm:
Event 1a) being some (what I assume to be wind speed related) drops in output at ~10:07 and ~10:30 on the morning;
Event 1b) being the unit being ‘lost’ at ~11:35; and
Event 1c) appearing to be a brief attempt to return to service at ~12:26 on the day, which would be:
(a) ~50 minutes after it had come offline; but also
(b) ~42 minutes before the larger disruptions of Event 2 at ~13:08 that included the trip of Dundonnell, Loy Yang A and Yaloak South + loss of load at 1,000MW of load ‘shaken off’, and so on…
It’s Event 1b and Event 1c that particularly interest me here in this article, so I’ll just ignore Event 1a (assuming it’s unrelated to Event 1b, and leave to some possible future/separate article?) and zoom in to show more detail:
This chart I’ve annotated with some quantities and times – coming from the AEMO’s 4-second SCADA Data (but note the caveats).
A first look at the AEMO market data in EMMS for STOCKYD1
With the above frame of reference in mind, I thought it would be useful to attach this snapshot from the ‘Bids & Offers’ widget in ez2view, which has been:
1) Time-travelled back to 24:00 at the end of Tuesday 13th February 2024;
2) Set to look back 24 hours, so covering the whole day; and
3) With the ‘Bid Table’ also showing, so readers here can see the bids submitted for the unit across this 24 hour period
… note that this also shows (in light grey rows) bids submitted but superseded before any point of dispatch.
I’ve also annotated with the same 3 events (i.e. Event 1a, Event 1b and Event 1c) noted above.
Interestingly the fleeting burst of production that appears in the 4-second SCADA data from 12:25:31 to 12:27:43 (i.e. a duration of 2 minutes and 12 seconds in Event 1c) is not seen in any dispatch interval data – which is not a total surprise, given (assuming the timestamps are correct) that it’s located wholly within the 12:30 dispatch interval.
Some of the rebids shown in the Bid Table we will refer back to in the chronological listing below.
Selected dispatch intervals (for STOCKYD1) from 10:30 onwards
With the above frame of reference in mind, I thought it would be useful to switch to use of the ‘Unit Dashboard’ widget (and other linked widgets) in ez2view, setting the look-back to only 3 hours and look-forward to 3 hours (i.e. into predispatch from the ‘now’ point of the window), and have captured a few dispatch intervals from 10:30 onwards…
I’ve picked out a few particular things … but there’s more that we could look at (alas, not enough time).
And remember that (as with this, similarly structured, article here) there are many caveats applying to this first pass … Reader Beware, there may be errors, and the picture is definitely incomplete at this point.
| Dispatch Interval on Tue 13th Feb 2024 remembering this is ‘NEM time‘, |
Development during this dispatch interval
(readers should remember … with the images shown below, we try to set them up so clicking on the image will open in a new tab as a larger size image … so if you have 2 windows you can more easily read). |
|---|---|
|
up until 10:25 |
We’re going to skip these earlier dispatch intervals, as Event 1a. |
|
10:30 … still one hour prior to the ‘loss’ of the STOCKYD1 unit in the 11:35 and 11:40 dispatch intervals |
It’s worth starting here with the ‘Unit Dashboard’ widget in ez2view:
In this image we see: 1) That the unit has 300MW bid at -$64.55/MWh and another 150MW of volume bid at -$61.80/MWh: (a) I don’t have time to consider why two bid bands so close together (perhaps PPA?) (b) But loosely speaking we’d call them both ‘negative LGC’ 2) We also see that the unit’s currently not constrained, but is forecast to be constrained in the AEMO’s P5 predispatch run for 10:35. So we click through to the ‘Constraint Dashboard’ widget in ez2view, focused on the ‘V>>NIL_MLGT_MLGT_R2’ Constraint Equation to understand more about this constraint:
The ‘Standing Data’ tab (not shown here) reveals that this constraint relates to ‘avoid O/L Moorabool to Geelong #1 or #2 on trip of other Moorabool to Geelong line, Yallourn W Unit 1 in 220 kV mode’ and is part of the System Normal ‘V-NIL_RADIAL’ Constraint Set that’s been invoked since 27th April 2021 and is currently scheduled to persist forever after. As identified in the image above, there are many DUIDs on the LHS of this constraint equation, with many with larger positive factors than the +0.3762 factor assigned to STOCKYD1, which means that (assuming all units bid the the same price) the other units with the higher LHS factor would be ‘constrained down’ before STOCKYD1. |
|
10:35 … five minutes forward, just because of the forecasted constraint |
Stepping forward with the the ‘Unit Dashboard’ widget in ez2view we see the constraint has bound, driving the CPD Price slightly lower … but it’s still above the units ‘most expensive’ volume (bid at –$61.80/MWh at the RRN) so it is not constrained down.
The P30 predispatch forecasts also show no forecasts for drop in output – but that does not mean it can’t happen, so it’s understandable (happens often) that there’s a rebid made (appears to come from an autobidder) arriving at 10:32 to be considered for the 10:40 dispatch interval shifting volume down to –$1,000/MWh. At this point, we’ll skip a few dispatch intervals … |
|
10:40 to 11:25 |
Skipping these Dispatch Intervals. |
|
11:25 |
No snapshots of the market for this point – just remember that the CFA Update copied above was published at 11:23 (i.e. in this dispatch interval) and it said: ‘Grassfire – Watch and Act for Stockyard Hill. You should leave Now if you are not prepared to stay’. Perhaps we will learn more later, if any particular instructions were provided to the onsite operators at the Wind Farm? |
|
11:30 … just prior to the ‘loss’ of the STOCKYD1 unit in the 11:35 and 11:40 dispatch intervals |
Stepping forward to 11:30 (i.e. just before the output dropped to 0MW, we again use the ‘Unit Dashboard’ widget in ez2view:
In this image we’ve highlighted: 1) A series of prior intervals where the unit has seen a positive Raw Off-Target (i.e. because its output was below Target). (a) In his earlier article, Allan noted that this was quite common through the day for VRE in Victoria (i.e. this was what Allan coloured red). (b) Note that the Conformance Status is derived as ‘Normal’ because the unit is under Target (i.e. not above, which would be more of a concern), and is presumed to be deviating as a result of the changing wind resource. 2) Also noted that at this point the CPD price is well under the ‘negative LGC’ points where the bids had previously been (i.e. prior to the rebid to -$1,000/MWh shown above). … If not for the rebid, this unit would have been ‘constrained down’. 3) In this dispatch interval, there’s a second constraint equation that has begun to bind … which, due to similarity in Constraint ID, suggests it’s related to the earlier one. |
|
11:35 … STOCKYD1 starts to come offline (Event 1b) |
Stepping five minutes forward and we see the unit start to come offline:
In particular: 1) The MaxAvail in the bid is still set at 511MW; 2) But the Availability is adjusted downwards slightly because UIGF (which comes from AEMO’s AWEFS system in this instance) is at 501MW. 3) As a result of that Availability, the unit is given a Target of 501MW (i.e. because it’s bid below its CPD price); 4) But, in contrast, the unit tracks the other way … from 474MW at the start of the dispatch interval down to 305MW at the end of the dispatch interval. |
|
11:40 … STOCKYD1 is fully offline (Event 1b) |
Stepping five minutes forward and we see that the unit is now fully offline:
In this dispatch interval, we see the UIGF from AWEFS (which is setting the Availability for the unit in the NEMDE run, being lower than the MaxAvail) is at 501MW … yet the unit is now offline. Interestingly, looking into the next dispatch interval there’s a rebid submitted at 11:37 (what a coincidence!) that will take effect for the 11:45 dispatch interval … |
|
11:45 … offline – but not in the bid |
So stepping forward we see there’s a key difference in determining Availability:
Quick notes: 1) The rebid reason was ‘Band adj to manage binding constraints – SL’, but 2) The effect of the rebid was ‘just’ to extend further forward the time range into the future at which volume was offered at -$1,000/MWh. 3) As such, the bid still says capacity is available … 4) … but thankfully AEMO’s AWEFS process now says the true availability is 0MW … how this comes about is a topic for a follow-on article, when time permits. 5) … so the Target is set to 0MW. |
|
11:50 … offline – but not in the bid |
Stepping forward 5 minutes, the picture is much the same:
… so moving along … |
|
11:55 … offline – but not in the bid |
Again at 11:55 the picture is much the same as the two dispatch intervals before:
… so moving along … |
|
12:00 … offline – but not in the bid |
Stepping forward to 12:00 we see a big change in colours for the bid bands to take effect in the 12:05 dispatch interval:
This is the result of the rebid submitted at ~11:55, and which we will review just below … |
|
12:05 … STOCKYD1 rebids, and sets MaxAvail to 0MW |
At the 12:05 dispatch interval, the picture changes:
Importantly, the available capacity shown in the bid from this point forward is seen as 0MW. Using the ‘Bid Comparison’ feature in the ‘Bids & Offers’ widget in ez2view shown above, we see a clear difference between the two bids:
Not only has volume been shifted to higher priced bid bands – MaxAvail has also been set to 0MW, as a reflection of the physical limitation of the plant. |
|
12:10 & |
Nothing to add at this point for these 3 x Dispatch Intervals. |
|
12:25 … just prior to the rapid spurt in output |
Returning to the ‘Unit Dashboard’ widget in ez2view:
There’s nothing shown here that suggests a spurt in output coming soon. |
|
12:30 … the dispatch interval within which there appears to have been a rapid spurt in output (Event 1c)
|
Stepping forward 5 minutes:
We can’t see a spurt in output … but we do see a rebid to take effect in the next dispatch interval… |
|
12:35 … afterwards |
So taking more of a look, nothing much looks different in the ‘Unit Dashboard’ widget in ez2view:
… except that I’ve flagged how MaxAvail from the bid is not 0MW anymore: Remember it’s always the lower of MaxAvail and UIGF, and UIGF has continued to be 0MW, so that means Availability continues to be 0MW. Taking a look at the rebid through the ‘Bid Comparison’ feature in the ‘Bids & Offers’ widget in ez2view shown above, we see a clear reversal in what happened before:
Max Avail is no longer 0MW … |
|
12:30 onwards |
And that’s where we’ll leave this chronological review. |
If you find any mistakes (or things that are not quite clear) in the above, please do let me know:
1) Either as a comment on the article (which is public); or
2) Privately via email (if you have my address) or LinkedIn.
Leave a comment