Yesterday we wrote ‘Three time-ranges of interest, with respect to aggregate ‘Dispatch Error’ by Fuel Type on Tuesday 27th January 2026’.

As noted in that article, what jumped out at me straight away was:

• Long contiguous periods of collective under-performance (i.e. positive aggregate Dispatch Error from the VRE fleet)
  • with aggregates for both (Semi-Scheduled) Wind and Solar:
    • sometimes overlapping in time;
    • so putting more of a collective drag on frequency (as we’ve already explored on many occasions, some of which are tagged and collated here).
  • though perhaps caused due to different factors
• With the Scheduled units (primarily BESS and Coal units) doing what they can, via PFR and Regulation FCAS arrest the frequency drag.
• On Tuesday 27th January 2026:
  • Which was a day of particular interest for other reasons
  • This seemed to break down into three discrete blocks of time

I thought that, from that starting point, it would be better (and clearer for readers) to separate out the two separate VRE legs and analyse and report on each separately.

So:

• this is the 2nd of two articles today … this one specifically focused on Wind (i.e. the Semi-Scheduled units only, and in aggregate).
• because we’ve already published ‘Delving deeper, into periods of under-performance across Semi-Scheduled Solar Farms on Tuesday 27th January 2026’.

Breaking the NEM-wide data down to Region specifics

Taking the ‘Trends Engine’ within ez2view, I took the analysis presented in the prior article and did two things:

• Dropped the other Fuel Types (retaining just NEM-wide Wind); and
• Ran the same query for aggregate Dispatch Error, but down at a region level in addition to a NEM-wide level:
  • Adding 5 more data series
  • i.e. one for each of the Regions

This is what we then see:

Note that

• I’ve kept the axes the same, so that we can easily compare between a few charts in terms of relative size
• Per usual, click on the image for a larger-resolution view
• For those with a licence to the ez2view software, you can access your own copy of this trend query here.

With respect to the results above:

1)  Unlike the case for Solar (where NSW deviations were dominant) in this case we can clearly see:

(a)  that collective under-performance in SA and VIC are the main drivers.

… remembering that a positive Dispatch Error represents under-performance (relative to Target, which is what NEMDE uses to make Supply balance Demand).

(b)  collectively Wind Farms in NSW, QLD and TAS almost don’t feature at all in the day.

2)  With respect to the three time-ranges noted here, we can clearly see that:

(a)  Time Range 1 (which was in the morning) was dominated by under-performances across SA Wind Farms.

(b)  Time Range 2 (the sharp spike, peaking at +673MW NEMwide at 15:15 (NEM time)) is dominated by under-performances across VIC Wind Farms;

(c)  Time Range 3 (the longer run of under-performance through the afternoon and well into the evening):

i.  Is mainly under-performances across VIC

ii.  But also featuring under-performances across SA

Remembering two things…

We’d ask readers to remember two things, at this point:

Remember #1)  Remember some key facets of Semi-Scheduled units:

(a)  Which (as clarified in the AER Rule Change) ‘benefit from’ a uni-directional, and looser, Dispatch Conformance requirement than is the case from fully Scheduled units

(b)  But:

i.  as Allan well illustrated in his earlier article ‘Semi-dispatch caps are loading the deviation dice’,  makes it much more likely … indeed certain … that collective under-performance of VRE will occur more often and in larger quantities;

ii.  with the results clearly seen in these articles progressively updating a trend of AggROT across Semi-Scheduled units

(c)  Which is:

i.  Creating much more consistent, and larger, pressures on system frequency control;

ii.  Which then need to be addressed by the remaining Scheduled units (primarily BESS and Coal units), via PFR and Regulation FCAS arrest the frequency drag.

… with the delivery of these services coming from units that are either:

>  increasingly ageing; or

>  increasingly energy-limited.

… and the Semi-Scheduled (which I understand should be providing PFR when they can) largely not providing that service.

(d)  and hence is one of the factors that has been making us question if the Semi-Scheduled category is scalable and sustainable, in its present form.

Remember #2) Please remember the higher-level context to what was happening on Tuesday 27th January 2026:

(a)  It was a very hot day in the southern regions

…. which, amongst other things, drove VIC demand to a new record.

(b)  and it also incorporated a substantial cool change.

Taking a close look at Wind Farms in South Australia

With the above in mind, here’s some more results from the ‘Trends Engine’ within ez2view, focused on the South Australian region:

For those (licensed users) who want their own copy of this query, click here.

In this chart:

1)  I’ve focused just on aggregate Dispatch Error (from Wind Farms) in South Australia

2)  Then I have added on three trends of different Air Temperatures from different measurement points around the region.

(a)  note that the ‘Trends Engine’ within ez2view contains data feeds for many different measurement points in each state and territory that span back almost 20 years

(b)  I did this quickly and found 3 that showed different stories that I thought were useful

3)  In particular, note how the air temperature at Port Augusta airport rose significantly to high levels, but then also fell (quickly and) significantly

4)  It’s a bit rubbery (because of distance, and also hub height) but …

(a)  if we assume that this is somewhat reflective of air temperatures at the hub heights of many Wind Farms in South Australia;

(b)  then we see:

i.  The ramp up in temperature correlates reasonably well with the ramp up in aggregate Dispatch Error;

ii.  The more rapid ramp down in temperature correlates reasonably well with the more rapid drop in aggregate Dispatch Error;

(c)  now:

i.  ordinarily we should remember that ‘correlation does not necessarily suggest causation’;

ii.  although we know that there are high-temperature effects on most generation technologies … including Wind Farms!

5)  So this, to me, suggests that high-temperature de-rating was a significant factor in the collective under-performance seen across SA Wind Farms on Tuesday 27th January 2026 … particularly in the morning period.

This might be something we might explore in more detail later with respect to individual DUIDs … or that might be better as one-on-one conversations with some?

Taking a close look at Wind Farms in Victoria

Moving east into Victoria, here’s some more results from the ‘Trends Engine’ within ez2view, focused on the Victorian region:

For those (licensed users) who want their own copy of this query, click here.

In this chart:

1)  I’ve focused just on aggregate Dispatch Error (from Wind Farms) in Victoria

2)  Then I have added on numerous trends of different Air Temperatures from different measurement points around the region.

… despite the notes above about the BOM data in the ‘Trends Engine’ within ez2view, I could cheat a bit in VIC because I’d already used the same data here

3)  In particular, note how:

(a)  The cool change came from the west:

i.  First in South Australia at Adelaide and Port Augusta;

ii.  Then in VIC … Portland and Mortlake, then Geelong, then Melbourne

iii.  Some hours to cross that distance

(b)  And, aligned with this, we see a pattern of aggregate Dispatch Error (from Wind Farms) in Victoria

i.  A rapid ramp up, and then drop, in (a shorter) time period 2

ii.  Then a long and lingering (though not as extreme) period of under-performance in time period 3

4)  So again we conclude that high-temperature de-rating was a significant factor in the collective under-performance seen across VIC Wind Farms on Tuesday 27th January 2026 … particularly in the afternoon/evening period.

This might be something we might explore in more detail later with respect to individual DUIDs … or that might be better as one-on-one conversations with some?