Where’d the wind go, during the last week of April?

We’re now well into Q2 2021 and (as announced here 11 days ago) we’re well on the way to completing some aspects of our up-coming GenInsights21 release.

(for those who wish to save on their access to GenInsights21 when it is released, you can pre-order today).

We’d commenced the work well before the ESB released its Options Paper for post 2025 design, but the quick skim I have conducted reinforces the importance of one of the Appendices we have coming together now – and that is with respect to the degree of correlation in wind production patterns across the NEM.

(A)  What did we see this week?

Quite topical, then, to look up on our (NEMwatch and ez2view) dashboards during the past week to see a consistent pattern – overlaid on top of the typical 24-36 hour cycle of output was a more pronounced stillness that speaks to what we had gone some way to analyse for the GRC2018, and which we are extending much further for GenInsights21.

Here’s a view (using this query) from NEMreview v7  that shows an aggregate NEM-wide view of what happened:


We can see that total installed capacity is up at 8,703MW currently – and that (along the lines of that ‘yin-and-yang’ animation from way back in 2015) there were both highs and lows in the month:

A1)  The high points

This period included the new all-time record for instantaneous production (5,430MW at 23:55 on Tuesday 13th April 2021).

On a daily aggregate basis, some days saw daily Capacity Factors approaching 50%, which seems to be a pretty impressive yield in historical terms (especially when also considering that those days also typically saw something like an additional 10% of possible harvest spilled* in response to a combination of factors:

Spillage Factor #1 = negative prices, in which cases some wind farms switch off to avoid paying to generate:

(a)  Some wind farms (such as those which cost AGL a write-down a staggering $1.9 billion) have legacy contracts in place, so don’t care about the spot price and run straight through

(b)  Other wind farms that are Non-Scheduled just switch off automatically to avoid the negative price with no need to advise the AEMO (Allan’s analysis of Easter Saturday 2020 here showed an example), which presents the AEMO with its own challenges as the incidence of negative prices increases.

(c)  Other wind farms that are Semi-Scheduled that want to switch off must only do this through bidding behaviour (following recent passage of a Rule Change that follows the AER’s request).

Spillage Factor #2 = constraints (with some of the factors explored in this historical record inspired by the GSD2020).

Caveat (*) on the estimate of spillage – readers are reminded of Marcelle’s warning that it’s ‘not as simple as it appears’ to estimate curtailment of renewable generation.

A2)  The low points

In the same month of these high points, we see that there was a 5-6 day period in the week that’s just passed where wind output significantly disappeared right across the NEM (this had been noted by a number of people I’d spoken with during the week, and I gather this had something to do with a high pressure system in the southern part of the NEM … but I am definitely not a meteorologist!).

1)  There were 5 consecutive calendar days with the NEM-wide daily Capacity Factor less than 15% (and with 3 of these days being under 10%).  That’s only 20-25% of what was delivered on those strong harvest days noted above, if we assume we add back in the curtailment noted above.

2)  As we can also see in the chart, there was negligible curtailment seen during this period, so it does seem to be mostly because of poor wind conditions.

3)  Furthermore, it’s fairly easy to see (from the regional trends in the chart) that output suffered across SA, VIC, TAS and NSW.

4)  Because QLD only has 2 wind farms operating at present, the aggregate consumption is small in absolute terms, so difficult to see from this chart.

(a)  Digging into the data (not shown here) we find the capacity factor was 17% on Sunday 25th April (same day when NEM-wide capacity factor was 11.8%), but then began recovering:

i.  27% on Monday 26th April

ii.  and then up to 43% on Tuesday 27th April and in the days following.

(b)  There was some curtailment on the Monday 26th April (at least in part due to negative prices – did not check constraint activity), but there was none on Sunday 25th April.

(c)  So it seems probable that (in this instance) more installed capacity of wind in QLD would have delivered higher NEM-wide capacity factor for much of the 5-6 day period … notwithstanding possible spillage.

(d)  Working with this hypothetical case and mentally projecting forwards (not even to the point of ‘back of envelope’), the additional harvest possible in QLD would mean a number of things including:

i.  The need for a sufficiently robust grid to ensure delivery of the power where it needs to be;

ii.  Not much change to the MW capacity required in dispatchable plant (because of low harvest on 25th April) compared to a ‘no steeper expansion in QLD’ assumption; but

iii.  Lower MWh energy stored requirement, because of only one ‘lowest harvest’ day to get through, rather than five in a row.


(B)  What have we already done

Two years ago, when we released the inaugural Generator Report Card 2018 (a very well received combination of Analysis and Statistics) we included an exploration of the challenge that is ‘when the sun does not shine and wind does not blow’:


We did this in Theme 10 within Part 2 of the 180-page analytical component within the GRC2018.

Two months after the release, we freely shared here one of the figures presented in the GRC2018 that summed up one of the central tenets behind that piece of the analysis:

1)  Our sense has been that a significant number of people (when thinking about diversity of wind harvest) think about this only as a continuum between two states (i.e. highly correlated, and random).

2)  However it is clear to us that it is the third state that is required to deliver any meaningful capacity benefit (i.e. to the market as a whole) from a deployment of wind farms across the NEM.

(a)  Some of the discussion I’ve seen in the 36 hours that has followed from the release of the ESB’s Option paper has centred around statements about the need for 19GW of dispatchable capacity (i.e. over and above the capacity of wind and solar to be deployed in this energy transition).

i.  On the basis of what we did for the GRC2018, and what I have observed in the market since that time (including in the prior week) makes me to wonder is that (already large) amount of capacity going to actually be enough?

ii.  A key aspect of determining just how much will be required is understanding the objective reality of the diversity patterns that we can expect in a NEM with significantly higher penetrations of wind farms.  Perhaps what is being presented in many quarters, to date, has been ‘sunny days’ predictions that don’t properly factor in High Impact (relatively) Low Probability events?

(b)  Anti-correlation would also be one factor supporting much better individual project economics as well (as Marcelle showed here for one particular example with Bald Hills Wind Farm).


We returned to this theme in May 2020 to note how diversity is not a magic wand– and reiterated on 14th April when we saw peak instantaneous wind output right across the NEM.




(C)  How are we extending this, for GenInsights21?

On page 14 of the ESB’s Option paper it is written:

This means there is a need for 6-19 GW of new utility scale, flexible and dispatchable resources, as up to 63% of the current coal and gas fleet in the NEM) retires by 2040.  To put these numbers into perspective, the average NEM demand is around 20 GW.

(Quotes the AEMO 2020 ISP – Central and Step Change Scenario – Transmission and Generation Outlook Files)

Given the criticality of our collective understanding of the size of this requirement, and ensuring we have a means for incentivising the capacity of this size (and the right performance characteristics) to enter the market, it seems quite important that what we’d delivered in the GRC2018 be subsequently extended to provide a deeper insight.

1)  We did see Cornwall Insight recently publish what appears to be very similar to what we delivered in the GRC2018 two years ago …

(a)  but we have not really seen anyone take what we did back then and extend it further, which does concern us.

(b)  Did we miss something?

2)  Given we see this as quite important, we have had an analytical exercise running now since the start of the year (with results to be published in GenInsights21) to extend this analysis in order to understand:

(a)  What would diversity patterns be of MWh harvest (as distinct from wind speeds) assuming full deployment of wind farms across the NEM … and even further afield?

(b)  What are the extent of the ‘wind droughts’ that 16 years of history have shown we can reasonably expect to occur in the future?

i.  What are their duration, and depth?

ii.  How do they correlate to consumption shapes (i.e. what would be the ‘net energy’ required from dispatchable resources)?

(c)  If that ‘net energy’ were to be substantially supplied from plant that is also itself energy-constrained, what is the shape of the storage required (i.e. GW and GWh)?

i.  batteries and hydro are perhaps more obvious in this energy constraint,

ii.  however gas and liquid fuelled plant can also have energy constraints in terms of accessible fuel supplies – as the outcomes from Texas showed us earlier this year.

(d)  In a separate appendix in GenInsights21 we are also looking to explore what history has shown us in terms of the ‘predictability’ of wind harvest (this piece is not as far advanced at this point).



We’ll look forward to sharing with clients when GenInsights21 is ready.  Don’t forget to send in your pre-order today, and save on the final release price.

About the Author

Paul McArdle
One of three founders of Global-Roam back in 2000, Paul has been CEO of the company since that time. As an author on WattClarity, Paul's focus has been to help make the electricity market more understandable.

8 Comments on "Where’d the wind go, during the last week of April?"

  1. We have known since Paul Miskelly’s paper in 2012 that there are always going to be prolonged and widespread wind droughts when high pressure systems sit over SE Australia.
    We also know that on windless nights there will be next to no RE generated.
    When there is no spare wind anywhere it will not help to have more transmission lines.
    We know that the so-called big batteries do not store power at grid scale.
    Pumped hydro costs the earth and wastes up to 40% of the power that is fed into the system.
    Join the dots:)

  2. Looking at the situation in SA (the wind-leader,) checking at breakfast and dinnertime – the peaks of demand coinciding with little or no solar, we find SA almost always importing from Victoria (with help from the battery of the nation quite often.) When the wind is low, look at the amount of gas used in SA and the coal plus gas in Victoria. Then contemplate the situation when a big chunk of Victorian coal capacity gone, bearing in mind that no amount of extra wind and solar capacity is any use on windless nights, or even windless breakfast and dinnertimes.

    Stop celebrating big wind periods and focus on the lows, like the low points of a flood levee where the water will penetrate regardless of the high points of the levee or the average height. It is the low points of the wind supply that kill the grid, regardless of the installed wind and solar capacity. Build as much as you like if you want to displace coal power, RE can displace it but not replace it.

  3. Have a look at SA and Victoria this morning, they are back in wind drought territory (less than 10% of plated capacity).

    Actually Paul knows all this and he has been dropping hints for years, but he doesn’t want to frighten the horses and he is waiting to see how long it takes the RenewEconomy people to work it out for themselves:)

  4. Here are the periods of maximum and minimum rolling 1,2,3 and 7 day NEM Wind means for the last year .. http://nemlog.com.au/nlog/nem-wind-summary/

  5. That Capacity Factor (CF) should be one of the most important considerations here.

    While there is a total Nameplate now of 8703MW, when CF is taken into account, that means that Nameplate equates to a power generation of only 2610MW

    Over the last 135 weeks (a bit more than two and a half years now) that long term CF is 29.6%, roughly the same as it was 135 weeks ago.

    In that time there has been a number of increases to that Nameplate, adding more and more ‘better technology turbines into the grid.

    In the most recent 12 Months, there have been four increases to that Nameplate, from 6960MW up to that total now of 8703MW, so an increase of 1743MW.

    However, that rolling CF for the most recent 12 Month to this date is only 29.2% in fact, so with the addition of better technology, the CF has actually gone down.

    As to the weather situation, it’s relatively simple to work out when wind generation will be low. 65% of all wind plants (Nameplate (around) 5700MW of that 8700MW) are located in that area of South East South Australia and Central Western Victoria. With the approach of those large High Pressure systems, and then very slowly passing over that area, it’s easy to see when wind generation will be low, sometimes falling by anything up to 3000MW in hours.

    Conversely, that in between times between those Highs and the following lows, when those isobars are closer together, hence higher winds, you can see large and sudden losses of power, ranging between 500MW and 1500MW in around the time spacing of ten to 15 minutes. Those sudden losses are due in fact to those newer technology plants, taller towers, larger nacelles with larger generators, and considerably longer blades. In those high wind situations, vast numbers of those newer tech turbines automatically turn off, so the blades, (higher tip speed) are not destroyed.

    Those sudden losses of power can be seen by looking at those wind graphs, and you can see numbers of those ‘spikes’ occurring in succession, off, on, off, on, etc.

    And no, I am also not a Meteorologist, just an observer with a bacground in the electrical trade who puts two and two together, having been doing this for so long now.

    That CF figure should be the most important figure in all of this.


  6. I know a bunch of wind farms schedule their annual maintenance shutdowns at this time of year (in preparation for the more lucrative winter months I guess). Could explain some of the low overall output that you’ve seen recently.

  7. WA has an island grid known as the SWIS. SEN modelling for 90% RE on an hourly basis using 2014 weather data (a low year for wind out of the previous decade) shows us that while there are winter wind droughts that can strike for ten days in a fourteen day period, it is manageable and even with these wind droughts, storage and overbuild still can provide 90% affordably, and 100% for a little bit more cost (summer is not an issue due to PV).

    I’d post an image of a summer and winter four week period, but I can’t. Here’s a tweet you can look at.

  8. I should add the wind correlation in WA wind farm locations on the SWIS is pretty high. Batteries and PHES really helps, but only goes so far (say 2 hours for batteries today, 4 hours in 5 years time and 10-14 hours for PHES in WA)

Leave a Reply to Dave Cancel reply

Your email address will not be published.