Opinion piece by Gordon Wymer, Chief Commercial Officer at Snowy Hydro published on 14th February 2022.
The root cause of the Global Financial Crisis was not, as “The Big Short” would have you believe, endemic corruption and fraud. The basic, primary, fundamental cause of the GFC was that correlated distribution tails that were considered to have zero probability of occurrence, all occurred, simultaneously. It turned out that the probability that Fred couldn’t repay his mortgage was actually highly correlated to the probability that Madge couldn’t repay hers either.
And the scenario in which one collateralised debt obligation (or CDO, just a bundle of lots of mortgages) defaults, and causes the simultaneous default of lots of other CDOs and CDO derivatives was not considered, most famously by the credit rating agencies who ascribed AAA ratings to a pile of financing instruments that turned out to be junk.
The Australian Energy Market Operator (AEMO)’s target for annual unserved load is 0.002%. Simplistically, that means you should expect a blackout for around 11 minutes a year. I reckon that’s AAA. I’m OK with that, because my milk won’t go off in 11 minutes and hospitals’ emergency generators won’t run out of diesel or gas.
But there is a GFC-type problem with this beautiful world. AEMO is tasked with writing the “Integrated System Plan”, or ISP, which plots the National Electricity Market (NEM)’s forecast generation, transmission and storage out to 2050, from the system administrator’s perspective. In writing the draft 2022 ISP, AEMO have set themselves a task of such fiendish difficulty (in modelling the NEM in granular detail, from the bottom up, for 30-odd years) that they have to make assumptions about many things that may look secondary but actually matter a lot. So their targets for electricity storage and transmission builds are based on averages rather than on testing whether a given scenario holds up to perceived low-probability outcomes.
For example, the draft 2022 ISP ranks future Renewable Energy Zones on the basis of average capacity factors, and on the basis of the correlation of native supply with native load (which is close to irrelevant) rather than on the correlation of supply between future Renewable Energy Zones. The modelling therefore assumes that output from Fred’s wind farm in Victoria and output from Madge’s wind farm in Tasmania are not correlated (or in AEMO’s words, Tasmanian wind will “provide greater resource diversity to mainland wind farms”).
For NEM reliability, these inputs are critical. The correlation between Victorian and Tasmanian wind is actually 0.55, using AEMO’s latest public data set. This is a high correlation factor. So this (and other similar) assumptions are incorrect, which leads to projections for future storage demand that are based on a key error. Compounding this issue, AEMO continues to employ a regulatory test for new transmission investment (the infamous “RIT-T”) that many industry participants, including Snowy Hydro, believe is terminally flawed. This leads to conclusions on transmission investment that are, practically, highly problematic.
The other GFC-type problem is that sections of the media are trying to tell us that a 100% renewable NEM will work just fine, with just a few batteries and a bit of “redundant” renewable capacity. They point to South Australia as a beacon for a new world. Nick O’Malley writes, in the Sydney Morning Herald, 16 January 2022, South Australia breaks record by running for a week on renewable energy.
Except that it did no such thing. Factually, this headline is untrue.
It is true that “the state produced on average 101 per cent of the energy it needed from wind, rooftop solar, and solar farms”. That’s because all that wind and solar wildly over-generated when it was “on”, to the point it had to be exported to Victoria or significantly “curtailed”, which is a polite way of saying that the solar and wind generators were told to turn their generation down, or off. But, for 45% of the time, supply had to be bolstered by imports from Victoria, and by local gas and diesel generation (by the way, there is absolutely nothing wrong with that).
In the NEM, the probability that the wind does not blow, and the probability that the sun does not shine, are correlated. Worse, these probabilities are also correlated with times of peak demand. What does this mean for a 50% (or more) renewable NEM? It means we need electricity storage. Lots and lots and lots of storage. Shallow storage, deep storage. And transmission. Lots of transmission. The more, the better; the sooner, the better. And optimal utilisation of the assets that will be there in 2030.
Like Snowy 2.0. Did you know, dear Victorians, that AEMO’s draft 2022 ISP prefers a new transmission link to Tasmania over the “VNI West” link that connects Snowy 2.0 to Victoria (and completes a very useful loop through NSW, Victoria and South Australia)? Snowy would love to buy more wind and solar electricity from Victorian developers. We already buy more than 1,100MW from Victoria, NSW and South Australia. But if VNI West is deferred until 2031/32, grid-scale wind and solar development in Victoria will stop, because the current transmission lines are full.
In the blackouts of 2019, Snowy had over 1,500MW of spare capacity at Tumut 3 in NSW, ready to save the day for Victoria. But the transmission lines were full. No more power could get through. Hello darkness, my old friend.
If the system planners don’t look at the probabilities of low, and correlated, wind and solar output, and at wind and solar droughts, and correlation of troughs in supply and peaks in demand, if they don’t start focusing on what happens when the correlated distribution tails that were considered to have zero probability of occurrence, all occur, simultaneously…
I’ve heard that before somewhere.
I look forward to going through these issues in detail at the conference.
Want to hear more? Join us in Melbourne on 16th – 17th March for Enlit Australia 2022 to hear Gordon deliver his presentation The NEM: It’s just not that into you.