When the wind don’t blow

Go back roughly half a decade, and the gripe you’d see most regularly cited about renewable energy was price. Since then, renewables are massively cheaper, in many cases over-taking thermal coal and gas, both for existing power stations and the cost of building new ones.

Accordingly, the critical narrative has changed to issues of reliability. The gripe is simple: solar panels and wind turbines only generate when the resource they use is available. The bit of earth that your solar panel is attached to spins away from the star in the centre of our solar system, and the speed of our planet’s swirling, roiling atmosphere speeds up and slows down. 

Something momentous has happened recently with the installation of a big, flashy battery in South Australia (at the time, the world’s biggest). It was announced by Elon Musk alongside then-Premier Jay Weatherill. But oddly, the successes of the machine, far exceeding expectations, haven’t had quite the same fanfare. 

It flows when the wind don’t blow

First up, you should know that battery’s primarily designed to function as a ‘fast response’ helper controlling the balance between consumption and demand in the grid. It’s not there to churn out megawatt hours for weeks on end – they’re ‘millisecond tweezers‘.

Not only is the battery doing this, a recent report shows it’s out-performing 19th century steam engines. A coal fired power station tripped offline on 2017-12-18 at 17:35. The report, from the Australian Energy Market Operator shows the battery’s response (it’s called ‘Hornsdale Power Reserve’), compared to the response of a thermal power station, with each facility (dark purple) trying as hard as it can to follow an instruction from the market operator (the red line):

This is what the battery was designed for. Not only does it do that well, it out-performs traditional power stations. This gives you a little taste of how incredibly wrong it is to say that renewables paired with storage technology like this is a ‘threat’ to grid reliability. 

But, something I was curious about was the use of the battery to generate electricity specifically when the wind isn’t blowing – a repudiation of that ridiculous catchphrase, ‘when the wind don’t blow, the power don’t flow’. This is different – it’s called ‘load shifting’, and it means storage sucking up giant clumps of energy when it’s freely available, and spitting it out into the grid when it’s most needed. 

So, here’s a little chart of the Tesla battery that shows a bunch of 5 minute time intervals – specifically, all the times the Tesla battery was generating electricity when wind speeds were low (as shown by a bunch of nearby wind farms at low output levels). The intervals are ranked, with the highest battery output on the left, decreasing towards the right. 

So that’s that. Sorry, everyone. We can say, with confidence, that when the wind don’t blow, this power still flows. There’s about 600 5-minute intervals, or roughly two full days, above that show it with absolute certainty.

Show the chart above to a grid engineer or an energy nerd, and they’d shrug. This stuff is totally unremarkable; but the belief that resource variability makes wind and solar pure evil is firmly held and angrily defended, and my bet is that that’s going to change in the next few years, just like the ‘renewables are expensive’ myth did in the previous few. 

Extra bits and pieces that you might like

First of all, figuring out when the wind doesn’t blow is trickier than you might expect.

The Bureau of Meteorology doesn’t make detailed historical wind speed data easily available. You can use the output of wind farms as a rough estimate of wind speed, but there’s a catch – wind farms can be offline for a bunch of reasons, including a generation target from the grid operator, a circuit breaker trip on site, or a transmission line problem somewhere else. If a wind farm’s offline, it might not just be low wind. 

There’s a wind farm near (ish) the battery called Mt Millar wind farm, and it’s ‘non-scheduled’ – this means it doesn’t receive generation targets from the grid operator, because it was built before this was a thing you could do.

To demonstrate that this wind farm operates freely, whilst others are sometimes ‘capped’ at a maximum output by the grid operator, here’s a scatter plot of the unconstrained wind farm (Mt Millar), vs the others (sometimes constrained): 

That dark green band along the top is roughly where those semi-scheduled wind farms are often limited in their output (though there’s an odd cluster for Mt Millar around 40-50 megawatts too, I wonder what that is?). Regardless, it shows that we can get a good guess at wind speed if we combine them all – and so, in the first chart, I filtered ‘low wind’ intervals by choosing times when output was low for all these facilities. 

Something else I noticed is that even though the battery isn’t really there to do load shifting, it still generates electricity roughly matched with SA’s demand: 

And, as you might expect, the battery’s consumption profile is the opposite of its output: 

Roughly, for 2018 at least, the Hornsdale battery’s been generating more when demand is higher, and consuming more when generation is higher – helping out in a small way with supply and demand imbalance like a conventional generator. 

These are all extra points for a system that wasn’t really specifically designed or installed for ‘load shifting’ – long, slow controlled contributions of big pockets of stored energy. You’ll also see this called ‘arbitrage’, basically, just serving power when prices are high specifically to make money: 

There’s a great piece from May in Renew Economy you can dig into further. Or, just soak up how great it is that something so simple and obvious is also so revelatory and game-changing. Peace out. 

Data sources

Hornsdale power reserve 5-minute data – from OpenNem 

Wind farm output (except Hornsdale): ARENA’s Australian Renewable Energy Mapping Infrastructure site (AREMI). 

SA electricity demand: Australian Energy Market Operator 

Header image – this website