Hornsdale Power Reserve: “outsmarting” lumbering coal units by four seconds?

This is an update on a previous post about the claim that the Hornsdale Power Reserve (in South Australia) is helping to prevent blackouts in Melbourne (in Victoria), roughly 1,000 km away from each other. In that post, I rejected that idea, saying that this was highly unlikely because the capacity of the Hornsdale battery is way too small to do so.

In the meanwhile, I got a link to an article that seems to describe such an event. At the end of 2017, just after the Hornsdale Power Reserve was put into use, a coal fired power plant unit in the state of Victoria tripped, causing a sharp drop in frequency and that triggered the Hornsdale battery to supply power to the grid. Its response was much quicker than that of a coal fired power plant commissioned to compensate for the loss.

Melbourne was not specifically named in the article, but the question doesn’t really change much: did the Hornsdale Power Reserve in South Australia actually helps to prevent a blackout in the neighboring state of Victoria after a coal fired plant unit failed there in December 2017?

Did this article really provide some solid evidence that this happened? And if so, how did the battery manage to do so, considering its capacity is only 100 MW and can store 129 MWh?

The article is published in RenewEconomy, it is titled “Tesla big battery outsmarts lumbering coal units after Loy Yang trips” and is written by Giles Parkinson.

Let’s first look at what happened exactly on December 14, 2017. This is how it is described in the article (my emphasis):

Last Thursday, one of the biggest coal units in Australia, Loy Yang A 3, tripped without warning at 1.59am, with the sudden loss of 560MW and causing a slump in frequency on the network.

What happened next has stunned electricity industry insiders and given food for thought over the near to medium term future of the grid, such was the rapid response of the Tesla big battery to an event that happened nearly 1,000km away.

Even before the Loy Yang A unit had finished tripping, the 100MW/129MWh had responded, injecting 7.3MW into the network to help arrest a slump in frequency that had fallen below 49.80Hertz.

Data from AEMO (and gathered above by Dylan McConnell from the Climate and Energy College) shows that the Tesla big battery responded four seconds ahead of the generator contracted at that time to provide FCAS (frequency control and ancillary services), the Gladstone coal generator in Queensland.

But in reality, the response from the Tesla big battery was even quicker than that – in milliseconds – but too fast for the AEMO data to record.

Importantly, by the time that the contracted Gladstone coal unit had gotten out of bed and put its socks on so it can inject more into the grid – it is paid to respond in six seconds – the fall in frequency had already been arrested and was being reversed.

Gladstone injected more than Tesla did back into the grid, and took the frequency back up to its normal levels of 50Hz, but by then Tesla had already put its gun back in its holster and had wandered into the bar for a glass of milk.

This is the graph that shows what happened:

Hornsdale battery "outsmarts" coal 2017-12-14

This is the order of the events described in the article:

  1. The Loy Yang A3 unit suddenly tripped
  2. The loss of output creates a drop in frequency
  3. When the frequency reached 49.80 Hz, the Hornsdale Power Reserve starts discharging to the grid within milliseconds
  4. The Gladstone coal generator starts injecting more power to the grid four seconds later.

That last part made clear why the author thinks that the Hornsdale Power Reserve “outsmarts” the Gladstone coal generator: the battery responded four seconds ahead of the coal generator. That is all nice and well, but the big question is whether that head start of four seconds actually made a difference when it comes to the recovery from the frequency drop?

Before I start throwing around numbers, a cautious remark. The author seems to use data with a 4 seconds interval:

As for timing, Loy Yang A started to lose it at 1.58:59 (NEM time). Its biggest drop, from 364MW to 176MW, was at 1.59:19, which is where frequency hit 49.8Hz, and which is where Tesla came in a matter of milliseconds (but recorded in the 1.59:23 time frame).

Gladstone 1, a contingency FCAS supplier, hopped in at 1.59:27. Loy Yang was down to 44MW by then and completely gone in the next 4 second period.

If I understand that correctly, then the Hornsdale battery discharged its load somewhere in the 1:59:20 to 1:59:23 interval and this was registered at 1:59:23. The Gladstone generator chimed in somewhere in the 1:59:24 to 1:59:27 interval and this was registered at 1:59:27.

If that is true, then it is not possible to make such fine-grained conclusions from this coarse-grained data. It is then for example not possible to claim that the Gladstone generator responded 4 seconds later than the battery, it is only possible to claim that the response of the Gladstone generator was registered 4 seconds later than the Hornsdale battery. However, for the sake of the argument, I will start from the assumption that the numbers given in the article are what actually happened and start calculating with these them as if these were correct.

According to the author of the article, the Hornsdale battery halted the drop and the frequency then went back up from then on. Looking at the graph, I see the frequency line crossing the 49.80 Hz level and shortly thereafter the frequency reaches its lowest level.

If the frequency crosses 49.80 Hz at 19 seconds, then the point where the frequency reaches its lowest values must be at 23 seconds (data from 20 to 23 seconds). That is the point where the response of the battery was first registered and where it allegedly halted the drop. From there, the frequency goes back up, this then should be the 27 seconds registration.

Hornsdale "outsmarts" coal 2017-12-14 detail

Then there is the capacity loss of the Loy Yang unit. There was a drop from 364 to 176 MW at 19 seconds, then it dropped to 44 MW at 27 seconds and nothing was registered at the next time frame (this must then be from 28 to 31 seconds, registered at 31 seconds). This means that the capacity of Loy Yang is hitting zero somewhere between 24 and 27 seconds after 01:59 am.

That is very interesting, but it is not good news for the claim that the battery arrested and even reversed the frequency drop.

Is it at least theoretically possible for the Hornsdale battery to stop this drop and even start reversing the frequency as claimed by the author? If the Hornsdale battery would be able to arrest the frequency drop, then it should be able to discharge at least as much as was lost and to reverse the drop, it should discharge more than what was lost.

We now know that the capacity already dropped to 176 MW at 19 seconds and that it was 44 MW at the registration of 23 seconds, so this means a 132 MW drop. Having a 100 MW capacity, it is not possible for the battery to fill in that gap, even if used its entire capacity for the entire time frame.

It obviously didn’t do that. According to the author, the battery was able to stop the drop by “injecting 7.3 MW into the network” (I think he is confused between capacity and output here). There is no way that the battery was able to fill in that drop by throwing only 7.3 MW of its capacity during those 4 seconds into the battle.

It might be possible to arrest the drop in the next time frame (from 23 to 27 seconds), but this is not because the battery would be so very powerful, but because the capacity of Loy Yang unit hits zero somewhere in that time frame anyway.

Looking at those numbers, I don’t see how the battery could arrest this drop by discharging at a capacity of 7.3 MW. The battery might be blistering fast and outsmart every other electricity generator in the market, but it is inadequate to arrest such a steep capacity drop. Also, the capacity of the Loy Yang unit was only seconds away from hitting zero by that time anyway.

That is why I still stick to my earlier standpoint that the battery didn’t have the means to do what is claimed that it did. Despite the hyperbole in the article, the battery played no significant role in the stabilizing nor in the reversing of the steep drop in frequency. The numbers don’t support this.

But then, why does that graph show that the frequency went back up in the interval just after the battery starts discharging (this uptick seems to be the reason why the author made the claim that the battery reversed the frequency drop)? We now know that the frequency goes back up again at 27 seconds after 1:59 am, but we also know that this was the interval where the first response of the Gladstone generator was registered. The Hornsdale battery might outsmart coal fired power units on speed, those coal fired power units surely outsmart the Hornsdale battery on raw power. Even if the battery was able to arrest or even reverse the frequency a tiny bit, the Gladstone generator was already at it and would overpower whatever the battery could do.

It could also be other generators providing FCAS services that came to the rescue. The author seems to suggest that the Gladstone generator was the only one that was triggered to compensate for the capacity loss. That is a very simplistic vision of the workings of a power grid. There are many layers of security and reserves, some with quick intervention time, other with slower intervention time. I would find it very unlikely that in both states (Victoria ánd South Australia) only the Gladstone coal generator was triggered to fill in the gap and no other generators would step in within the roughly 30 second period from the beginning of the tripping (1:58:59) until zero output (1:59:27). That is hard to believe. Chances are that this arresting and reversing of the frequency was partly done by other generators springing into action, already triggered by the frequency loss at an earlier stage than the Hornsdale battery did.

There are other indications that the battery didn’t do that much. The Hornsdale battery was not even part of the line of defense against such event and the author mentioned this twice in his article. This should also be clear from the rather minimal response of the battery to this steep frequency drop. It discharged for about 3 minutes and there were only a few peaks above 7 MW. My guess is that it used on average about 5% of its capacity. If it was actually in the line of defense against such events, then it surely would have reacted with more than a fraction of its capacity (unless it was out of juice at that time).

It also most probably would have responded earlier. According to the author, the battery responded at the 49.80 Hz threshold. This is rather late in the game. That is already outside the normal operation range of the South Australia grid (the normal range is between 49,85 and 50.15 Hz). If it really would be in the line of defense, then it would been set to a much tighter threshold (but then it might come into trouble with its limited capacity).

A better explanation for the underwhelming response of the battery is that it is commissioned to respond to (small) instabilities on the grid. That is where it is very good at and what it probably tried to do, whether it could solve the problem or not. It might even have earned quite a buck providing FCAS services to the grid at that moment.

The final blow for this claim comes from the author himself. This is what he stated in an update below the article (my emphasis):

And no, we never suggested this averted a blackout. The point of the story was what Tesla could do.

He indeed never said that the battery averted a blackout. He however did suggest that the battery single-handed stopped, even reversed, a catastrophic event happening 1,000 km away and then left it to that “lumbering” coal unit to wrap it up. Some seem to conclude from this that the contribution of the battery was somehow crucial in resolving the frequency drop.

The author apparently meant to tell the story what the battery COULD do, not what it actually DID. The Hornsdale Power Reserve definitely might have a blistering fast response time, but the numbers show that what it did during that head start was surprisingly little.

Concluding, did the Hornsdale battery prevented a blackout in the state of Victoria on the 14th of December 2017? The article didn’t give any proof of that and even the author had to admit that this wasn’t the case. The author however used plenty of hyperbole describing the Hornsdale battery response to the incident and that may have fooled readers into believing that the battery played a crucial role in the recovery of the frequency drop, therefor coming to the (wrong) conclusion that the battery helped preventing a blackout after the failure of a coal fired power unit 1,000 km away.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s