I was puzzled late last year when a bit of hype emerged about the performance of South Australia’s “Big Battery” (Hornsdale Power Reserve) following the trip of a large coal-fired unit in Victoria early on the morning of December 14th.
An article appeared on the Reneweconomy energy news site which might have been read as crediting the “millisecond response” of the battery for saving the grid from some sort of catastrophe, and contrasting this with the slower response of “the generator contracted at that time to provide FCAS (frequency control and ancillary services), the Gladstone coal generator in Queensland”.
A few things here just didn’t sound right. For a start, at any one instant there are always many generators enabled in the NEM to provide FCAS – services which come in eight different flavours, from second-to-second frequency regulation, through to delayed-reaction contingency response over timeframes of up to five minutes, as explained in Jonathon Dyson’s excellent WattClarity articles on FCAS. So focusing on what Gladstone did or didn’t do clearly couldn’t tell the whole story. And whatever the speed of HPR’s response, being a much smaller facility (100 MW) than the tripped Loy Yang A unit 3 (560 MW) it was never going to be able to stabilise system frequency all by itself. That requires replacing essentially all the tripped unit’s lost power output with increased output from other sources.
At the time I was busy with other stuff and thought I’d go back and have a closer look a little later. Or a lot later as it turns out.
A chart in the original Reneweconomy article showed the timing of events with the Loy Yang A unit’s output falling away, and a sudden injection of power from the battery (HPR):
Source: Reneweconomy, 19 Dec 2017
Looks impressive, but a closer glance at the scaling of the chart shows that the maximum injection of power from HPR was less than one sixtieth of what was lost when Loy Yang A3 tripped. If we use a common scale for both facilities then the picture is a little different:
But for at least some Reneweconomy readers– see the voluminous comments thread under the article – this was unimportant detail and it was the speed of that response that must have been decisive. After all, according to the article “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”. Now, speed of response is a factor in response to contingencies but it’s far from being the only one.
To be fair, a subsequent update to the article makes it clear that there was never an intention to suggest that the response of the battery “averted a blackout”, and that the author’s point was to show “what Tesla could do”.
But to see all this in better context, the point of this post is to show what the system as a whole does when a large generating unit trips, covering the response of a broad set of generators.
Doing this requires grappling with AEMO data that is a lot more obscure than the usual 5-minute and 30-minute market data dealt with by products like NEM-Watch and ez2view (as if that data isn’t obscure enough!). Another reason for the delay in getting this article together.
A quirk of the FCAS markets means that AEMO publishes operational data from its Energy Management System which captures SCADA data every four seconds on generator outputs, system frequency, and other arcana. For very good reasons commercial electricity market viewer / analysis products don’t deal with this level of detail, but it’s the best available public source to understand what happens in response to a large unit trip. It’s this sub-5 minute world that FCAS is concerned with. I’m not sure if market operators anywhere else in the world routinely publish data at this level of operational detail, but I’d be surprised if many do.
For a start we can look at what NEM generation levels and system frequency were doing for the five minutes up to and just after the Loy Yang A3 trip:
Here we are looking at a single five-minute dispatch interval. The blue Generation line is essentially total scheduled generation in the NEM, and the red Target line is what AEMO has instructed that generation to produce by the end of the interval. The fact that generation exceeds the target level by around 80-100 MW for most of the interval could reflect a number of factors, for example scheduled demand being higher than forecast by AEMO (requiring some generators to be dispatched upwards in the raise regulation FCAS market, to maintain system frequency at 50 Hz).
The Frequency line shows system frequency which is nominally 50 Hz, with a normal tolerance range of +/- 0.15 Hz, indicated by the upper and lower reference lines. The grey shaded time interval covers just over 30 seconds (from 1:58:55 AM to 1:59:31 AM) during which the Loy Yang A unit’s output fell from nearly 550 MW to zero as the generator went offline.
You can clearly see the effect on system frequency – as input to the grid drops, its frequency begins to slow down, just like a vehicle travelling steadily up a hill would decelerate if its engine power fell. In fact power system frequency is the direct analogue of speed, because it reflects the rate of rotation of all the spinning synchronous generators and motors connected to the system.
You can also see a notch in the overall generation level corresponding to the falloff of the Loy Yang A unit, but interestingly this is nothing like as large as the 550 MW loss of generation from that unit, and quickly bounces back to close to where it was – so what’s going on? And finally as generation bounces back, system frequency stabilises although it still sits below the lower tolerance band.
We know the tripped Loy Yang A unit didn’t suddenly come back online, so other generators must have made up the deficiency – which ones? The next chart, covering 30 seconds either side of the trip, is a bit busy, but it highlights the fact that many generators across the NEM responded, prompted by the falling system frequency. This is a great example of FCAS in action:
The top panel is essentially a closeup of the same Reneweconomy chart that prompted all this, showing the rundown of the Loy Yang A unit (LH axis) and the response of the Tesla battery / Hornsdale Power Reserve (RH axis). The middle panel shows the increase in output from the whole subset of generators around the NEM that materially responded to the event (plenty didn’t). To be clear, except for the battery and the Jindabyne hydro unit in blue, these generators were already online and producing before the trip – they simply increased their output, using what’s known as “spinning reserve”. In terms of FCAS markets, this response would be primarily the fast raise (6 second) and slow raise (60 second) contingency services enabled by AEMO. For emphasis, I’ve highlighted in red the responses from the SA battery and also from Gladstone in Queensland. It’s very clear that they were only a small part of the overall story, nor was either the first to respond to the event. In fact most of the response came from old-fashioned coal-fired stations. That’s not at all surprising, because in this overnight timeframe with low demand it’s predominantly coal generation that’s online, and contrary to what you might expect from reading elsewhere, during low to moderate demand periods online coal fired generation can usually vary its output fast enough, in aggregate, to respond perfectly adequately to this kind of event.
[Finally a note on chronology – it’s curious that system frequency appears to start falling well before the major changes in generation, by about 12-20 seconds, and then to stabilise before the bulk of the generation response arrives. Whilst the data used in these charts is all timestamped according to AEMO’s public files, I strongly suspect there are some measurement lags or timing offsets present in this raw operational data, and it’s likely that in real time the inflection points in frequency and generation levels respectively line up much more closely than shown above.]
About our Guest Author
 |
Allan O’Neil has worked in Australia’s wholesale energy markets since their creation in the mid-1990’s, in trading, risk management, forecasting and analytical roles with major NEM electricity and gas retail and generation companies.
He is now an independent energy markets consultant, working with clients on projects across a spectrum of wholesale, retail, electricity and gas issues. You can view Allan’s LinkedIn profile here. Allan will be sporadically reviewing market events here on WattClarity Allan has also begun providing an on-site educational service covering how spot prices are set in the NEM, and other important aspects of the physical electricity market – further details here. |
Absolutely brilliant analysis.
thanks Alan. I am not sure I understand your point here. So thermal generation on AGC responds. Yes, we know that. Surely the interesting here, as pointed out in Renew is the relatively superior responsiveness of batteries. The fact that the volume of their response is small misses the point.
Bruce, in line with WattClarity’s theme of “making Australia’s energy market understandable”, the purpose of the post is primarily to present a clearer overall picture of how the system responds to a trip, especially for readers not as familiar as you might be with how the electricity grid operates and the characteristics and behaviour of different types of generator. I don’t think anyone could say that the original Reneweconomy article had that aim in mind!
In this particular incident, as Andy points out below, the response of the battery is neither remarkable (a battery can quickly switch on from zero output – who knew?!) nor critical, and of course nor is the response of any other individual generator. There are definitely all sorts of valuable characteristics and demonstration effects that HPR and other new technologies bring to the energy transition – but this event doesn’t really illustrate them. In fact it’s my understanding that the battery’s capabilities have had to be “hobbled” somewhat to fit in with operating in an old-style predominantly thermal system.
Allan
(PS Update 7/4 – see the just released report by AEMO on initial operating experience of HPR at http://www.aemo.com.au/Media-Centre/AEMO-Hornsdale-report – gives a much fuller overview of the battery’s capabilities and services)
Thanks Allan, perfectly reasonable points. Thank you.
The fact that several generators started to increase load before NEM frequency decayed thru 50Hz implies their loading instructions were being received via AGC as participants in FCAS (AGC usually also receives digital signals from stations such as trips to pre-empt frequency decay). The rate of these load increase will most likely be driven by the control algorithms in the AGC rather than the governor actions of the active units. Little natural governor action appears to be present. (And many generators in the NEM not participating in FCAS market will actually have their governors switched off) Hornsdale response seems to have commenced after frequency fell below 50 Hz and looks to be proportional to the (0.2Hz-deadband)/2 x Capacity – about 9MW based on 4% governor droop. All looks about right. Not sure there is a story here for the battery guys (or anyone else really). Very good, practical analysis however Allan.
Great writeup Alan, your stacked bar chart of unit responses is excellent. Have to point out something this explainer overlooked though: there’s two ways to rebalance supply and demand, and recover to nominal system frequency following a generation contingency: you can increase generation, or reduce demand (or both). Indeed, in this trip, both types of responses would have occurred. In the dispatch interval where this trip occurred NEMDE had enabled 19/50/58 MW of R6/R60/R5 FCAS from demand side loads (across three dispatchable units)… so the generation responses in your chart are only part of the story. (Of course, you’re forgiven for being unable to represent the demand side contribution in your chart, as the data isn’t immediately published in AEMO’s 4s SCADA logs)
Really good point Matt – I should have mentioned this, along with the intrinsic “load relief” that accompanies a frequency dip. I did actually check APD (Portland Smelter) which does show up in the 4s data and was enabled for 17/11/11 contingency raise, but its response was minimal. As you say, the other demand side DUIDs enabled in that interval don’t appear in the 4s data set.
The world’s best engineers and technologists have an inborn aptitude for the disciplines in which they have studied and worked.
Until such people are given their due respect and tasked with planning the future for Australia’s electricity supply, we are going to slip further into the fiasco now unfolding.
It’s also clear from the many comments posted to its website, that ignorance and naivete are prerequisites for believing Reneweconomy’s propagandist messaging.
With the design and development of the nation’s electricity supply being taken over by politicians and activists, is it any wonder that the horrendously expensive and fragile grid that has resulted comes about?
It appears that collapse and chaos is the only way to bring the nation to its collective senses.
That would be great Rob if some of the ultra conservative energy professionals of yesterdays-year hadn’t been so resistant to the need to replace fossil fuel generation with renewables and storage. Like it or not means big changes in the operation of the grid.
Fortunately the tide seems to be turning and even the older engineers resistant for a great time to renewables are finally seeing that back to the future is not gonna fly, New solutions to old problems and new problems will need to be found. That’s surely what engineering is about, sticking with what they already know (i.e. fossil generation only) is not engineering so much as a basic fear of change, and it’s no longer an option for engineers, politicians or finance people.
It’s a great shame it took a rag-tag crew of activists to smoke out the blockers within the gentaliers, network owners and operators and regulators. I work with a brilliant group of engineers, coders and finance people who model what energy grids of the future may look like and assess where the risks and opportunities lie. It’s great when engineers can solve problems in this way and show creative initiative in contradistinction to modelling ostriches with heads in the Nullarbor sand.
Yet even with these great solutions, engineers, modellers and financial people are still needing to play activist to get the recognition and reform at the level of government policy and mechanisms put in place to take advantage of the now cheaper renewable generation and storage possibilities for better economics. If RenewEconomy becomes cheer squad at times, you need to recognise the mammoth task in converting old thinking about energy in not just the energy industry but also finance, heavy industry, resources and politics.
Thanks Allan for the great analysis. Very clear. I echo with you on the problem with the time tags, which makes difficult to see how much increased power of generators is from the governor control and how much is from 6sec and 60sec services. The 6 sec and 60 sec services should kick in only when the frequency deteriorates to a level below 49.85Hz. Is that right? Although many generators choose to quit their governor control because of wear-and-tear cost and not being properly awarded in the market, I believe there are still some royal generators fulfill their responsibilities.
Thanks for the feedback Jin. You’re correct that in most cases the 6s and 60s raise services would not kick in until frequency drops to below 49.85 Hz, although I think in the case of providers with Switching Controllers (smelter potlines might be an example), there might be lower frequency setpoints before raise services are triggered.
While I didn’t go into this in the article, looking at which generators were enabled for fast and slow raise services vs the responses actually seen was interesting – there were some significant responses from units not enabled for 6s/60s raise – HPR fell into this category – while other units enabled for 6s/60s raise appeared to provide smaller amounts than expected, based on their enablement level. However this last is difficult to be certain of from the 4s data, and accurate verification of contingency responses requires higher resolution data as set out in the MASS.
Thanks Allan! Great comments!
Allan,
Great article.
http://www.abc.net.au/news/2018-04-06/tesla-battery-outperforms-coal-and-gas/9625726
Our ABC giving their un-bias view as usual – I don’t know why they get away with it unchallenged.
I’m not sure of your point there Dave.
The ABC article was based on a report just released by AEMO (http://www.aemo.com.au/Media-Centre/AEMO-Hornsdale-report). On a first skim I think the ABC article is a pretty fair summary of that report.
As the AEMO paper makes clear, the battery does provide very fast and high quality response for both regulation and contingency FCAS, and in very different circumstances from the LYA3 trip referred to in my post, that fast response (harnessed via the newly developed SIPS scheme, see p4 of AEMO’s report) could be very important in maintaining South Australia’s connection to the rest of the NEM and to keeping the lights on in that state.
The point of my post was not to knock the battery at all, simply to explain in full context what happens when there’s a large generation trip in the NEM, and to correct the possibly misleading impressions people might have gained from the original Reneweconcomy article.
A huge problem in the energy ‘debate’ at the moment is people’s tendencies to take sides and hold to their positions regardless of facts – something applying across the spectrum of what Paul has termed the Emotion-O-Meter (https://wattclarity.com.au/2016/07/the-emotion-o-meter/ and covered in many subsequent posts).
Allan
Thanks Allan. I fully agree with you that people should not take sides when looking at the FCAS. Although battery storages could kick in very quickly, but it is unfair to say the conventional generators are too slow to respond. Although governor control is slower than the battery, but think about the big inertia that kicks in immediately after the big disturbance. The kinetic energy stored in the conventional generators go nowhere, but convert to the electric energy trying to fill in the gap. Compared to the battery contribution to capture the frequency drop, the contribution of conventional generators should also be applauded. Saying they are slow is a little bit unfair to them.
Fantastic analysis Allan.
That second graph with the same scale really shows how much the battery has been overrated in the media. In the first graph, it appears that the battery ‘prevented a blackout’ by meeting the shortfall when in fact it was only a small part of many FCAS responders. The capability of the battery to respond so fast to stabilise the frequency should not be ignored however.
This shows the bias towards renewables on the RenewEconomy website, they disregard selected facts to make renewables look good.
My project is on FCAS and FFR markets and I’ve been reading plenty of material online but never seen such clear information. Thank you.