Towards net zero: is battery storage leading the way?

Volumes of Dynamic FFR by technology type since 2014. EFR and DC are provided by battery storage.

As the UK decarbonises and real time balancing of the electricity system becomes more challenging, battery energy storage will play a crucial role in maintaining a stable system. The UK’s Electricity System Operator, National Grid ESO, has the ambition of operating a zero-carbon electricity system by 2025. This growing requirement for robust real time balancing of the system has been the dominant revenue driver for battery storage projects over the last few years, via Firm Frequency Response. As the ESO begins a journey of reform in UK frequency regulation via new services such as Dynamic Containment, and 2025 draws closer, we look back at how the system has changed and the impact that batteries have already had, through the lens of grid frequency. 

The UK’s electricity system has been rapidly decarbonising over the last few years: renewables accounted for 19.1% of generation in 2014. In 2020, this figure stood at 44.1%. As we move towards a zero-carbon grid, a higher proportion of electricity comes from renewables with no rotating mass, and this has an impact on how the system is operated – not just in dealing with GW swings in wind generation within a few hours, but in the delicate balance of supply and demand over a matter of seconds. 

The spinning turbines of traditional power generation give rise to system ‘inertia’: similar to a bike wheel that keeps turning when you stop pedalling, inertia is an important part of the stability of a power system. Grid frequency is then the needle showing the stability of that system; when all things are equal, it is 50Hz. When they are not, blackouts can occur – frequency plummeted to 48.8Hz on August 9th 2019, leading to nationwide power cuts. 

Here, we look in some detail at grid frequency since 2014: at the trends in frequency ‘events’ (when frequency spikes or dips in response to an outage), and in the way the system recovers. Because we have many more renewables now, system inertia has decreased. We see that year on year, grid frequency is becoming more volatile (see Figure 1), and events are becoming longer. However, the rate of change of frequency (RoCoF) is becoming less severe (Figure 2).  

No alt text provided for this image
No alt text provided for this image

So why would RoCoF, a key indicator of system stability, be getting ‘less’ bad with decreasing inertia? One explanation for this is in the new technologies which have been coming onto our system in recent years. National Grid ESO, the UK’s Electricity System Operator, procure an array of services to balance supply and demand in real time. One of the most important of these is dynamic firm frequency response (FFR), in which a plant moderates its output to help balance the system in real time, given the system frequency. It has become dominated by batteries, which, when operated well, can respond reliably and nearly-instantaneously to frequency events – and crucially for net-zero, cleanly.  

Figure 3 shows the volumes of frequency response from different technology types. Since 2014, batteries have gone from providing no FFR volume to now providing virtually all FFR volumes. This, alongside frequency regulation volumes delivered by batteries via Enhanced Frequency Response (EFR) tenders, and more recently the Dynamic Containment (DC) auctions, mean the UK now has close to 1GW of low carbon, ultra-fast battery storage providing real time frequency regulation to balance the electricity system. 

No alt text provided for this image

Looking at two comparable frequency events, one from summer 2017 and one from summer 2019, we can corroborate this theory. There were similar conditions on these days – a similar national demand (INDO), wind outturn, and a sudden power loss equivalent to around 2% of total demand (Table 1). The event in 2019 is after the evening peak, when national demand is decreasing, while the one in 2017 is during the evening ramp. Figure 4 shows the frequency trace for these two events, on a common time axis. 

No alt text provided for this image
No alt text provided for this image

Despite the slightly smaller loss in generation causing the 2017 frequency event, it was more severe – with a larger RoCoF and reaching a lower frequency: 49.57 vs 49.70 Hz. The post-event response overshoots and subsequently the frequency remains high for a couple of minutes, whereas the 2019 event returns to a frequency close to 50Hz. While time of day considerations may be at play, it is interesting to consider the volumes of frequency response on the system during both of these periods: see Figure 5. The earlier event had around 30% more dynamic FFR volume. Logic then says the system should stabilize more quickly to a comparable loss, but the opposite is true. However, there is a marked shift in the makeup of that volume; batteries in both FFR and EFR provide significantly more of the stack in 2019.

No alt text provided for this image

The fault recovery of the system appears much better in 2019, suggesting all of these batteries are really improving our response to unplanned plant outages. Returning to the bicycle analogy, it’s much easier to stop a bike from falling over if you catch it just as it starts to topple rather than just before it hits the ground. Batteries are so much faster as compared to the gas, pumped storage and hydro plants which dominated just a few years ago, and it means the system operator needs less volumes to provide the “same” service, so it’s more efficient. Batteries are also very happy performing these low utilisation services – we see very low levels of cell degradation in battery systems performing FFR over long periods of time, and it can be easily stacked with other services. 

All this is great news for net-zero. Leaps and strides in energy storage technology over the last few years, alongside the platforms which operate them, mean we can integrate far more intermittent renewable generation into our electricity mix – whilst ensuring the system remains robust and secure, crucial in our highly electrified society. As we build more wind and more solar, the importance of battery storage technologies in operating a decarbonised, digitalised, democratised and decentralised system will continue to grow. And, not just in frequency regulation but across the board of balancing requirements. 

Written By Grecia Monsalve

Dynamic Containment – a look into the crystal ball

After the excitement of January 2021, which saw Day-ahead prices of £1,500/MWh and Balancing Mechanism (BM) prices of £4,000/MWh, it feels strange to write about the comparatively mundane topic of Dynamic Containment (DC). However, it’s perhaps easy to forget that current prices of £17/MW/hr (resulting in battery revenues of £140+/kW/yr.) are just as exceptional and are going to be vastly more important to the returns from battery systems this year than value obtained through trading -something almost no one was saying would be the case in 2021 a few years ago when frequency response was old news.

The market is clearly in a phase of transition so the most important questions for investors and developers are: how long are these prices likely to last, and what will things look like afterwards?

What is Dynamic Containment?

Dynamic Containment is a new, fast-acting, dynamic post-fault frequency response service, the first of National Grid ESO’s (NGESO) new suite of frequency products. It is designed to stabilise grid frequency in case of large generation or consumption drop-outs, such as the interconnector trip on 28th Jan.

Dynamic Containment response from Open Energi unit during IFA2 trip on 28th January 2021.

Anyone who has been around the market for a while will know that trying to predict frequency response prices has bitten people badly in the past – we only need to look back to the crash of Firm Frequency Response (FFR) rates in 2018 and the current rates observed in DC. However, improved transparency from NGESO and the dominance of lithium-ion storage now makes it a little easier to understand the market and predict where things are heading (famous last words…).

So, let’s look back at the recent past and the three main frequency response markets to understand the dynamics at play.

Graph showing Frequency response prices from March 2020 to February 2021.

As expected, supply and demand volumes are the driving factor for prices, illustrating a competitive and functioning market. Price discovery varies by market based on the frequency of procurement – the monthly dynamic FFR market takes much longer to settle at a new price than the daily DC market, which found its cap within a few days. DC is more valuable to NGESO than Dynamic FFR meaning those who can have left FFR to provide DC. In turn this has reduced the competitive pressure in the FFR market and raised prices for the monthly tenders. Essentially, the additional 500MW requirement which arrived with DC implementation has massively unbalanced the market in favour of providers.

This means that for forecasting the price forwards in the near term, the most significant elements are: NGESO’s requirement for frequency response volumes, the volumes of storage competing for these services, and the relative price caps NGESO sets for each service.

Dynamic Containment requirement for 2021

On the demand side, luckily NGESO have published their forecast requirement for DC for the entirety of 2021, along with dynamic FFR requirement through to July 2021. Interestingly this shows a significantly higher requirement in the summer, which is perhaps to be expected given lower system inertia at this time.

Forecasting supply is a little more tricky. We know no other technology than storage looks able to meet the increased technical requirements of DC (namely speed and response) at scale and therefore we can use the pipeline of storage being developed over the next few years. Given the favourable investment environment and current sky-high DC prices, we can expect most of these to be bought forward to completion (this could see 500MW+ of storage being built this year!) and all of this capacity should be able to immediately enter the DC market. We expect a portion of dynamic FFR participants to continue to switch across to the higher value DC service. There are also some long-term FFR contracts coming to a close, along with EFR contracts doing the same towards the end of 2021.

Putting all of this together, and we start to get a picture of what things might look like in 2021.

Graph showing the supply imbalance up to Dec 2021.

As we can see, this current market oversupply looks likely to remain in place until at least Q4 2021. It is actually incredibly hard to create scenarios that brings this point forward, given the higher requirement across summer and capacity not expected to enter the market until the autumn. This leads to the current high prices being maintained until late 2021 (as long as the price cap remains the same). When market saturation does occur, we would expect DC prices to quickly fall towards those available in the weekly and monthly FFR markets, and track these downwards from that point on at a rate similar to what has been observed historically.

As we can see, this current market oversupply looks likely to remain in place until at least Q4 2021. It is actually incredibly hard to create scenarios that brings this point forward, given the higher requirement across summer and capacity not expected to enter the market until the autumn. This leads to the current high prices being maintained until late 2021 (as long as the price cap remains the same). When market saturation does occur, we would expect DC prices to quickly fall towards those available in the weekly and monthly FFR markets, and track these downwards from that point on at a rate similar to what has been observed historically.

Other things which could affect the value in 2021 (likely in a positive way) would be the introduction of high response DC and moving to EFA block procurement, albeit neither of these is certain to happen this year.

2022, by contrast, is a whole lot more complicated. The biggest factor at play here is the possible introduction of the remaining two new frequency response services: Dynamic Moderation and Dynamic Regulation. Initially scheduled to be introduced by March 2022, we now expect this point could be delayed towards the end of 2022. This then brings a few questions which right now are difficult to answer: will this see additional volume as with DC? What price will NGESO be willing to pay for these services? What assets will even end up being able to provide the high utilisation regulation service? If these products are introduced along the published timescales, we would expect to see a similar distruption of the market as seen with the introduction of DC with high prices in the short-term, albeit perhaps one resolved quicker as there will be more storage online to meet the requirement.

Other big questions are: what state do the 200MW of EFR batteries emerge from their contracts? Will they be able to jump immediately into providing these new services? And finally, will we see another bumper year of newly built batteries as we expect in 2021?

To keep things simple, if we assume the new products won’t be introduced for much or all of 2022 then as long as dynamic FFR volumes are maintained this should lead to similar market dynamics being at play in 2022. All in all, this could see 100MW or so swing the market significantly one way or the other, depending on whether the higher summer requirement sees the supply imbalance returning and prices moving towards the cap again.

Graphs showing the energy supply type differences and pricing forecasts for 2021.

Frequency response prices have shown previously they have the potential to fall as low as £3/MW/hr in a saturated market, given the operational costs of providing frequency response are so low for batteries (a bit of efficiency losses and degradation). This time however arbitrage markets will provide significant opportunity cost – given the improved access to and consistent value demonstrated by trading and the Balancing Mechanism in particular. Batteries no longer have to accept rock bottom  revenues and become price makers in the frequency response market.

Beyond 2022, this optionality is going to be the main factor driving frequency response prices, as we expect storage capacity to start to significantly exceed the requirement for frequency response services. This will see prices starting to reflect the value available in these other markets, especially with EFA block (or even half-hourly) procurement. At this point, we would really hope to see price caps on frequency response prices removed or lifted to enable the market to function efficiently – otherwise next time prices of £1,000/MWh occur in the day-ahead expect to see much more volume exiting the market to chase this value. An example of a market functioning across multiple revenue streams can be found in Australia, where system operator AEMO manages procurement of both frequency response and energy, with price caps of $15,000. This sees prices vary significantly on a half-hourly (or even 5-minutely) basis in response to the requirements of the grid and value from arbitrage in the spot market.

Graph showing the Australian FCAS price fluctuations November 2019 to Jan 2020.

So, in summary: we can expect 2021 to be a very good year for storage assets capable of providing Dynamic Containment, but if anyone is telling you we will see the same in 2022 treat this with a huge pinch of salt. Be sure to engage system integrators and optimisers early to ensure assets are ready to go into DC on day 1. And ultimately, optimisers having the ability to seamlessly trade across 5+ potential markets whilst managing state of charge and warranty constraints will be vital to maximising battery returns.

Scarcity pricing in the Balancing Mechanism

The last couple of weeks saw some power stations make some bumper profits as cold weather drove scarcity pricing. The greatest profits were made by some large gas (CCGT) plant in the balancing mechanism on Friday 9th and Wednesday 13th January, earning up £4,000/MWh (around 100 times the normal cost of power), so what was it that enabled these plants to earn large revenues compared to other assets? 

The first thing to bear in mind is that the Balancing Mechanism (BM) is not a single market with a well-defined supply and demand but a marketplace for several system needs. The BM is used to correct supply demand imbalances that naturally occur (eg from demand forecasting errors) but crucially also to manage constraints, voltage, frequency, inertia, reserve (and more) in real time (sometimes augmenting specific tendered markets for these services). 

No alt text provided for this image

If we examine the accepted actions by the System Operator (NGESO) on the afternoon Wednesday 13th after the event we can see this immediately given the large volume of both Bids and Offers accepted. If the only purpose was to correct the overall system imbalance it would be very inefficient to pay one plant £4000/MWh to turn up while turning others down at £0/MWh.

 Bids and Offers accepted 1700-1730 on Wednesday 13th. Source: LCP Enact.

These actions only start to make sense once we consider other requirements which the BM is used for, namely Reserve. Operating Reserve is needed to ensure sufficient backup from sudden impacts (like an interconnector tripping off or wind generation dropping off) and therefore GWs of back-up power is required to be available and able to respond in minutes. Reserve can be created by the System Operator through tenders ahead of time (STOR & Fast Reserve), bilateral contracts (SpinGen) and by creating headroom in the BM.

Headroom refers to difference between current output level and the Maximum Export Level (MEL) on thermal plant, the amount they can ramp up to and hold indefinitely. CCGTs take hours to start from zero ouput, but once operating above a Stable Export Level (SEL) they can change output in minutes and so contribute to Operating Reserve.

No alt text provided for this image

A CCGT brought on to provide Headroom. Source BM Reports.

This is where we see the perverse incentive of being a large, slow moving beast sometimes emerges in the BM. Faster (traditionally smaller) plant like OCGTs, batteries and pumped storage can ramp up in a timescale of minutes (or faster) from a standing start which means they are naturally providing Reserve simply by offering volume into the BM. Whereas CCGTs must be turned on to SEL (and paid for this) to create Reserve.

Put yourself in the shoes of a Control Room engineer as the evening peak approaches: if you are looking at the resource available and think you might come up short clearly the best thing is to start up the slower plants in preparation for this. But if no incident does happen or the demand turns out lower than expected then the faster assets will likely not have been used during this period, whereas the CCGTs will have been renumerated heavily from the start-up instruction. 

This is what happened on Wednesday where each of four CCGTs were ramped up from zero to their Stable Export Limit (SEL), ie the least amount NGESO could procure, to create as much Headroom as possible. However, it turned out the system was long over the peak and with no incidents occurring meant many OCGTs received no instructions, despite offering much cheaper volume into the BM (it should be noted that size also plays a role here, the CCGTs offer much more volume). 

No alt text provided for this image

 Accepted Offers from CCGTs to SEL 1700-1730 13/01/21. Source LCP Enact.

Offers from OCGTs not accepted 1700-1730 13/01/21. Source LCP Enact.

Offers from OCGTs not accepted 1700-1730 13/01/21. Source LCP Enact.

The problem is because in a utilisation payment only market the insurance value provided by fast responding assets isn’t valued if they aren’t called into action. NGESO are taking the right steps to remedy the issue by reintroducing STOR at Day Ahead (which allows providers to value in tight system conditions to their tenders) and investigating new methodologies like the BM Reserve Trial; both of which feature availability payments for those best able to provide Reserve. These are part of a wider package of Reserve reform which is fundamentally about the strategic shift to managing the system with predominantly duration limited resource, like batteries, instead of traditional notions of Headroom (and Footroom).

But this still doesn’t answer why the CCGTs were able to command such high prices. The answer is of course scarcity, cold days with little wind can create tight system conditions and we saw this reflected in the Day Ahead wholesale price at 1700-1800 (£1500/MWh on the N2EX). These CCGTs made the calculation they could earn more by not self-dispatching against the peak hour-long wholesale price and instead Grid would need to bring them online and pay high prices for an entire six hour run. It’s a gamble which doesn’t always pay off but did here, with a single unit at West Burton earning £3.78m from the BM that day compared to £0.53m if it had made the same run against wholesale.

No alt text provided for this image

West Burton Unit 3 profits in the BM compared to wholesale. Source BM Reports.

The wholesale market and Balancing Mechanism are linked of course by the imbalance price and it is the SIP calculation formula which drives these trading decisions between the wholesale market and the BM. The SIP didn’t clear at high levels on Wednesday afternoon because ultimately there wasn’t a shortage of energy so the large costs of balancing on this day, shown below, are put down as Reserve costs and so instead pass through via BSUOS. Interestingly SIP did reach £990 earlier at 1pm that day when similar actions were being taken but the system was short (also on the 8th Jan when it cleared at £400/MWh). Whether this is the SIP calculation working well, or a sign it is broken, however is for another post.

No alt text provided for this image

Balancing costs on week starting 11th Jan. Source NGESO.