What lockdown has taught us about the future electricity system
National Grid gave us a slot during its Summer Insight podcast series, to talk about the challenges of performing grid balancing during the COVID-19 lockdown. We began by assessing what impact lockdown has had on our own operations?
By Seb Blake, Head of Markets and Policy, Open Energi, and Robyn Lucas, Head of Data Science, Open Energi
We’ve had to undergo very little operational change as a result of lockdown. Our platform, Dynamic Demand 2 .0, is at its core, automated, which means that when everyone suddenly had to go home, all our technology continued to operate in exactly the same way as before. All our tech is also cloud-based, so the transition to working from home has been really straightforward.
Lockdown has also had very limited impact on the availability of our portfolio. A lot of the demand-side response (DSR) assets in our portfolio are in the waste water sector and clearly there hasn’t been much impact on the amount of waste water processing that needs to go on, so all of those pumps have kept performing in a way similar to what they did prior to lockdown, which has meant that our availability has been pretty consistent. The rest of our volume is from batteries and because batteries are exclusively flexibility assets, they haven’t been impacted at all.
Co-located assets lasting longer
The only real difference has been where we’ve seen a reduction in site demand where we have a co-located asset. This might be a battery sitting next to a factory with some solar on the roof, for example, where we perform some automated optimisation to reduce the reliance on grid imports at that site. What that’s actually meant is that we’ve been able to take those sites offline for longer, because the energy that we have stored in the battery prior to the peak can last a little bit longer because the site is drawing less power from it.
Other than that, it’s been a little more challenging to plan and carry out site visits, but all in all there’s not been a huge amount of impact on our operations.
It’s interesting to look at the contrast between a purpose-built flexibility asset like a battery or a gas peaker and DSR. The former clearly aren’t so susceptible to changes in their primary operation in the way that DSR is. That doesn’t mean we no longer see any future value in those variable asset types. We are still huge believers that demand-side assets like water pumps, or things with innate flexibility that heat, pump or cool, can make a very important contribution to grid balancing, particularly as newer asset types like electric vehicle (EV) charge points come in. These are assets that already exist on the system and there’s a huge amount of untapped potential.
A role for DSR?
What we have seen is some of the market environment shifting over the last few years to favour the front-of-the-meter (FoM) storage applications over DSR behind-the-meter, more variable assets. Things like the Targeted Charging Review (TCR), which has changed the revenue stack, and also some technical changes, which have tilted the balance slightly.
Batteries have effectively raised the technical standards of services like frequency response. Batteries are brilliant at responding really quickly and having large power changes. For demand-side assets like water pumps, which are, by their nature, a bit slower to respond, it’s become difficult to compete with battery assets.
It’s a shame because there’s so much potential, but they really just need to find their niche in the market. They clearly can have a role to play and it might not be these super-quick, super-precise services that batteries perform so well, but there should be a role for these flexible assets, operated in a smart, flexible, invisible manner, as our platform does.
So where does this leave DSR, considering that some of these changes are down to fundamental grid requirements like the need for faster response? Where does DSR fit in the flexibility landscape and the scope of services that National Grid requires to balance the system?
We don’t see DSR fitting into the very fast frequency response services that are coming through. Locational requirements make it much more difficult to aggregate up small assets and the fundamental need for a very quick, very precise response to signals – probably less than one second, for example – is a real challenge for demand-side assets.
BM Baselining tricky for DSR
It’s hard to see the Balancing Mechanism (BM) working well for DSR assets and that’s really down to baselining. With the BM you have to declare what you’re going to be doing an hour ahead of time and it’s quite difficult to predict accurately what a demand-side asset is going to be doing an hour ahead of time. That’s not to say there isn’t a huge amount of flexibility that you could expend at that kind of timescale; it’s all about committing to this active power level before gate closure and then delivering something against that baseline within the settlement period of delivery.
It’s notoriously difficult to define what a baseline is and the rules around this are quite grey, so while there is a huge amount of potential for despatching flexibility, within the timescales and the volumes required in the BM, this difficulty about baselining does make the BM particularly tricky for DSR.
Perhaps we need lower grade services with less stringent technical characteristics, which might pay less but be more suitable for assets like water pumps or EVs. Using existing flexibility rather than building out huge FoM systems is clearly a much more efficient use of resource (and with scalable technology platforms like Dynamic Demand 2.0 that can interact well with smaller smart devices like EV charge points, very achievable), but ultimately it is a question of market design, and aspects like moving the procurement of balancing services closer to real-time will have a big effect too.
COVID-19 and the flexibility markets
Lockdown has reduced demand significantly, so we can use it as a window to a future Grid with high renewables penetration. We’ve seen a much reduced level of transmission system demand so our renewables are making up a much larger proportion of the total demand on the grid and the response to it has been really interesting.
We’ve had this new Optional Downward Flexibility Management service (ODFM), which has been a way for National Grid to access an amount of negative reserve at really short notice, and the scale-up of this service has been really impressive – there’s something like 4GW accessible. ODFM is principally provided by wind and solar, so it’s a form of renewables curtailment, and I think this is something we are going to see a lot more of in a future energy system, and probably a lot more than people realise. It’s just not going to be efficient to build the amount of storage to capture all of the renewable excess and there are going to be times when the most economic thing to do is shut them down.
An interesting aspect of the ODFM is the timescale at which it’s despatched. It’s done a day ahead, as opposed to other balancing services, which are much more real-time events. This may be due in part to the technical characteristics of the assets – there are some renewables sites which require fairly manual shutdowns – but fundamentally there is no reason to think it should be. We’ve integrated with renewable sites through Dynamic Demand 2.0 and know that we can turn them off in a matter of seconds.
So, looking forward, it does raise some interesting questions around the notion of when National Grid should be taking balancing actions, given at day ahead doesn’t really seem to fit into our preconceived notions of the flow of the wholesale market. Normally the wholesale market should take care of 95% of balancing up to around an hour before real-time, at which point National Grid takes over, whereas with this ODFM service, it’s a day ahead.
A question of timing
So perhaps it’s more efficient for National Grid to act early if it does need to curtail some assets. That’s the interesting question here: do we need to preserve our traditional balancing timelines, where National Grid leaves the wholesale market as much time as possible before finally stepping in and rearranging to ensure reliability and security, or is it more efficient for National Grid to act early when it identifies a demand – for example, a particularly low demand weekend with a high amount of wind – and, therefore, it can lower the final cost to the consumer by responding quickly to that particular set of circumstances?
That’s something we’d like to see more discussion on moving forward. The Arenko trial in the BM gives an interesting foresight into one potential direction of travel here, whereby, the BM being a utilisation only service, National Grid, if it is going to wait to the last minute to call on providers to solve a problem, it does need to have confidence that people are going to be there and able to offer availability.
So potentially what we might see is a bit of an availability payment being introduced into the BM to ensure providers like a large battery is at right stage of charge level and able to offer in the right availability, so National Grid can get that confidence ahead of time that they’ll be able to manage an event when it does occur.
Impact on frequency
In terms of grid frequency, we continue to observe a general upward trend in the amount of volatility. What this means is the throughput required for a storage system to perform dynamic frequency regulation continues to rise year on year. For example, back in 2014, a one hour battery system would have performed 390 cycles over the course of a year, whereas this figure is now more like 490 – an increase of 25%. This means that the random walk that frequency does around 50Hz is getting longer, and this is due to a lower inertia system, as renewables have been increasing over this period and fossil fuel based thermal generation (which historically produces inertia) has been reducing.
Also, the number of times frequency has dipped below the 49.7Hz regulatory marker has actually increased fivefold in this period, so frequency is getting more and more volatile generally. So the procurement of real-time flexibility has never been more important in ensuring security of supply, particularly as we bring more and more renewables onto the system.
What’s really interesting is that we haven’t observed a huge change in grid frequency since mid-March, as compared to the same period last year, which, given that it’s been a much lower inertia system at times, is not what you might expect. We think this might be because, while the system demand has been lower and at times the amount of renewables on the system has been higher, Grid has been procuring the same volume of frequency response because the volume procured is driven by the biggest loss of load on the system, which hasn’t changed. So effectively you’ve had a higher proportion of frequency response for your same demand, which may have acted to increase the synthetic inertia and reduce that random walk of frequency.
Lessons for the future
So when we think forward, it’s clear that we might already have the tools to manage higher renewable penetrations. The question is whether we might be able to do it in smarter, more efficient ways, given we certainly have seen a fair amount of renewable generation curtailed off and thermal plant brought on. Can we bring in actual synchronous inertia as well as the synthetic use mentioned? But clearly, on the timescale of an instant economic lockdown, this is the best that can be done and it’s just about applying these learnings of future market designs.
We need to ensure a reliable, secure system, where we have these volumes of wind and solar, where we’re not having to bid them off, and we can continue to have them generating power whilst balancing the system at the same time. That’s where smart, distributed technology and platforms like ours that can operate these storage assets in an autonomous way are really important for the future of the electricity system.