Nodal pricing - also known as Locational Marginal Pricing (or LMP) - is a way of determining the price of electricity that varies locationally. In Part One, we looked at the implications of nodal pricing for the whole energy system in Great Britain. Then, in Part Two, we explored what it might mean for battery energy storage.
For Part Three, we’ve modelled what a battery energy storage system might do across the course of a single day, at a simulated node. In this scenario, we’ve imagined the node as a single Grid Supply Point (GSP). As you can see in figure 1 (below), there are 362 of these across Great Britain.
Figure 2 (below) shows the modelled make-up of embedded generation in each of the GSPs by winter 2025/26. This data is taken from NG ESO’s Future Energy Scenarios 2021 (specifically, the Leading the Way scenario). The projected average demand is overlaid (the purple line).
It may be helpful to view this piece not as a forecast, but rather as an addendum to Parts One and Two. Taken together, all three articles aim to paint a holistic picture of how battery energy storage may operate under a nodal pricing system.
Spoiler alert
In this article, we show how a battery might be optimised in a nodal pricing system. A specific node has unique characteristics which lead to certain price incentives. We then have the ability to layer on nationwide ancillary services to increase the revenue stack. This begins to look quite similar to a behind-the-meter (or co-located) optimisation - in which a site's particular characteristics can lead to significant differences in a storage asset’s optimal dispatch.
What will these GSPs look like?
Figure 3 (below) shows the modelled capacity at five individual Grid Supply Points. As above, the projected average demand is overlaid in purple. Each GSP has a different make-up of embedded capacity, demand, and flexibility (defined here as storage). We’ve picked these five as they are particularly interesting GSPs - one has huge demand, one has lots of wind, one has loads of embedded solar, one has lots of baseload capacity, and one has a nice mix.
These individual GSP models can give us a good idea about what batteries might do, based on their location. (In this piece, we will focus on BOLN_1, MITY_1, and particularly CREB_1. We have shown the others to give a picture of the types of make-up we will see at these GSPs.)
For example, the GSP at Minety in Wiltshire (MITY_1) - which currently has a 100 MW operational battery installed, as well as another 49.9 MW battery energy storage site that has recently come online* - is projected to have nearly 700 MW of solar by 2025/26. This will outstrip the fairly low demand there during the sunniest times of the day. Therefore, a battery here could charge up using the excess cheap solar power in the middle of the day, discharge during the peak when demand is likely to outstrip supply, and provide ancillary services for the rest of the time.
* The graph above models 131 MW of storage in the Minety GSP by 2025/26. Keen mathematicians will note that it already houses more than 131 MW of storage capacity. This is because we have used the 2021 FES models for the graphs in this article, and they are just that: models, with scope for error.
On the other hand, the GSP at Bolney in West Sussex (BOLN_1) is projected to have high demand with little embedded generation. It will therefore be a net-importing GSP. Storage here is likely to derive its revenues from ancillary services. It could also capture arbitrage opportunities from the import price at the node.
What does our node look like?
For our modelled case study, we have looked at what a battery might do at a GSP node across a single day in November 2025. Figure 4 (below) shows what this day could look like. This simulated node is based on Creyke Beck (CREB_1), although it does differ slightly. We chose to base our node on Creyke Beck due to its mixed generation make-up.
- Solar generation is quite small, with a load factor of 5% (due to it being winter). We have some peak ‘other’ generation coming from biogas-fed plants, and it’s pretty windy.
- Residual demand on the node over most of the day is negative. This means the node is net-exporting, with some import overnight.
- As a result, the modelled price is high in the periods when power is imported, and low during periods when embedded renewable generation is exceeding demand.
Figure 5 (below) shows the same node, but this time the import and export connection can only take 200 MVA.
- Residual demand exceeds the export limit at some times of the day (between 10:00 and 15:00).
- During these periods, consumers within the node are motivated to use more power. Rather than curtail generation, it could be cheaper to pay users to use power.
- This is how we arrive at a ‘nodal price’, which spikes when power consumption is forecast to approach the export or import limit of the node.
What does our battery energy storage system do?
We’ve modelled a 50 MW / 100 MWh battery energy storage system, with a round trip efficiency of 85%. We allow the system to do 1.75 cycles on this day. Figure 6 (below) shows our battery charge and discharge schedule, given the nodal price in our example node.
(At the time of writing, there is a single 49.9 MW / 50 MWh operational battery energy storage asset at GSP CREB_1, or Creyke Beck.)
- When the price spikes overnight, the battery discharges.
- During the middle of the day, when the export constraint would be breached, the battery charges (and gets paid to do so). As a result, the export limit at the node is not exceeded, so no curtailment is necessary.
- The battery then discharges during the higher price periods later in the day.
This optimisation is based purely on the modelled nodal price: ie. the charge and discharge schedule above assumes our battery is only pursuing arbitrage opportunities around the nodal price. This schedule results in a long period where the system has very little charge. If other factors were also considered, such as a requirement on the minimum state of energy, or additional revenues associated with having availability for ancillary services, the optimisation would change to have less time spent at 0% state of charge (but the arbitrage revenues would decrease)!
What does our battery energy storage system do in frequency response?
Availability for ancillary services
So how would this schedule affect our asset’s availability to provide ancillary services? Figure 7 (below) shows this.
In theory, a 50 MW battery doing nothing at an unconstrained site would have 50 MW availability to charge, and 50 MW to discharge. At our node, the availability is limited - by the residual demand, the import and export constraints, and the energy available in the battery (e.g. when the battery is empty, after the night-time discharge, it cannot discharge any further).
It is also dependent on the power output of the battery. If the battery is charging at 50 MW, it can stop charging and start discharging. In this situation, it can discharge 100 MW, but has 0 MW availability to charge (as it is already charging at full power).
By moderating its scheduled power output (based on the nodal price optimisation), the battery could provide ancillary services in nationwide markets, as long as those ancillary services don't normally require a high power output (i.e. if they are 'low utilisation'). This means that participating in those ancillary services won't cause the energy of the system to deviate too much from its planned schedule (for most of the time). This way, the battery can still provide the flexibility the node requires, while stacking with another service - and can earn revenues for doing so.
What do clearing prices look like?
Figure 8 (below) shows nationwide frequency response clearing prices on our modelled day.
- Prices average £10/MW/h for the high-frequency service, and £8/MW/h for the low-frequency service.
- Since this service is low utilisation (similar to Dynamic Containment), our battery is able to tender into the service on top of its wholesale trading activities.
Figure 9 (below) shows our battery’s ancillary service availability with those clearing prices overlaid. This availability is dependent on the scheduled charge and discharge power, the energy of the system, and the capacity at the node for importing and exporting power (once the residual demand is factored in).
- When the battery is fully discharging, around midnight, the availability to discharge further is zero.
- There is only one settlement period (30, 0r 13:30-14:00) in which our battery has zero availability for charge or discharge. This is because it is fully charging, from empty - so cannot charge anymore, and also cannot discharge as it has no stored energy.
We assume that this ancillary service is procured by EFA block, instead of by settlement period (in the same way the auctions are designed now). We must be able to provide the volume we sell for the whole of the EFA block. Therefore, we have to take the minimum availability per EFA block as the volume we can tender into the service, both for the high (charging) and low (discharging) services. In the scenario in figure 9 (above), this means that the battery is unable to tender any frequency response during EFA 1 and EFA 4 - as at some points within these EFA blocks both the high and low availabilities are zero.
How would settlement period procurement impact this?
These limits on how much and how often our battery can provide frequency response are due to the market being procured by EFA block. As stated above, if there is any point during a 4-hour block in which a battery has no availability, it cannot provide the service at all during those four hours. If this sounds inefficient, that’s because it is.
If the market were set up differently - so that frequency response was secured by half-hour settlement period, instead of by 4-hour EFA block - what could our battery do? Figure 10 (below) shows how much frequency response our battery could provide during our simulated day if the service was procured by settlement period.
- On this day, settlement period procurement would mean our battery being able to provide 120% more volume than it can with EFA block procurement.
If a battery were able to change its tender on a settlement period basis, it would be able to sell much more of its available flexibility. In theory, moving to settlement period procurement would also benefit the ESO. This ability to procure the volume needed on a more granular basis would allow for more accurate projections of capacity. They’re not having to procure for a rainy day, so to speak, as they can adjust amounts nearer to real-time. On top of this, the ability of more assets to tender their available flexibility - i.e. more potential market participants - could drive down prices (as there’s increased competition). However, it would likely lead to an overall decrease in the amount of capacity procured - meaning a smaller market for battery energy storage to play in.
So how much money does our battery energy storage system actually make?
So now we’ve seen the optimisation strategy of our battery, let’s examine the revenue stack. Figure 11 (below) shows the modelled revenues of our battery energy storage system across our November day at our imagined node. There are big spikes from arbitrage revenues during certain EFA blocks, with ancillary services providing steady income during other times of the day.
Of course, we need to take the above with a massive pinch of salt. These prices aren’t necessarily accurate representations of the opportunities that may arise under nodal pricing. As a benchmark, two-hour battery energy storage systems made an average of £311.40/MWh/day in November 2021 when we had incredibly high wholesale volatility. Under nodal pricing, we would expect this to be much higher in some nodes, and much lower in others.
Conclusion
Having the right data available, in a common and timely format, will be key to making the most of nodal flexibility. While the job of the optimiser (and anyone else who is trying to model the opportunities across different nodes) could become far more complicated, it could also be more interesting. Ultimately, the system overall should benefit from cheaper balancing costs!