Thought piece by Will Stephenson, Research Associate for Great Britain Flexible Energy Markets
On 2nd November, National Grid ESO (NGESO) first procured ancillary services via its new auction platform. Much has already been said about this Enduring Auction Capability (EAC), particularly regarding the three key changes it brought in:
- Negative pricing: in the design of the EAC, NGESO removed the existing price floor of £0/MW/h; instead, there is now a minimum price limit of £-999.99/MW/h and a maximum limit of £999.99/MW/h. This has opened the door to negative prices in High services, where providers would thereby be paying to charge.
- Service Splitting: following the move to EAC, it is now possible to split between different Response products in the same or opposite directions. Legal bids could include pairing dynamic containment low and dynamic moderation low, or dynamic containment low and dynamic regulation high.
- Submission of multiple orders: under the new system, sellers can submit many different orders and NGESO must select between these via co-optimisation. Sellers can submit mutually exclusive or substitutable orders and, on the buy side, NGESO also submits substitutable orders. These orders are then co-optimised; the EAC algorithm attempts to select the set of orders that, taken together, maximise welfare.
The immediately obvious impact of EAC was seen in how prices for both DM (dynamic moderation) High and DR (dynamic regulation) High took no time in falling below £0/MW/h (as shown in Figure 1). In effect, this heralds the end of a period of artificial bonus value for battery storage providers of the Dynamic suite of frequency response services.
However, perhaps the most interesting trend has been the increasing complexity of orders being made by suppliers.
As suppliers can split bids between different services, they are likely to make more bids, with the capacity of each being smaller. This is because they can combine different Dynamic suite service options, allowing them to exploit multiple profit opportunities at a time. We also see this in the data, with average bid volumes falling. Meanwhile, the number of bids being made since the introduction of EAC has exploded, increasing by 359% across the six services. These trends are shown in Figures 2 and 3.
Complexity of bidding
Who did it well?
With negative pricing becoming a part of service procurement, revenue ceases to be a useful metric with which to define success. It is debatable as to whether it ever was, given the way in which these services generally supplement a wider optimisation strategy that includes energy trading as the main component; frequency response services, particularly DR High, are as much a means to an end as the end itself.
Therefore, in this analysis, acceptance rate is instead used as the metric for success. If a unit bids in at a particular price, they are presumably happy to be accepted for it. We have therefore pulled out some trends regarding which units were successful in the first month of EAC in Table 1 and Figure 4 (with a threshold for qualification of 1,000MW of executed capacity).
This suggests that the dominant strategy for units with the highest acceptance rates is to specialise in the DC (dynamic containment) services, with 6 of the 7 firms exclusively making DC orders. There is an obvious reason for this: DC is by far the biggest service. However, we also see that it is possible for a unit to pursue a complex strategy, as seen by VEST-4, with its 99.45% acceptance rate, despite DM representing 67% of its executed volume.
Since DC dominates the strategy of the most successful units, the next logical step is to examine each service’s average acceptance rates to identify whether DC Low and DC High are the markets with the overall highest acceptance rates (Table 2).
The results are surprising. While DC High and DC Low have relatively high average acceptance rates compared to other services, DR High has the highest. Both DC services have a high level of standard deviation, which implies that units either have extremely high or low acceptance rates for DC services.
Due to the discrepancies in statistics for each Dynamic Suite service, it is important to obtain a more granular view of the distributions. In Figure 5, we show what proportion of each service has had at least a certain proportion of units accepted. For example, we show that 35% of units have an acceptance rate of at least ~60% for DC High, and that there is a large cluster of units with a 7-9% acceptance rate for DR Low.
These results carry information about the distribution of the acceptance rate of each service: steeper portions of each curve indicate that a service has a high proportion of a specific acceptance rate and every service’s curve becomes shallower at the higher acceptance rates, as it is less likely to have a high portion of units achieving this.
What is National Grid ESO looking for?
To examine what services NGESO is prioritising, we need to find the total volume it orders for each service (which can be thought of as the maximum volume it desires), and its maximum willingness to pay. It is possible to find the maximum that NGESO is willing to pay via its Buy Orders, in which it is willing to pay higher amounts for its first ‘x’ MWs of capacity, to ensure availability of suppliers. Finding a daily average for the maximum willingness to pay and an EFA Block average for the maximum volume desired, we see that results vary significantly across services
These numbers confirm something that we all know to be true; NGESO is looking to procure far greater volumes of DC than of DR or DM, and it is willing to pay some eye-watering prices for it. This suggests that grid inertia, the alternative to DC in terms of arresting the rate of change of frequency (RoCoF) following a potential loss from the system, is expensive.
As shown in the table above, NGESO’s willingness to pay for DM is significantly lower than for both DC and DR. With NGESO’s reduced present willingness to pay for DM, there is likely to also be a lower acceptance rate from the EAC algorithm for a given bid price, which we see in the results. However, this has the potential to change in the future, depending on the share of renewables in power generation. Dynamic Moderation is designed as a pre-fault service to deal with large frequency fluctuations. As we progress towards a generation mix dominated by renewable energy sources, lower inertia will result in more significant frequency deviations, potentially increasing future demand for DM.
What does the future hold?
Upcoming Services
Quick Reserve (QR) and Slow Reserve (SR) are also coming to EAC; they will be replacing Fast Reserve and Short Term Operating Reserve (STOR), respectively, beginning in Summer 2024. These two new markets will not always clear simultaneously; Quick Reserve consists of 12x 2-hour blocks, whereas Slow Reserve consists of one 8-hour block from 23:00 to 07:00, followed by 8x 2-hour blocks. While splitting between Frequency Response and Reserve services will not be allowed (for example, it would not be possible to split between DC and QR), suppliers will be able to split between the opposite directions of the same service e.g., Positive Quick Reserve and Negative Quick Reserve. Therefore, at least at first, Reserve services are unlikely to see the same level of complexity as Response services.
In addition, Balancing Reserve (BR), from Spring 2024, will also be conducting auctions using the same auction algorithm as the EAC, but independently run. Here, the ESO will again submit its Buy Orders, while suppliers will submit their availability prices; the auction will accept a cohort of Sell orders and Buy orders such that welfare is maximised and determine clearing prices that minimise the cost of provision.
BR will allow for significant added complexity in the market: it can be split/stacked with Capacity Market contracts, Bid Offer Acceptances (BOAs) in the BM, and real-time ancillary services such as the Obligatory Reactive Power Service (ORPS) or Mandatory Frequency Response (MFR). It will not initially be stackable with the Dynamic Suite, due to performance monitoring methodologies not being coherent enough to avoid penalising stacking, however this may change in the future. Therefore, there is potential for further future sophistication of bidding strategies from market participants.
Implications for the power system and battery operation going forwards
The EAC has shown how greater market automation has a positive impact on market liquidity (shown by higher bid counts). This is because it has encouraged potential providers to submit more bids into the market, by removing the exclusivity between DC, DM, and DR. This in turn helps to achieve more efficient outcomes, as NGESO can fulfil its dynamic requirements for various services from a larger, more competitive pool of bids. Additionally, the introduction of negative pricing prevents providers from artificially gaining inframarginal rent, particularly in the DR High service where the “free” charging offered will tend to make negative pricing sensible economically speaking.
We expect to see this trend continue, meaning that the main profit-making opportunities will move from capitalising on market inefficiencies towards using more complex strategies that increase profitability across not just the Dynamic Suite of frequency response services, but also Reserve (Balancing, Quick, and maybe even Slow), and the Balancing Mechanism.
Speaking of complexity and the Balancing Mechanism, the Open Balancing Platform, launched on 12th December 2023, automates the selection of actions taken in the Battery Storage and Small BMU zones of the Balancing Control Room. In place of manual acceptance of Bids and Offers from these BMUs, acceptance will be algorithmically decided, raising the potential frequency of Bid/Offer acceptance enormously.
Ultimately, perhaps the main beneficiary of these developments will be consumers, via lower costs. A world in which essential grid services are procured more efficiently is one in which fewer costs need to be levied on end consumers, which must be good news. That doesn’t mean it’s bad news for the batteries, though. As we’ve shown, there is significant opportunity if you optimise well. With an increasing number of services available to providers, for which NGESO is willing to pay handsomely, and more automation than ever in presently disappointing areas like the Balancing Mechanism, there is much for battery storage owners and optimisers to be excited about.
