31 Dec 2022
Flora Biggins

Signal methodology - including the Balancing Mechanism

The Balancing Mechanism has to date made up an insignificant proportion of battery revenues. In 2022, it accounted for <1% of revenues across the battery fleet. However, National Grid ESO have planned control room improvements (namely, a new control room system called the Open Balancing Platform due for release in September 2023). This will allow batteries to make a lot more money in this market in the future. As such, we now include Balancing Mechanism revenues in Signal - our 3-year battery revenue projection model.

This article outlines how we calculate Balancing Mechanism (BM) revenues in our Signal forecast. Check out this article for a more general overview of our recent changes to Signal.

Overview of modeling steps

To model revenues from batteries in the BM to include in our Signal forecast, we:

  1. Generate bid and offer price projections.
  2. Calculate potential revenues for batteries in the Balancing Mechanism, assuming all bids and offers are accepted.
  3. Model bid and offer acceptance rates in line with the system improvements due in the National Grid control room.
  4. Model Balancing Mechanism participation rate across the battery fleet.
  5. Combine prices and acceptance and participation rates to quantify revenues.

The remainder of this article expands on these steps in more detail.

1. Generating bid and offer price projections

Assets generate revenue in the Balancing Mechanism through the acceptance of bids or offers. In this process, assets submit a price (£/MWh) and quantity (MW) that they are willing to adjust their scheduled production by. If accepted, the ESO pays the asset in line with the price submitted.

This means that there is no single price associated with the cost of balancing. However, taking a volume-weighted average of the accepted bid and offer prices indicates the average price assets were paid.

In line with the industry norm, our calculation of the average price will exclude any actions taken to manage system constraints. This is so that the values would be reflective of balancing prices across the whole of Great Britain, irrespective of asset location.

Relationship with day-ahead prices

In order to forecast volume-weighted average bid and offer prices, we first examine how they have historically varied with day-ahead prices. This is shown in figure 1 (below).

Figure 1 - Relationship between day-ahead prices and volume-weighted average bid-offer acceptance prices. Note: outliers have been excluded from this plot for illustrative purposes. Data from January 2018 - October 2022.
  • Volume-weighted average bid and offer prices both have a strong positive correlation with day-ahead prices.
  • Day-ahead prices explain 86% of the variance of offer prices and 68% of the variance of offer prices (as measured by R² value).

Modeling bid-offer prices

Figure 2 (below) shows both day-ahead and volume-weighted average bid-offer acceptance prices across an illustrative 2-week period in 2022.

Figure 2: Illustrative period showing the daily relationship between day-ahead prices and volume-weighted average accepted bid and offer prices.
  • Offer prices track the daily shape of day-ahead prices but are consistently higher.
  • Bid prices are more static (showing less daily variability) than both offer prices and day-ahead prices. On closer examination, they correlate more closely with the minimum daily day-ahead price.

Therefore, we use a linear regression trained on historical day-ahead and Balancing Mechanism data to model bid and offer prices.

  • Offer price calculations use the day-ahead hourly prices, with an intercept value corresponding to the average uplift of offer prices.
  • Bid prices use a rolling minimum of day-ahead prices, resulting in lower intraday variability of pricing.

We use these two regression models and the hourly forward power curve within Signal (see for more details) to project the volume-weighted average bid and offer prices.

2. Modeling Balancing Mechanism revenues to include in Signal

Strategy assumption

Given the limited historical data available, we need to make some assumptions about how batteries will likely be operated in the BM. One strategy (which we saw play out during the summer 2022 heatwave) is to use the BM to ‘unwind’ or ‘buy back’ physical positions made in the wholesale markets. This strategy is illustrated below in figure 4.

Figure 4: Dispatch profiles for a trade made in the day-ahead market, subsequently unwound in the BM.
  1. Two trades are made to charge and discharge the battery in the wholesale market.
  2. A bid-offer pair is accepted in the BM, amending the scheduled dispatch profile of the asset. During the charge, the accepted offer effectively cancels the assets import. Similarly, for the discharge, the accepted bid effectively cancels the export.
  3. As a result, the state of charge for the site remains unchanged, and no physical cycling occurs. The original wholesale position is said to be ‘unwound’ or ‘bought back’.

Modeling unwinding/buy-back value

Using the above assumption for BM strategy and the forecasted bid and offer prices outlined in part 1, we can calculate the future value associated with unwinding a wholesale position in the Balancing Mechanism as follows:

  1. Use the hourly power curve to find the optimal battery dispatch in the wholesale markets.
  2. Calculate the corresponding value associated with buying back the positions in the balancing mechanism. This means finding the bid and offer prices that coincide with the discharge and charge positions, respectively.
  3. The difference between offer and bid prices at these times gives the unwinding/buy-back value.

This unwinding calculation is shown below in figure 5. It shows an example where power is bought in the day-ahead market in the morning (to charge up a battery) and sold in the evening (discharging the battery).

Figure 5: Unwinding spread due to selling and purchasing power in the BM (price data from 19/09/22 is used in this example).

Using the calculation outlined above, we can quantify the total value of unwinding wholesale arbitrage positions in the BM. This value represents the theoretical maximum value a battery could expect to earn if all positions were unwound. The theoretical maximum revenues over the forecast horizon are shown in figure 6 below.

Figure 6: Annualized theoretical maximum unwinding revenues for a 1h battery.

In reality, however, the above value cannot be entirely realized by batteries for two reasons:

  1. Acceptance rate - Not all bids/offers in the BM are accepted.
  2. Participation rate - this strategy is only available to assets pursuing a wholesale strategy in the first place.

These factors are accounted for in the following modeling steps.

3. Modelling bid and offer acceptance rates

Battery acceptance in the BM is a hotly discussed topic. So much so that in December 2022, National Grid ESO held an event at the Wokingham control room to discuss the matter in more detail. One key idea that was discussed was skip rate: a measure of how frequently bids and offers are rejected in favor of other (sometimes more expensive) units.

At this event, the ESO announced a new IT system called the ‘Open Balancing Platform’, which is set for release in September 2023, with a full rollout in 2027. This new system is intended to improve the simultaneous dispatching of hundreds of assets by enhancing visibility and optimization decisions.

With this new system, we expect the acceptance rate of batteries in the BM to increase, reflected below in figure 7.

Figure 7: Indicative acceptance rates (%) for batteries in the Balancing Mechanism, with the introduction of the Open Balancing Platform.
  • We expect that until September 2023, bid and offer acceptance rates will be in line with historical rates. That is, Balancing Mechanism acceptances (and consequently revenues) will be minimal.
  • With the introduction of the Open Balancing Platform in September 2023, we would expect a step-change in bid and offer acceptance rates for batteries, with a gradual increase until implementation is complete in 2027.

4. Merchant participation rate

The second driver of low BM activity for storage has been its relatively low value compared to other revenue opportunities. Historically, large frequency response requirements relative to installed battery capacity have led to undersupply and high prices. As response markets saturate and prices fall, more and more assets will move toward merchant strategies. These assets will be increasingly likely to participate in the BM.

As such, we have modeled the percentage of the GB battery fleet pursuing merchant strategies in the wholesale markets and the Balancing Mechanism. We calculate this based on our expectations for future response requirements in addition to future battery buildout. An example market participation rate is shown below in figure 8.

Figure 8: Example merchant strategy participation rate. Calculated as the remaining battery capacity without response contracts as a percentage of the entire fleet.
  • Participation rates exhibit seasonality in line with higher requirements for Dynamic Containment Low during the summer months.
  • Across the time horizon, wholesale participation increases in line with increased deployment of BESS relative to response requirements.

5. Bringing it all together in Signal

We arrive at our Balancing Mechanism revenue projections for Signal by combining the results of the three previous modeling steps: potential unwinding revenues, acceptance rates, and participation rates. Multiplying these together gives our forecast for the fleet-wide average BM revenues, shown below in figure 9.

Figure 9: Average battery fleet Balancing Mechanism revenue projections for Signal’s p50 scenario, assuming 1h battery duration.
  • Prior to September 2023, we expect acceptance rates in the Balancing Mechanism to be in line with historical rates - with negligible revenues across the fleet.
  • We expect average battery Balancing Mechanism revenues to increase from September 2023 onwards, with the introduction of National Grid ESO’s Open Balancing Platform.
  • By 2025 our Signal forecast projects annualized revenues in the Balancing Mechanism will increase to £10k/MW for 1h duration batteries.
  • We also anticipate some slight seasonality in these revenues (higher in winter, lower in summer) as day-ahead prices are seasonal, and bid/offer prices correlate with day-ahead prices.

Copyright© 2024 Modo Energy. All rights reserved