In January, we started modeling asset-specific revenues for batteries from wholesale trading. Since then, we have updated our asset-specific revenues to include Dynamic Containment. This methodology update explains how we modeled the following:
- Dynamic Containment Low and High prices at an EFA block granularity.
- Asset-specific revenues in these markets.
You can check out the results of our updated asset-specific revenues, including Dynamic Containment revenues, here.
Modeling Dynamic Containment prices
Figure 1 below shows the rolling 100-day correlation between Dynamic Containment Low prices and the daily wholesale spread.
- In August 2022, we saw a weak correlation between mean daily Dynamic Containment Low prices and daily wholesale spread. Since then, the correlation has increased to 75% for the 100-day rolling period ending 31st January.
- The trend of high correlation between Dynamic Containment prices and wholesale spreads reflects the increasing saturation in Dynamic Containment. The auction has become more efficient with operators pricing in opportunity costs.
Dynamic Containment saturation will continue, especially with expectations for battery buildout and National Grid ESO reducing the volume requirements over the summer months in 2023. Therefore, we have chosen the wholesale opportunity cost (using our hourly forward power curve) as the basis of our Dynamic Containment price model.
Modeling steps
The steps we used to model Dynamic Containment prices on an EFA block granularity are outlined below. We use Dynamic Containment Low prices from November 2022 as an example to illustrate the performance in more detail. We show performance over the full back-tested period for Low and High prices afterward.
1. Wholesale opportunity value of each EFA
Firstly, we identify the maximum potential wholesale revenue for each EFA block. We do this by finding the best wholesale spread (maximum - minimum price) for each pair of EFA blocks daily. We attribute half of the spread for each EFA block in a pair. An example is shown below (Figure 2) using N2EX price data for 1st November 2022.
- On this day, the greatest wholesale spread (£74/MWh) is achieved by charging in EFA block 2 and discharging in EFA block 5.
- We then find the average potential wholesale revenue in each EFA block. We refer to this as the wholesale opportunity value. It quantifies the average net revenue achievable by trading in this block and another EFA block (including itself).
Figure 3 (below) shows the wholesale opportunity value for each EFA block and each day in November 2022, compared against the Dynamic Containment Low clearing prices.
- We see strong correlation between the EFA wholesale value and the Dynamic Containment Low clearing price.
- The high clearing prices at the end of November are reflected in the wholesale opportunity value.
- Dynamic Containment Low prices in EFA block 5 are usually much higher than the rest of the day, and the wholesale opportunity value consistently underestimates this.
We apply the observed trends in the Dynamic Containment Low pricing per EFA block to this wholesale opportunity value to better predict EFA block clearing prices.
2. Adjust wholesale opportunity values
We adjust the wholesale opportunity values in line with two historical factors:
- Clearing prices in the Low and High services: The sum of Dynamic Containment Low and High clearing prices should equal the wholesale opportunity value. Therefore, we must divide the wholesale opportunity value between the Low and High services, historically representing 70% and 30% of the total (Low + High) clearing price.
- EFA block clearing price ratios: This adjustment increases prices in EFA blocks that have historically had higher prices (e.g. EFA 5) and decreases prices in EFA blocks that have historically had lower prices (e.g. EFA 3 and 4), while keeping the sum of prices constant. We do this separately for the High and Low Dynamic Containment services because they typically have different profiles. The High service has higher prices in EFA 2, and the Low service in EFA 5.
Finally, we divide the adjusted wholesale opportunity value by 4 to convert to an hourly clearing price - since each EFA block lasts 4 hours.
Model performance
Figure 4 compares back-tested Dynamic Containment revenues (using steps 1 and 2 to model prices) against actual revenues (using actual prices).
- Back-tested Dynamic Containment revenues are generally consistent with actual revenues, especially for Dynamic Containment High.
- The notable exception is in December 2022 for Dynamic Containment Low, where the model significantly overestimates revenues.
- This may be due to less manual bidding in December, with people taking time off for the holidays. Additionally, December saw high wholesale spreads, which could have caused errors as the model is based on wholesale opportunity value.
Figure 5 (below) shows the daily wholesale spread plotted against the absolute daily error in modeling Dynamic Containment Low revenue.
- The Dynamic Containment Low model performs better when wholesale spreads are lower. We can see that larger errors occur as spreads increase.
- This explains why the model overestimates Dynamic Containment Low revenues in December 2022.
We use the model (described by steps 1 and 2 above) with our hourly forward power curve to produce forward Dynamic Containment prices, which we use to determine revenues in this market.
Dynamic Containment revenues
We calculate asset-specific Dynamic Containment revenues using our forward prices and the EFA blocks when our asset provides this service. We assume symmetrical delivery of Dynamic Containment (e.g., the battery provides both Low and High services) and a de-rating factor of 0.9. The revenue is therefore equal to:
Revenue = (DCL price + DCH price) x battery-rated power x 0.9 de-rating factor
This is calculated for each hour of Dynamic Containment delivery. Outside of these hours, we optimize the asset’s wholesale revenues while ensuring Dynamic Containment delivery requirements are met. For more detail on our dispatch model and assumptions, look at our help center article.
Our method is configurable for different battery parameters and different assumptions (battery duration, cycling constraints and Dynamic Containment EFAs). If there’s any particular scenario that you’d like us to model, drop us a line - we’d love to chat!