It’s a cold December afternoon in 2019, and the Modo team (all 4.5 of us) around huddled around a tiny desk in a tiny office. We've just come off a call with another investor that wants to build battery energy storage assets but can't get comfortable with the revenue streams.
The industry is awash with buzzwords like ‘neural networks’, ‘trading algorithms’, ‘auto optimisation’. Everyone is very excited about the potential of energy storage operating in the BM and merchant markets (and rightly so, too!). Battery business cases are full of assumptions, but nobody knows whether they're realistic.
Investors are struggling to make sense of it all and keep circling back on the same questions:
- How much money are batteries really making?
- Which assets are performing better than others?
- What is the range of potential returns?
- What does my upside business case and my downside case look like?
So, we decided to find out. Enter the Leaderboard.
Leaderboard V1
Key features:
- Balancing Mechanism only.
- A limited number of assets.
- It runs on Alex's laptop (sometimes).
In December 2019, we started to build a picture of how assets were performing in the Balancing Mechanism. To begin with, we looked at all BMUs and ranked them on activity - bid/offer acceptances (BOAs). We were largely focusing on which operators were most active, and most utilised by National Grid ESO. (Back then, the merit order wasn't quite as, well, ordered as it is now.) We ranked all the BMUs against each other to see what the field looked like.
We created our first ‘Modo Metric’, which we called Percentage of Perfect (PoP). We then used PoP to see which assets were being dispatched at prices closest to the marginal price (i.e. the most expensive offers and least expensive bids) - in simple terms, a high PoP was a good thing.
The first iteration of the Leaderboard was rough-and-ready, completely unscalable, running on a local machine, scripted solely in Python, and visualised on Plotly (an open-source plotting library).
Note: We also took words that ended with the letters ‘cs’ (like ‘Metrics’) and replaced the ‘cs’ with an ‘x’ (so that ‘Metrics’ became ‘Metrix’). For this, we can only apologise.
Leaderboard V2
Key features:
- New markets added - firm frequency response (FFR) and enhance frequency response (EFR).
- PDF report distributed to customers via e-mail.
- Asset IDs mapped to asset names.
- Ranking of Top 20 assets.
We showed V1 of the Leaderboard to a couple of potential customers who liked the idea, and said that if we built it, they would come. Awesome! Our next step was to create something that they would actually pay for (a very different challenge). We decided to launch a ‘Monthly Leaderboard Report’ and flesh it out to include more detailed analysis - including new markets, more assets, more accurate revenues. V2 pushed the Leaderboard in two opposite ways - it become both more complex, and at the same time simpler.
More complex: more markets, more assets
The revenue figures in the Leaderboard (and their supporting calculations) became more advanced as we added in ancillary services (weekly FFR; monthly FFR; EFR). We normalised revenues on a £/MW basis, so that we could compare apples with apples. After a lot of requests from (at this stage, potential) customers, we also decided to include annualised revenues. Note - this split the Modo team between the purists and the pragmatists. Should we do what our customers want and provide annualised figures (knowing they are sometimes misquoted to over-estimate earnings)? In the end, we decided that all we could do was provide the information we had in the most transparent way possible, and that what people chose to do with that information was entirely their choice.
In V2, we made the Leaderboard much more readable too. For example, we mapped market IDs to physical assets, so that you could see tangible, easily understandable asset names such as ‘Rock Farm’, rather than abstract IDs like ‘ANG_334’.
Simpler: a printable PDF report distributed via e-mail
We decided to strip down the interface to the rawest acceptable level for our early customers. At this stage, we were still testing our hypothesis, deciding whether companies would pay to see a Leaderboard of battery assets in GB.
We copied the outputs of our code, Leaderboard.py, into Microsoft Word and distributed it via e-mail on a PDF report - old-school. The Leaderboard wasn't interactive, but this turned out to be a good thing - we were forced to fit everything on one A4 page and focus on what mattered.
We also had to set some rules about what the Leaderboard should be, and what it should become - should we normalise on £/MW or not? Yes. Would we include metrics like PoP? No. Should we normalise based on an asset's nameplate power rating or on the size of its maximum frequency response contract? All were critical questions in our pursuit of the perfect Leaderboard.
Leaderboard V3
Key features:
- We moved the Leaderboard to our online data platform.
- Interactive - toggle long term contracts and choose whether to normalise revenues (£/MW).
- Wholesale (traded) markets added.
- Energy throughput (MWh) estimations added.
It's Q3 2020, and we've been distributing the Leaderboard to our (four!) paying customers for six months via e-mailed PDFs. We're supplementing the reports with tons of analysis, commentary, and sub-leaderboards for each market. The Monthly Modo Report has grown from a 5-page summary to a 21-page behemoth full of content. The reports won't fit on regular emails because they're bigger than 10MB, so we need to compress them. You get the idea. And, our customers are asking for more - more data, more assets, more historical data. We’ve reached the limits of PDF reports. we built a user interface.
We moved the Leaderboard to our online platform and added filters, date pickers, toggles, and more.
At this stage, batteries were beginning to operate in merchant markets (day-ahead and intraday). We made our first stab at estimating revenues from these markets and added those to the Leaderboard. Our first go at this wasn't perfect, but it got us thinking about what we could do with data from the power exchanges.
We launched Leaderboard V3 on the platform in Autumn 2020.
Leaderboard V4
Key features:
- Fully interactive with all revenue streams and use-of-system charges.
- Increased accuracy of merchant market revenues.
- Frequency response energy throughput calculations.
- Asset tracking and performance benchmarking.
- State-of-charge management costs.
- Improvements in round-trip-efficiency estimations.
- Site connection (fixed DUoS) and use-of-system costs (BSUoS, DUoS).
We launched V4 of the Leaderboard in Summer 2021, which included updates across the board - significant improvements in our calculations of revenues, battery energy throughputs, asset mapping, and a substantial expansion of our user interface (light mode, anyone?).
After collecting two years of performance data, we were able to build in functionality to compare asset revenues and the Modo average. We also built screens for detailed asset information and ranked owners and optimisers on market share.
We also made massive server-side upgrades to the granularity of data that we analyse (e.g. moving from 15-second frequency data to second-by-second frequency data) to estimate energy throughputs. These improvements enabled us to report broader BESS costs such as BSUoS and DUoS, which significantly impact earnings. We also improved our methodology for wholesale market revenues, particularly for assets that choose to opt-out of ancillary services (e.g. Dynamic Containment) during periods of high prices in the day-ahead and intraday markets. All of this has brought us closer to our goal of creating the perfect Leaderboard.
The future of the Modo Leaderboard 🚀
We have so much more that we're going to do to support our customers as they build, own and operate energy storage assets in GB. In the future, we've got plans for deeper analysis on assets' physical and commercial performance, alongside the performance of trading entities such as optimisers, technology providers and utilities.
We're so privileged to have the support of the GB energy storage community as we experiment, build and grow. Wherever the Leaderboard takes us next, we'll never forget everything we've been through to get here.