Data Visualization Art: Turning Data into Great Visuals
CfD Market Monitoring Dashboard
By Sophia Adler
During my time in the WMS Cost Stack Team, a data-driven team utilizing multiple data sources to improve non-energy cost (NEC) forecasts, I focused on the Contracts for Difference (CfD) cost. The CfD scheme, which promotes low-carbon electricity investment, offers renewable energy generators a fixed price for electricity production via an auction process, ensuring revenue through 15-year contracts with the Low Carbon Contracts Company (LCCC). Suppliers, like EDF, are responsible for covering the difference when the market price falls below the fixed price, making it crucial to track CfD generators closely.
My task was to create an interactive dashboard to monitor and predict the behavior of CfD generators entering the scheme. Since these generators have the option to delay the start date of their contracts, it is essential to account for potential delays that could lead to cost uncertainties, thereby affecting supplier’s price forecasts. Leveraging data visualization, this project aims to track CfD units expected to join the scheme within the next three years, refining start date predictions and adjusting our CfD cost forecasts accordingly. This approach fosters faster decision-making, deeper insights, and a data-driven culture.
Defining The Steps for Successful Data Visualisation
Three key steps were defined to navigate this data visualization project, as illustrated in the diagram below.
Stage 1
The primary goal of this project was to monitor the progress of upcoming CfD generators and predict the likelihood of them altering their contracted start dates, which would subsequently impact our cost forecasts. The initial step involved identifying the factors that could influence generators to change their contracted start dates. Key influencers identified were; generated metered volume, wholesale electricity prices, and outages. This identification guided the third step data collection process, which in turn defined the depth of our analysis. The methods and visuals used to display this information will be detailed further below.
Stage 2
Understanding the audience profile was crucial in selecting and presenting data effectively for the third data visualization project step. The primary audience, the Cost Stack team, is highly data-literate and familiar with various data sources and factors affecting the start date of a CfD generator. Given the team’s knowledge of the energy industry, tailoring visualisation to include detailed technical information aided informed decision making, enabling the transformation of intricate datasets into actionable insights.
Stage 3
The third step involved data collection, preparation, and visualization. Multiple external and internal data sources were collated, including the internal EDF market data database and the Low Carbon Contracts Company (LCCC), which provides a regularly updated CfD register. A meaningful relationship was formed between these two data sources, via manually mapping the BMU ID (balancing mechanism unit ID) to the CfD units in question. This information was sourced from Neta Reports website. Since our internal database has metrics for BMU IDs we could now extract data specifically for these CfD generators. The diverse data was meticulously cleaned, processed, and formatted enabling the transformation of raw data into visually engaging formats for enhanced decision-making. Other meaningful relationships could be determined now our CfD units had mapped BMU IDs, for example determining outage information from the Elexon website.
Despite the complexity of the data, visualizations were kept simple and readable to facilitate the use of metrics to provide evidence for the project’s defined purpose. This approach transforms intricate datasets into actionable insights. The technology used was minimal, utilizing easy-to-code graphs with Python and SQL and displayed using Streamlit.
The CfD Market Monitoring Dashboard
The home page displays a simplified version of the LCCC CfD Register displaying only the upcoming CfD units within the next 3 years and filters the data to display only immediate useful information such as the technology type. Interactivity was an important thing to consider when designing the dashboard and the table can filtered based on Expected Start Date. The dashboard also displays new BMU IDs to then prompt the user to search if it matches a CfD on the LCCC register. The dashboard includes dedicated pages for each individual factor or objective identified in Step 1.
Generated Metered Volume
To track the generated metered volume, the internal database was queried using SQL and filtered by the CfD’s BMU IDs. The resulting data was transformed into an effective and visually appealing format. It was further aggregated monthly to highlight high-level trends while maintaining the ability to drill down into more detailed information.
The Metered Volume Tracker graph on the left has enabled a clear depiction of the volume (MW/h) generated by the generators who are on track to join the CfD scheme, which is clearly colour coded and can be filtered by each BMU. There are some estimates where there are multiple BMU IDs for one CfD unit. For a generator to join the scheme, it must achieve 80% of its target generation capacity, measured by the load factor. This is determined by combining generation data extracted from the internal market data and the contracted capacity stated in the LCCC CfD register. The accompanying graph visually represents these figures, allowing users to hover over the bar chart to extract exact values. By tracking this data, we can assess if the generator is on track to reach the required 80% of its contracted volume, thereby estimating the likelihood of joining the CfD scheme. Additionally, a flag alerts users if a generator has exceeded the 80% load factor. This information is crucial for adjusting our forecasts accordingly. For example, if the load factor of a generator was consistently at 20% and they were expected to join the CfD scheme in a month, it would be unlikely they would join. This dashboard page also features the capability to drill down to half-hourly load factor granularity, providing detailed insights into specific periods of high and low load factors.
Wholesale Electricity Prices
Our forecasts can also be impacted by wholesale market prices of electricity hence the project was also driven by a comparison between Capture price and Strike price. The Strike Price is a defined as the pre-agreed fixed price a generator will earn per MWh of electricity generated for the lifetime of the contract. This is set via competitive auctions where renewable energy projects bid to secure CfD contracts. The Capture/Reference Price, on the other hand, is the volume-weighted average wholesale price that a given generator or technology is expected to achieve (wholesale).
If the agreed Strike Price is higher than the Capture Price, the generator will receive the difference as a top-up payment. This makes it financially optimal for the generator to join the scheme. However, if the Strike Price is lower than the Capture Price, the generator must pay the difference, making it suboptimal for the generator to join the scheme, potentially causing delays. The example below on the left provides insights for a closely watched generator, Dogger Bank A, expected to join the scheme soon. The Capture Price is consistently above the current Strike Price (which is indexed to inflation. This suggests that Dogger Bank will likely delay their contract, providing evidence that adjustments to our forecasts are necessary.
The choice of plots was important, it allowed a clear comparison between the real time capture price data and the Strike price of each upcoming CfD generator stated in the LCCC CfD register. Obvious trends can be spotted and there is a flag to alert the user if the mean difference is significant and worth incorporating into CfD cost forecasts.
Outages
The Elexon website provided data to track both planned and unplanned outages for the upcoming CfD generators. The API provided a dataset which was refined to contain the relevant fields and whittle down the plethora of data to the relevant generators to ensure readability and valuable insights were delivered. Additionally, a filter option was implemented, allowing users to zoom in on specific generators such as Dogger Bank as seen in the table below. This enabled us to clearly identify examples of planned outages that could impact our forecasts.
This CfD Market Monitoring dashboard helps the team with CfD forecasting and through this project it’s been clear it is a requirement to establish a clear purpose, understand the audience and complete meticulous data preparation. Despite some upcoming CfD units lacking BMU IDs for querying the internal database, focusing on the available data set can still yield valuable insights. By turning complex data into simple, readable graphs, this dashboard enables deeper analysis and can facilitate faster decision-making to enhance the accuracy of our CfD forecasts