Understanding the true scale of solar generation on the electricity network is crucial for supporting the transition to Net Zero. AI for Visibility and Forecasting of Renewable Generation is exploring the potential of artificial intelligence to accurately predict existing solar capacity on the network. By improving visibility of installed solar, we can make smarter planning decisions—avoiding unnecessary spending on network upgrades or flexibility services, and ensuring customer money is used efficiently.
What is the project about?
To keep the electricity network running smoothly and reliably, UK Power Networks needs a clear picture of how much energy is being generated — especially from solar — and how much will be used on any given day. This balance is essential to avoid overloading equipment and to make sure supply meets demand.
One way this balance is managed is through buying flexibility — essentially paying customers to shift their electricity use to off-peak times, like moving usage from busy winter evenings to quieter overnight hours. However, a lot of solar generation is “invisible” to the network because it sits behind customers’ electricity meters. This makes it difficult to accurately forecast generation and demand.
This project will use artificial intelligence to uncover and map hidden solar generation and improve forecasting accuracy. With better forecasts, UK Power Networks can make smarter decisions — only upgrading infrastructure or buying flexibility services where it’s truly needed. This ensures customer money is spent wisely while supporting a low-carbon, high-performance network.
How we’re doing it
The project is developing machine learning models that can estimate how much solar generation capacity is already connected to the network by analysing historical satellite imagery and weather data.
Once we can reliably estimate this existing capacity, we’ll take it a step further—training artificial intelligence to forecast how much solar energy is likely to be generated on future days. To test how well the AI performs, we’ll give it past scenarios — like specific weather conditions from previous years — and check how closely its predictions match the actual generation that occurred.
This approach will help UK Power Networks plan more accurately, reduce unnecessary upgrades, and make better use of customer flexibility—all while supporting the transition to a greener energy system.
What makes it innovative
While solar capacity forecasting isn’t new, the sophistication of our approach sets this project apart. Most forecasting has historically been done at a national level — but we’re taking it further by applying these techniques at the site level, providing a much more granular and locally relevant view of solar generation.
What’s truly groundbreaking is our novel use of historical substation meter data to infer behind-the-meter solar capacity. As far as we know, this method hasn’t been used in this way before — not just in the UK, but globally. By combining this unique data insight with machine learning, we’re creating a pioneering solution that could transform how networks manage solar generation at the local level.
What we’re learning
The main area of learning is about how to create and improve AI algorithms to make forecasting solar generation and capacity more accurate. The algorithms that are produced will be made available to other organisations so that they can also benefit from the learnings made by the project.
Share your ideas
If you have an idea that could reshape or revolutionise the way we work, we want to hear from you.
We use cookies to give you the best possible experience when using our website. Many of these cookies are essential to the efficient operation of this site.
