We're hiring in our ML team! 💻
Learn more
Solar power is an essential part of the UK’s energy mix, providing a sustainable and clean source of electricity that enhances both our energy security and our climate goals.
Solar power generation is growing exponentially and is forecast to be the largest form of electricity generation capacity globally by 2040.
With this vast potential comes an obvious element of unpredictability: the weather, with solar panels generating far less electricity when a cloud moves overhead.
This uncertainty comes at a cost – millions of pounds and hundreds of thousands of tonnes of CO2 in the UK alone – hampering the efficient utilisation of solar energy.
Forecasting solar energy generation is vital to maximise the efficiency of this power source.
But what if we could more accurately predict future cloud cover, allowing solar energy to be more efficiently integrated into the electricity grid?
For the last two years, Open Climate Fix has been running the forecast of solar energy generation used in the control room of the UK’s National Grid, helping their engineers balance the grid and save cost and carbon.
Current AI solar generation models assimilate multiple weather forecasts; however, traditional weather forecasts have limitations. Firstly, they focus primarily on wind and temperature prediction, and clouds are particularly difficult to predict at the fine scale required.
Secondly, many of the decisions required to run the electricity grid efficiently – like when to turn on back-up fossil fuel generation – are made hours or even minutes ahead of time. But traditional weather models take hours to run on massive supercomputers, meaning they are out of date as soon as they are produced.
Open Climate Fix researchers have incorporated satellite imagery of clouds from up to an hour in the past into their AI model, which reduced the errors in forecasts by 15-20%.
However, a more desirable solution would be to predict future cloud cover rather than using only past cloud information from satellite data. In this endeavour Open Climate Fix has joined forces with the Turing to launch a project called Quartz Solar AI Nowcasting, which aims to explicitly predict cloud movements for up to four hours ahead.
This involves using powerful AI to analyse vast amounts of historical satellite imagery and weather data to identify patterns of cloud movement and evolution. This brings greater precision, speed and predictive power, with these cloud predictions then fed into the existing solar generation forecasts. Forecasts will need to run fast enough to work in an operational environment, taking minutes rather than the hours of traditional weather forecast models.
Greater accuracy will bring a range of benefits. Better forecasts will help manage the energy grid, cutting down the need for back-up power, reducing costs and ensuring a steady supply of electricity for our homes and businesses.
Of course, more efficient use of renewables like solar means lower emissions, helping us hit carbon reduction targets, reduce air pollution and mitigate climate change.
It’s not just solar farms and energy grids who could benefit; forecasting cloud cover could facilitate anything from better flood warnings to predicting whether you need to apply sun protection when you’re out and about.
Crucially this project is scalable, and if successful would allow us to produce a cloud forecast for anywhere in the world, with the potential to save tens of mega-tonnes of carbon emissions.
We are excited about the potential of AI to maximise renewables, and ongoing research will explore the application of these approaches to other renewable energy sources, such as wind power.
Partnerships will be key to our success, and by maximising the reach of the Turing and Open Climate Fix we hope to mobilise academia, industry and government to advance the technology further.
Policy makers and foundations can also play a crucial role by providing funding, shaping policy frameworks and promoting ethical deployment of these technologies.
With this in mind, we were delighted recently to see Quartz Solar AI Nowcasting recognised as a finalist in the inaugural Manchester Prize which recognises cutting-edge AI solutions for public good, including through an award of £1M to develop the project further.
Ultimately, we believe AI will be pivotal in helping to achieve a step change in green energy and by working together, and with other partners across the ecosystem, we have the best chance of driving innovation within an ethical framework, moving the world towards a more sustainable future.
