At a glance
Open Climate Fix, a nonprofit founded in 2019, built Quartz Solar, an AI system that combines live satellite imagery, weather data and machine learning to predict solar power output. It is now embedded in the control room of the UK's National Energy System Operator, where it is 2.8 times more accurate than the previous forecasts, halves large errors, and helps avoid around 300,000 tonnes of CO2 and 30 million pounds in costs every year.
- Quartz Solar is 2.8 times more accurate than the solar forecasting tools NESO used before
- The system halves large forecast errors and predicts solar generation up to 36 hours ahead
- Better forecasts mean less backup fossil fuel generation, avoiding around 300,000 tonnes of CO2 a year
- The tool saves an estimated 30 million pounds in imbalance costs annually
- Open Climate Fix shares its data and code openly so other grid operators can replicate the approach
Solar power is one of the cheapest sources of electricity in the United Kingdom, but it is also one of the hardest to predict. Clouds can cut the output of solar panels within minutes, and when a grid operator does not know how much solar energy is coming, it keeps fossil fuel plants standing by as backup. A nonprofit called Open Climate Fix has solved a large part of this problem with AI. Its Quartz Solar forecasting system now runs inside the control room of the UK's National Energy System Operator, and it is cutting around 300,000 tonnes of CO2 emissions every year, roughly the carbon impact of 111 million litres of diesel.
The numbers behind the project are remarkable for a small nonprofit. Quartz Solar is 2.8 times more accurate than the solar forecasting tools NESO previously used, it has halved large forecast errors, and it saves the grid operator an estimated 30 million pounds in imbalance costs annually. It is one of the clearest working examples of AI reducing emissions on a national electricity system today.
Why Solar Forecasts Are Hard
The UK now has tens of gigawatts of solar capacity, much of it installed on rooftops with no central monitoring. The operator cannot simply read a meter to know how much solar power is flowing at any moment. It has to estimate, and those estimates depend on weather forecasts.
Traditional forecasting relies on numerical weather prediction models that refresh only every few hours. These models are good at broad patterns but routinely miss fast moving changes in cloud cover. When a cloud bank rolls over the Midlands at noon, solar output can drop sharply within minutes. If the operator has not anticipated that drop, it must suddenly buy expensive reserve power, which in practice often means firing up gas plants.
Satellites, Machine Learning and Minute by Minute Updates
Quartz Solar takes a different approach. It combines machine learning with live satellite imagery and weather data to generate solar forecasts up to 36 hours ahead, then refines those forecasts minute by minute as new satellite frames arrive. Because the model can literally see clouds approaching solar farms, it catches the rapid shifts that older methods miss.
The system began as an innovation project supported by NESO's AI Centre for Excellence and funded through the Network Innovation Allowance. After extensive testing it moved into full production use, and today it touches the whole operating cycle, from day ahead scheduling to real time balancing in the Electricity National Control Room. It also feeds NESO's Dynamic Reserve Setting model, which decides how much backup capacity the grid needs to hold, and it improved NESO's total system demand forecast by 8 percent, generating further savings.
A Nonprofit Built by Insiders
Open Climate Fix was founded in 2019 by Jack Kelly, previously a researcher at Google DeepMind, and Dan Travers, who worked in energy trading and software before joining the cause. Their thesis was simple: the gap between cutting edge machine learning research and the daily reality of running an electricity grid was wasting huge amounts of renewable energy, and a small focused team could close it.
The organization runs on open principles. It shares data and code publicly so that other researchers, and crucially other grid operators, can inspect, reuse and improve the work. Dan Travers described Quartz Solar becoming core national infrastructure as a milestone for nonprofit climate tech, proof that open science AI can deliver results at scale every single day.
Why This Example Matters So Much
In November 2026, MIT Technology Review examined whether AI is actually reducing emissions in the real world, and the findings were sobering. A peer reviewed study led by researchers at the London School of Economics estimated AI could help cut global emissions by 3.2 to 5.4 billion tonnes of CO2 equivalent annually by 2035, with the single biggest opportunity in the power sector, where better forecasting could increase renewable electricity delivered from already built solar and wind facilities by up to 20 percent.
Yet when reviewers looked at grid operators in China, the United States and Europe, only one had deployed a known meaningful, working AI program of this kind: the UK's partnership between NESO and Open Climate Fix. The technology works, the savings are measured, and the playbook is public. What is missing is adoption everywhere else.
A Template Every Grid Can Copy
The 300,000 tonnes of CO2 avoided each year in the UK is meaningful on its own, equal to the electricity used by about 62,000 homes. But the bigger prize is replication. Solar capacity is expanding fastest in countries with abundant sunshine and far less forecasting capability than the UK. Every grid operator that adopts similar AI forecasting can squeeze more usable power out of panels that are already installed, without pouring a single additional centimetre of concrete.
Open Climate Fix has said openly that it is looking for new places and conversations with electricity operators around the world. Because the code and methods are shared openly, the barrier to entry is lower than almost any comparable climate technology. The story of Quartz Solar is not just about one clever model in one control room. It is about what happens when someone takes the world's most hyped technology and points it at one of the most mundane, most important problems on the path to clean energy: knowing, ahead of time, where the sun will be.
Common Questions
What is Quartz Solar?
Quartz Solar is an AI powered solar forecasting service built by the nonprofit Open Climate Fix. It combines machine learning, live satellite imagery and weather data to predict solar power output up to 36 hours ahead and refines its forecasts minute by minute.
How much CO2 does AI solar forecasting save in the UK?
According to Open Climate Fix and NESO, Quartz Solar helps avoid around 300,000 tonnes of CO2 per year, roughly the carbon impact of 111 million litres of diesel or the electricity used by 62,000 UK homes, by reducing reliance on fossil fuel backup generation.
How accurate is Quartz Solar compared with older forecasts?
Quartz Solar is 2.8 times more accurate than the solar generation forecasting tools NESO previously used and has halved large forecast errors. It also improved NESO's total system demand forecast by 8 percent.
Who runs the UK electricity grid?
The National Energy System Operator, or NESO, balances electricity supply and demand across the UK, managing tens of gigawatts of power in real time from the Electricity National Control Room.
What is Open Climate Fix?
Open Climate Fix is a UK nonprofit founded in 2019 by Jack Kelly, formerly of Google DeepMind, and Dan Travers. It builds open source AI tools for energy systems and shares its data and code so other grid operators can replicate its work.
Sources: Open Climate Fix NESO case study and Quartz Solar control room announcement; National Energy System Operator innovation programme reports; MIT Technology Review, The climate case for AI has not been made (November 2026); Nature Climate and Systemic AI study on AI emission reduction potential (LSE Grantham Research Institute and Systemiq).