Author | Elvira Esparza
The United Kingdom has improved its photovoltaic production forecasting with the use of Quartz Solar. This suite of AI tools enables more efficient grid management, as well as lower costs and reduced carbon emissions.
The United Kingdom’s energy system operator, Neso, has begun using the AI-powered Quartz Solar tool to forecast photovoltaic power generation and thus calculate more accurately the energy reserves needed to balance the grid. The goal of using this tool is to improve grid stability, reduce operating costs, and lower carbon emissions by minimizing the use of more polluting energy sources to compensate for inaccuracies in solar power forecasts.
The main advantage offered by Quartz Solar is that it speeds up and improves weather predictions several days in advance. With this lead time, Neso can respond more effectively to fluctuations in solar power generation, resulting in significant savings. Quantitatively, the use of this tool makes it possible to save about 30 million pounds (34.3 million euros) per year in backup energy costs and avoid the emission of roughly 300,000 tons of CO2.
Quartz Solar was developed by Open Climate Fix, a nonprofit climate tech organization, in collaboration with Neso’s Artificial Intelligence Centre of Excellence. The organization was founded in 2019 to connect cutting-edge research with the world’s energy systems. As a nonprofit, it shares data and code to encourage collaboration, maximize transparency, and accelerate the impact across the sector. It began working with Neso in 2021, and this year the tool has been integrated into the activities of the UK’s electricity system operator.
How does this tool work?
Quartz Solar works by using machine learning to analyze complex patterns in solar power generation. It incorporates real-time satellite images that capture clouds and weather conditions affecting solar output, along with the integration of meteorological data that enable more accurate solar production forecasts up to 36 hours in advance.
Quartz Solar uses, among other information sources, 12-channel satellite data with different wavelengths; meteorological data from the ECMWF (European Centre for Medium-Range Weather Forecasts) and the UK Met Office; and photovoltaic generation data provided by PV Live.
Artificial intelligence analyzes how clouds move and calculates the amount of sunlight reaching solar plants. This allows for accurate estimates of how much electricity these plants will generate in the coming hours. It also provides a national forecast and a regional breakdown based on supply points across the country’s grid.
What advantages does this AI technology offer in the power system?
The use of Quartz Solar has improved weather forecasts thanks to faster updates and a wider range of data and probabilistic models. Unlike traditional forecasting methods, which rely on numerical weather predictions updated every few hours and fail to capture sudden changes in cloud cover or solar output, the Quartz Solar model updates every few minutes, making it easier to track changes accurately. This tool has cut errors in half and is 2.8 times more accurate than traditional solar generation forecasting tools.
With this information, Neso can respond more effectively to fluctuations in solar power throughout the day, reducing reliance on fossil fuel backup used to balance the grid when solar generation is uncertain. The annual savings of 30 million pounds in grid imbalance costs could rise to 150 million pounds (171.4 million euros) by 2035, as more solar power is connected to the grid. The goal is to reach 70 GW of solar energy by 2035.
Other AI tools used for photovoltaic energy forecasting
In addition to Quartz Solar, which is planning to expand in India, there are other AI tools for solar energy forecasting currently operating in different countries:
Solcast. Founded in Australia in 2016, it is now used in numerous countries with solar plants across Europe, the United States, Latin America, and Asia. It uses satellites to monitor clouds and solar radiation and employs mathematical models to calculate current solar energy production and forecast output for the coming days. AI is applied to process large volumes of weather data and optimize predictions at different scales, from individual plants to entire power grids.
Tomorrow.io. Originating in the United States, it is used by energy companies worldwide. It leverages satellite and radar data, along with machine learning models, to forecast key variables such as irradiance, cloud cover, and temperature, which are used to estimate photovoltaic production. This technology helps reduce losses during sudden weather changes and allows for more advanced energy production planning.
SolarAnywhere. It is a cloud-based solar data and forecasting platform. It provides historical and real-time irradiance, photovoltaic production simulations, and forecasting tools used for planning, financing, and managing solar plants worldwide. It is applied in both solar project planning and plant operations.
Vaisala. Founded in Finland, it now operates worldwide. Although it is not an AI model itself, its sensors and radars form the foundation of modern forecasting systems. It combines data from weather stations, satellites, and atmospheric models, which are integrated into forecasting platforms that use machine learning to reduce errors and improve both solar and wind power predictions.
Thanks to these AI tools, data is now more accurate and available in advance. The most immediate result is savings on backup energy costs used to cover solar power variability, along with reduced carbon emissions by lowering the use of more polluting energy sources. Additionally, solar energy is better integrated into the energy mix, considering it is the fastest-growing renewable energy source worldwide.
Photos | Andreas Gücklhorn, Quang Nguyen Vinh
