# Renewable energy forecasting with satellite data Satellite data improve solar forecasts in two ways. Geostationary weather satellites observe cloud fields over wide areas, and forecasters extrapolate the cloud motion to predict irradiance, typically four to six hours ahead. For longer horizons, satellite observations feed the numerical weather prediction models that forecast from several hours to days ahead. ## Why forecasts matter for the grid Solar and wind power vary with the weather. The resulting fluctuations in supply strain grid stability, and grid balancing has become a major issue as these intermittent sources expand [1]. Near-term forecasts help operators decide when to charge or discharge storage and when to ramp other generation up or down [1]. They also support generator scheduling and reduce the need for costly fast-ramping plants [3]. Assessing renewable production in real time from space data helps anticipate near-term energy prices as well [2]. ## How it works A forecast is built from different inputs depending on how far ahead it looks [3]: | Horizon | Main input | How it works | | ------------------------------------------------------ | -------------------------------- | -------------------------------------------------------------------------------------------------------------------------- | | Up to a few hours, in relatively stable sky conditions | Measurements at the plant | Statistical or machine learning models extend the measured time series; when the sky is variable, they rarely perform well | | 10 to 20 minutes | Sky cameras on the ground | Images of the local sky show clouds approaching the plant | | Typically 4 to 6 hours | Geostationary weather satellites | Cloud motion vectors extrapolate the observed cloud field, with updates every 10 to 30 minutes | | Several hours to days or weeks | Numerical weather prediction | Physical models simulate the atmosphere, including the formation and dissipation of clouds | Satellites play two roles in this chain. Geostationary satellites observe clouds over wide areas, and their images are the main input for solar forecasts up to several hours ahead [3]. Satellite observations also enter the weather models: at ECMWF they make the biggest contribution to forecast quality of all observation types [5]. State-of-the-art PV forecasting services do not rely on one model. Intraday systems combine online measurements, satellite-based forecasts and weather model forecasts, and use statistical or machine learning methods to optimise the result [3]. **Figure:** Typical horizon of each forecast input, on a logarithmic time axis. Measurements at the plant only reach a few hours ahead when the sky is stable. ## Which data and satellites ### Satellite-derived irradiance The Copernicus Atmosphere Monitoring Service (CAMS) provides global horizontal, direct and diffuse solar irradiation; the all-sky values are derived from Meteosat and Himawari observations. The record starts in February 2004 for the Meteosat field of view and is available from 1-minute to monthly resolution. CAMS names the monitoring of solar energy systems and their integration into supply grids among the uses [4]. Using such data to plan a new plant is covered under [site selection](https://leto.space/use-cases/site-selection/). ### Production estimates During operation, Earth observation supports yield estimation and short-term production forecasts for solar plants, and for hydropower it monitors precipitation, snowpack, soil moisture, river extent and reservoir levels to predict inflows [1]. ### Climate information for energy The Copernicus Climate Change Service delivers climate indicators relevant to the energy sector, such as electricity demand and renewables production [1]. ## What you get - intraday forecasts of PV output for a plant or a portfolio, typically at hourly or 15-minute resolution, which matters for intraday power markets [3] - day-ahead forecasts that combine several weather models with statistical or machine learning methods [3] - estimates of regionally aggregated PV power, which grid operators need for marketing PV power and for grid management [3] - satellite-derived irradiance time series for monitoring the plants you operate [4] ## Limits **Limits** - **Satellites cover the next hours:** Satellite cloud motion forecasts typically reach four to six hours ahead. Beyond that, numerical weather prediction is the main approach \[3]. - **No forecast without measurements:** Time-series models need on-site measurements to be trained \[3]. Satellite data do not replace your own metering. - **Accuracy depends on the climate:** The errors of weather model forecasts depend strongly on climatic conditions; the relative error is typically smaller in sunny areas \[3]. Ask providers for error figures from sites with a climate like yours. ## How we help Whether a better forecast pays off depends on what forecast errors cost you today, in balancing, storage or trading. We help you quantify that, define the required horizon, resolution and error measure, and compare providers on the same terms in a [requirements and procurement](https://leto.space/consulting/procurement/) project. [More on implementation and pilots](https://leto.space/consulting/implementation/) The marketplace of [Spacegraph](https://leto.space/spacegraph/), our discovery platform, has a category for energy and utilities. ## Related use cases - [Selecting sites for solar parks and wind farms](https://leto.space/use-cases/site-selection/): How satellite data help choose generation sites. - [Precise timing for power grids](https://leto.space/use-cases/grid-timing/): Why grid measurements depend on GNSS time. - [Automating renewable energy sales](https://leto.space/use-cases/renewable-energy-sales/): Which data help automate the sale of renewable energy systems. - [Energy and utilities](https://leto.space/industries/energy-and-utilities/): All use cases for energy companies and utilities. ## Frequently asked questions ### What data do we need to bring? The location and technical data of the plants, and their measured output. Forecasts built from local measurements need on-site observations of irradiance or PV power [3], and measured output is also what you test any forecast against. For forecasts of a whole region, a common difficulty is that PV power is not measured at sufficient resolution for most plants in many countries [3]. ### Do satellite data help with wind power forecasts too? Yes. Space data and services track weather patterns, cloud cover, solar irradiance and wind speed, which supports near-term forecasts of renewable production [1]. Radar data from Sentinel-1 are used for offshore wind resource mapping [1]; that question belongs to [site selection](https://leto.space/use-cases/site-selection/). ### How do we start a pilot? Pick one plant or one portfolio where forecast errors already cost money, for example through balancing energy or curtailment, and agree on the error measure you will use. Our [consulting overview](https://leto.space/consulting/) shows the path from the problem to operation, and an initial call shows whether a pilot is worth it. ## Sources 1. [EU Space Market Report 2026, Issue 1, EU Agency for the Space Programme (EUSPA), 2026-05-26](https://www.euspa.europa.eu/sites/default/files/documents/EU_Space_Market_Report_2026.pdf) 2. [EUSPA EO and GNSS Market Report, Issue 2, 2024, EU Agency for the Space Programme (EUSPA), 2024](https://www.euspa.europa.eu/sites/default/files/external/publications/euspa_market_report_2024.pdf) 3. [Best Practices Handbook for the Collection and Use of Solar Resource Data for Solar Energy Applications, Fourth Edition (Report IEA-PVPS T16-6:2024), International Energy Agency, Photovoltaic Power Systems Programme (IEA PVPS), Task 16, 2024](https://iea-pvps.org/wp-content/uploads/2024/09/Best-Practices-Handbook-for-the-Collection-and-Use-of-Solar-Resource-Data-for-Solar-Energy-Applications-Fourth-Edition-.pdf) 4. [CAMS solar radiation time-series (dataset overview), Copernicus Atmosphere Monitoring Service (CAMS), Atmosphere Data Store, accessed 2026-10-06](https://ads.atmosphere.copernicus.eu/datasets/cams-solar-radiation-timeseries?tab=overview) 5. [Fact sheet: ECMWF's use of satellite observations, European Centre for Medium-Range Weather Forecasts (ECMWF), 2020-07-06](https://www.ecmwf.int/en/about/media-centre/focus/2020/fact-sheet-ecmwfs-use-satellite-observations) Page: