In this document
This article explains why solar power forecasting matters for grid stability, economic benefits, and operational efficiency, and outlines best practices such as combining forecast models, choosing appropriate temporal resolution and horizon, and quality-controlling reference data. It then describes the Solargis approach delivered through the Solargis Forecast service: the Solargis Cloud Motion Vector model combined with four Numerical Weather Prediction models, the resulting spatial resolution and update rates, the parameters and subscription options available, data collection and quality control, and how forecast accuracy is validated. It closes with a comparison of Solargis against Reuniwatt, Meteologica, Solcast, and Enercast.
Solar power forecasting
As the world increasingly relies on renewable energy sources, solar power has become a crucial component of the global energy mix. However, integrating solar energy into the grid poses challenges due to its intermittent nature, which is heavily influenced by weather conditions. This is where solar power forecasting plays a vital role. Accurate forecasting is essential for maintaining grid stability, optimizing energy trading, and ensuring the economic viability of solar projects.
Importance of solar power forecasting
Grid Stability: Forecasting helps Distribution System Operators (DSOs) and Transmission System Operators (TSOs) balance energy supply and demand, preventing power imbalances and potential blackouts
Economic Benefits: Accurate forecasts reduce financial penalties for PV asset operators by ensuring that the generated power aligns with contractual obligations. It also enables traders to make informed decisions, maximizing profits in energy markets
Operational Efficiency: Forecasting aids in sizing battery energy storage systems effectively, smoothing power output fluctuations, and scheduling maintenance during periods of low energy production
Best practices in solar power forecasting
Following is a list of best practices for optimal solar power forecasting service:
Combining forecast models: Utilize a combination of different models (e.g., Cloud motion vector and Numerical weather prediction model) to leverage their strengths.
Temporal resolution and forecast horizon: Provide forecasts with appropriate temporal resolutions (e.g., hourly or 15-minute) and forecast horizons (from intraday up to 14 days) to meet specific operational needs.
Reference data quality control: When evaluating forecast accuracy against reference data (actual measured solar radiation or PV power output), ensure the reference data is quality-controlled to provide a real picture of the forecast accuracy.
Forecast accuracy evaluation: Use consistent metrics such as Mean bias deviation (MBD), Mean absolute deviation (MAD), or Root mean square deviation (RMSD) to assess forecast accuracy. Normalize these metrics to facilitate comparison across different PV power plants.
The Solargis approach
At Solargis, we employ a comprehensive approach to solar power forecasting, combining advanced models and rigorous data quality control to provide accurate and reliable forecasts. Our process is designed to support efficient grid management, optimize energy trading, and enhance the economic viability of solar projects.
By following this structured approach, we provide high-quality forecasting services that enhance the efficiency and profitability of solar energy projects.
The forecasting solution is currently provided via the Solargis Forecast data service.
Forecast model selection and combination
Cloud Motion Vector (CMV) Model: Utilized for short-term predictions (up to approximately 3 hours), this model tracks cloud movements using satellite imagery to predict immediate changes in solar radiation.
Numerical Weather Prediction (NWP) Models: These models simulate atmospheric conditions over longer horizons, providing accurate forecasts beyond 3 hours and up to 14 days ahead.
Combining CMV and NWP models allows us to leverage their strengths, offering more accurate and reliable forecasts across different time horizons.
Models used: Solargis Forecast uses our proprietary CMV model and four different NWP models: HRRR by NOAA, ICON and ICON EU by DWD, IFS and IFS Ensemble by ECMWF, and GFS by NOAA.
Spatial resolution: Nowcasts and short-term forecasts for solar and PV parameters are derived from the Solargis CMV model at 0.033° (~4 km). In the continental USA, the forecast for the next 2 days is available with an hourly update rate and approximately 3 km spatial resolution. The spatial resolution of the final delivered data (solar and PVOUT parameters) is improved by the Solargis downscaling process using horizon shading, based on the near-global SRTM terrain database with a spatial resolution of 90 m (3 arc-seconds). This step is particularly important in areas with complex terrain.
Update rates: CMV data is updated every 5 minutes in the contiguous USA, every 10 minutes in the Americas and East Asia/Pacific, and every 15 minutes in Europe, Africa, and Asia. NWP-based forecasts are updated every six hours, and every hour in CONUS up to D+2.
No production data required for calibration: Solargis Forecast does not require real production data for calibration, relying on advanced simulation and robust quality control.
More about the modelling is described in the Forecast modelling document. Full model coverage, resolution, and update rate details are listed in the Forecast subscription details document.
Forecast generation and delivery
Forecasts are generated in a Time Series format, which facilitates easy integration into operational systems. Temporal resolutions can vary from 5-minute to hourly or daily intervals, depending on customer needs.
Forecast horizons extend from Nowcast (minutes to a few hours ahead) to 14 days, enabling both short-term operational decisions and medium-term strategic planning.
Conversion of solar irradiance forecasts to PV power forecasts is based on physical models that consider site conditions such as albedo and shading, and up to 20 configurable PV system parameters.
In addition to PVOUT (PV power output), the service provides access to over 30 solar, meteorological, and environmental parameters, enabling a comprehensive understanding of site conditions and supporting effective operational planning.
Clear sky and uncertainty variants are available for PVOUT, GHI, DNI, DIF, and GTI.
Geographical coverage: Land surface and coastal seas between latitudes 60°N to 56°S, and between 60°N and 65°N in Scandinavia and America.
Delivery methods: All Solargis Forecast services can be delivered via SFTP (a secure file exchange server) or via API, using an access token.
Subscription options: Forecast Basic provides a selection of forecast parameters in hourly resolution from today to D+7. Forecast Professional and Forecast Enterprise offer all forecast parameters with up to 5-minute resolution from today to D+14. Historical Forecast covers the past two complete calendar years for horizons H+0, H+1, H+2, H+3 and D+0, D+1, D+2. Portfolio Aggregation provides up to 5-minute resolution from today to D+14.
Portfolio aggregation: For portfolios that manage several PV power plants together, individual plants are grouped into Virtual Power Plants (VPPs) based on technical parameters and location, allowing Solargis to generate forecasts for large numbers of power plants while maintaining delivery speed and accuracy. Forecasts are aggregated across all VPPs, and the resulting PVOUT representing the whole portfolio is delivered to the customer.
Service tiers and delivery methods are described in the Introduction to Forecast document, and the full technical specification is in the Forecast subscription details document.
Data collection and quality control
To evaluate the accuracy of our forecast, we collect historical and real-time data from various sources, including satellite imagery and ground measurements.
Ensuring data quality is crucial; We rigorously check the data we use to evaluate our models and to ensure a high-quality baseline and accurate performance assessment.
For solar power forecasts, we integrate data from global NWP models such as IFS, GFS, and ICON, along with regional NWPs like ICON-EU and HRRR. We acquire these data from systems operated by international and national agencies ECMWF, NOAA, NASA, and DWD.
We use solar and meteorological measurements from public meteorological networks and private projects to validate and calibrate our models and processing algorithms.
The forecast data streams are systematically quality-controlled and their accuracy is enhanced through internal validation procedures.
Details about the data sources and its processing is described in the Origin of the data document.
Forecast accuracy evaluation
We use metrics such as Mean Bias Deviation (MBD), Mean Absolute Deviation (MAD), and Root Mean Square Deviation (RMSD) to assess forecast accuracy.
Normalizing these metrics against installed capacity or AC power limitation enables comparison across different PV power plants.
In our published validation study, GHI forecasts are validated against Solargis historical time series data generated by the satellite-based solar model, and PVOUT forecasts against Solargis PV power historical time series data generated by the satellite-based solar model and PV simulation model.
The study covers 153 global sites across all continents for the calendar year 2025, evaluating "hour-ahead" (nowcasting) and "day-ahead" (planning) horizons at hourly and 15-minute resolutions. Results are grouped into five major climate zones: temperate, cold, arid, tropical, and polar.
Detailed validation methodology is outlined in the Validation of forecast data document. Factors that influence achievable forecast accuracy are described in the Managing forecast accuracy expectations document.
Comparing Solargis with leading solar power forecasting providers
The solar power forecasting market is competitive, with several reputable providers offering advanced services. Here, we compare Solargis with some of the leading competitors, highlighting key features and benefits.
Note: the information below is taken from public documentation of the service providers, and we cannot guarantee that it is up-to-date.
Feature | Solargis | Reuniwatt | Meteologica | Solcast | Enercast |
|---|---|---|---|---|---|
Forecast models | Satellite-based nowcast + NWP | Nowcast + NWP (Sky camera as an extra) | NWP | Nowcast + NWP | NWP |
Temporal resolution | 5-min to daily | Up to 10-min (up to 1-min for Sky camera) | hourly | 5, 10, 15, 20, 30, 60-min | 15-min |
Forecast horizon | Up to 14 days | Up to 10 days | Up to 14 days | Up to 14 days | Up to 31 days |
Data delivery method | API, SFTP | API, SFTP, email (http, MODBUS available for Sky camera) | Not stated | API, web download | API, SFTP, email |
Global coverage | Land surface and coastal seas between 60°N and 56°S, and between 60°N and 65°N in Scandinavia and America | Yes | Yes | Yes | Yes |
Key differentiators
Each provider offers unique strengths in solar power forecasting, but several key differences emerge:
Solargis stands out for its comprehensive model combination (satellite-based nowcast plus NWP), which enhances both short-term and medium-term forecast accuracy. Its flexible temporal resolution (5-min to daily) and extended forecast horizon (up to 14 days) cover both operational and trading needs. The data delivery methods (API, SFTP) ensure seamless integration into diverse workflows.
Reuniwatt uniquely offers integration of sky cameras, making it ideal for short-term (up to 30 minutes), high-frequency forecasts for sites where sky cameras can be supported. However, the temporal resolution of its general forecast starts at 10 minutes, and the forecast horizon is limited to 10 days.
Meteologica offers short- and medium-term forecasts through NWP models. Its forecast horizon is up to 14 days. Due to limited public information about the parameters of the service it is not possible to evaluate its advantages.
Solcast provides high-frequency forecasts (5-min to hourly) and matches Solargis in forecast horizon (up to 14 days), utilizing satellite-based nowcasting and NWP models.
Enercast offers the longest forecast horizon (up to 31 days) and 15-minute resolution, focusing on longer-term planning. However, it uses only NWP models, which may not capture rapid cloud dynamics as effectively as hybrid approaches.
Solargis combines the strengths of both satellite-based nowcasting and NWP models across short- and medium-term horizons, without the need for any on-site equipment, like sky cameras, and without requiring real production data for calibration. Its flexible temporal resolution, extended forecast horizon, and multiple delivery formats make it adaptable for a wide range of applications - from real-time plant operations to market trading.