The SolarForecast is based on the global weather forecasting models IFS (ECMWF) and GFS. The output of the GFS forecast model is then statistically optimised using measurement data from weather stations. This system allows for the production of an irradiation forecast with high accuracy specifically for your site.
The GFS model is updated 4 times a day. The GFS MOS (model output statistics) forecast data which the SolarForecast relies on is updated hourly based on the latest measurement data from weather stations. This improves the forecast significantly for the first few forecasted hours. The figure below shows a comparison between the forecast based on the 00:00 UTC run of the GFS (available around 4 am UTC+1), the values of the hourly updates, the measured values and the clear sky radiation.
Comparison of day-ahead forecast and last hourly update with measured values and clear sky radiation for Bern on June 25th 2013.
The accuracy of the SolarForecast is in general site dependent. An internal benchmark for 17 sites in Germany and 16 sites in Switzerland shows relative root mean square errors (rel. RMSE) for hourly forecast values based on the 00:00 GFS run for the first day of 33% to 43%. For southern Spain the rel. RMSE is 20% to 25%.
With this accuracy the SolarForecast by Meteotest is among the most precise available.
The SolarForecast Standard delivers hourly values for the next 72 hours for the following parameters:
- Global radiation [W/m2]
- Air temperature [°C]
With the SolarForecast Advanced you receive in addition the following parameters and options:
- Direct and diffuse radiation [W/m2]
- Direct normal radiation [W/m2]
- Global radiation on the inclined plane [W/m2]
- Reference yield (for standard module and standard inverter) [Wh/kWp]
- Additional meteorological parameters such as precipitation, relative humidity, wind speed and wind direction
- Sub-hourly forecast for any parameter can be delivered with a resolution of 10, 15, 20 or 30 minutes. Sub-hourly values are obtained through interpola- tion for the specific site.
- Forecast for more than 72 hours (up to 7 days)
The product SolarForecast Kalman uses real-time on-site measurements of radiation and/or yield to improve the forecast with a post processing using a Kalman filter.
A Kalman filter is a statistical algorithm which improves the forecast by eliminating systematic deviations between forecast and measurement data. In order for SolarForecast Kalman to function properly, the measurement data also needs to have hourly resolution and be updated every hour. The measured data needs to be submitted to the webservice using the HTTP Post protocol. Accepted data formats are JSON and CSV (comma separated values). Meteotest does not check the submitted data for completeness and consistency.