ARTIFICIAL INTELLIGENCE SOLAR FORECAST FOR TODAY AND TOMORROW

Last update: September 2020

presentation

⇒ This service provides time series of Artificial Intelligence solar radiation forecast for today and tomorrow.

This service is based on a machine learning method called Gradient Boosting.

⇒ The features of this service are:

  • Solar radiation components: global, direct, diffuse and reflected irradiation over horizontal, tilted and normal planes
  • Time step of 1, 5, 10, 15, 30 and 60 minutes
  • Time reference is Universal Time (UT)
  • Meteosat 11 satellite images with 3 km resolution
  • Consider the shadowing effect due to far horizon
  • Aerosols and pollution provided by CAMS IFS weather forecast
  • Forecast uncertainties measures
 

WHY USING THIS SERVICE?

This service is useful to :

  • Integrate solar radiation forecast for storage and sales revenue optimization
  • Manage today deficit of production with the forecast values made the day before
  • Optimize the electricity storage system associated with the power plant production
  • Manage efficiently your electricity consumption from your off grid solar panels
  • Complement meteorological forecast applications.

This service can provide along with solar radiation data the meteorological data from GFS: Temperature (at 2 m), Relative Humidity (at 2 m), Pressure (at 2 m), Wind speed and direction (at 10 m), RainfallSnowfall and Snow depth.

 

Solar AI forecast demonstration updated in real time

The graph hereafter is the Global Horizontal Irradiation 15 minutes forecast over Carpentras, France (44.083, 5.059).

It presents in real time the current forecast for today and tomorrow as well as the real-time HelioClim3v5 radiation.
 
⇒ The forecast is made using GFS Numerical Weather Prediction (NWP) forecast updated 4 times a day.
The GFS solar radiation forecast is represented as black dashed line. It shows the variability induced by the machine learning method.
 

On the following graph over Petrolina, Brasil (-9.068, -40.319), the blue and green color filled intervals represent the forecast uncertainties. It is useful for the electricity storage system.

methodology

⇒ Solar AI forecast service uses as inputs :

 

QUALITY RESULTS OF THE SERVICE

This service has been evaluated to check the quality of the output data. We compared HelioClim-3 ai solar forecast radiation values with in-situ measurements from pyranometers. All the validation results are available here.

 

⇒ This service requires a learning step and is available on-demand.