Short-Term Forecasting of PV Power with Sky Imagers

Recording of cloud cover
© Fraunhofer ISE
Recording of cloud cover by a cloud camera.

We use sky imagers for PV power forecasting with high spatial and temporal resolution. The camera images have a higher resolution than satellite images and offer grid operators and PV plant operators forecasts which are more accurate up to 15 minutes ahead. The forecasts can be used by grid operators to ensure grid stability or to control systems such as hybrid power plants. PV plant operators use the forecasts to control their PV plant performance.

The spatial and temporal resolution and the range depend on the cloud situation and the time of day. Typical values are, for example, a temporal resolution of 10 seconds for a forecast period of 15 minutes. The spatial resolution is usually around 10 meters, and a cloud camera can record an area of about 10 x 10 kilometers.

We install sky imagers directly at your PV plant location or at different points within a power distribution network.

Our sky imagers are equipped with a fisheye lens and can capture an image of the entire sky. Based on the sky images, we develop algorithms that predict the movement of individual clouds. We extrapolate the cloud motion into the future and use it to predict solar irradiance. The short-term solar irradiance predictions are then fed into PV power simulation models to determine the expected PV power.

We use PV power predictions in the PV2WP project for example. The aim of this project is to control the operation of a hybrid PV/heat pump system in such a way so that self-consumption is optimized and steep ramps in the electricity feed-in to the grid are reduced.


Our Services:

  • Installation of the sky imager(s)
  • Perform forecasts of solar irradiation and PV power in real time
  • Predict the PV power output in Freiburg and Ulm, based on sky imagers which are already installed
  • Develop algorithms

Forecast Periods

We offer forecasting systems for the following time ranges

Short- and Medium-range Forecasting