Towards seamless irradiance nowcasting on intrahour to intraday scales using all-sky and satellite imagery

Photovoltaics (PV) power generation depends on atmospheric conditions and especially cloudiness causes strong fluctuations. Accurate nowcasting of the cloud situation and resulting solar irradiance for the next minutes and hours is thus essential for successful integration of PV power plants into the electricity grid. Ground based all-sky imagers allow for spatial and temporal high resolution snapshots of the cloud situation at the position of the plant at low costs. However, the limited area of the sky visible limits the possible forecast time when using cloud motion extrapolation. Commonly used consumer grade cameras further on do not easily allow for elaborate retrieval of cloud optical characteristics necessary for quantitative irradiance forecasts. In contrast, Meteosat satellite data allows for large spatial coverage and estimation of cloud optical properties suitable for intraday nowcasting but with lower spatial and temporal resolution. Due to the different resolutions, joint beneficial usage of all-sky and satellite imagery as well as smooth transitions between all-sky and satellite based nowcasting are challenging.

This work demonstrates the possibility of supplementary usage of satellite imagery for all-sky image based intrahour nowcasting using a simplified 4D-Var assimilation framework. The use of satellite imagery also allows for a seamless transition from intrahour to intraday irradiance nowcasting.

For the intrahour period, all-sky images from a pair of imagers are evaluated to obtain stereo cloud heights, cloud motion vectors and cloud masks utilizing a synthetic clearsky library. This data is combined into a 2D cloud mask advection model using a simplified 4D-var assimilation scheme. Data from multiple imagers and timesteps as well as satellite derived irradiance maps can be put to use within the assimilation framework and as boundary condition.
Within this work, satellite based irradiance maps are computed using methods based on Sirch (2018). Cloud optical properties are retrieved from Meteosat Second Generation (MSG) images with the APICS (Bugliaro et al. 2011) and CiPS (Strandgren et al., 2017) algorithms. These are horizontally advected using two separate layers for thin cirrus and thicker clouds, which are then input to radiative transfer calculations. Resulting irradiance maps can be used for the intrahour assimilation as well as seamlessly following intraday nowcasts.

 

References:

Sirch, Tobias (2018): Multi-resolution nowcasting of clouds and DNI with MSG/SEVIRI for an optimized operation of concentrating solar power plants. Dissertation, LMU München: Faculty of Physics

Bugliaro, L., Zinner, T., Keil, C., Mayer, B., Hollmann, R., Reuter, M., and Thomas, W.: Validation of cloud property retrievals with simulated satellite radiances: a case study for SEVIRI, Atmos. Chem. Phys., 11, 5603–5624,2011.

Strandgren, J., Bugliaro, L., Sehnke, F., and Schröder, L.: Cirrus cloud retrieval with MSG/SEVIRI using artificial neural networks, Atmos. Meas. Tech., 10, 3547–3573, 2017.