Geographic validation of Satellite-Based Solar Irradiance Forecasting Models Across Europe

Solar forecasting is a critical tool for mitigating the variability of solar power availability within the electric grid, notably influenced by atmospheric conditions. These models often fail to adapt to the diverse climatic and cloud conditions across different regions [1], complicating efforts for PV project developers to estimate forecast accuracy without historical dataset of in-situ and hindcast.

Among a great number of solar forecasting applications and approach, we focus on intra-day solar forecasting based on geostationary satellite. We examine the sensitivity of solar irradiance forecasting accuracy to varying geographic locations and cloud conditions across Europe. Our goal is to assess how these models perform under different settings and their generalizability across diverse European regions and weather situations.

To evaluate our methodology, we employed a cloud motion vector (CMV) algorithm through the high-resolution visible (HRV) channel of the Meteosat Second Generation (MSG) SEVIRI instrument [2]. The scope of our study was confined to Western and Central Europe to ensure comparable satellite spatial resolutions, thereby eliminating spatial resolution as a significant source of variability in performance comparison. We selected six BSRN stations [3], which underwent advanced quality checks [4], with half located in Cfb (temperate oceanic) climate zones and the other half in Dfb (humid continental) zones. We analyzed a complete year's dataset from these stations to evaluate the forecasting model’s performance across the different seasons, cloud coverage, meteorological regimes and spatial locations.

Preliminary analysis using the persistence method has identified three distinct performance clusters of geographical locations, each represented by one meteorological station from both Dfb and Cfb climatic zones. To enhance this analysis, the CMV algorithm will be employed as it quantifies advection movements of clouds, thereby improving our understanding of performance variability across diverse sites. 

[1] Liu, Bai, et al. "Predictability and forecast skill of solar irradiance over the contiguous United States." Renewable and Sustainable Energy Reviews 182 (2023): 113359.

[2] Cros, Sylvain, et al. "Reliability predictors for solar irradiance satellite-based forecast." Energies 13.21 (2020): 5566.

[3] Driemel, Amelie, et al. "Baseline Surface Radiation Network (BSRN): structure and data description (1992–2017)." Earth System Science Data 10.3 (2018): 1491-1501.

[4] Blanc, Philippe, et al. "Data sharing of in-situ measurements following GEO and FAIR principles in the solar energy sector." (2022).