Verification of sectoral cloud motion based direct normal irradiance nowcasting from satellite imagery

Abstract

Abstract. The successful integration of solar electricity from photovoltaics or concentrating solar power plants into the existing electricity supply requires an electricity production forecast for 48 hours, while any improved surface irradiance forecast over the next upcoming hours is relevant for an optimized operation of the power plant. While numerical weather prediction has been widely assessed and is in commercial use, the short-term nowcasting is still a major field of development. European Commission's FP7 DNICast project is especially focusing on this task and this paper reports about parts of DNICast results. A nowcasting scheme based on Meteosat Second Generation cloud imagery and cloud movement tracking has been developed for Southern Spain as part of a solar production forecasting tool (CSP-FoSyS). It avoids the well-known, but not really satisfying standard cloud motion vector approach by using a sectoral approach and asking the question at which time any cloud structure will affect the power plant. It distinguishes between thin cirrus clouds and other clouds, which typically occur in different heights in the atmosphere and move in different directions. Also, their optical properties are very different -especially for the calculation of direct normal irradiances as required by concentrating solar power plants. Results for Southern Spain show a positive impact of up to 8 hours depending of the time of the day and a RMSD reduction of up to 10% in hourly DNI irradiation compared to day ahead forecasts. This paper presents the verification of this scheme at other locations in Europe and Northern Africa (BSRN and EnerMENA stations) with different cloud conditions. Especially for Jordan and Tunisia as the most relevant countries for CSP in this station list, we also find a positive impact of up to 8 hours.

The Seamless Solar Radiation (SESORA) Forecast for Solar Surface Irradiance—Method and Validation

“…In the literature there are many possibilities of validating a solar irradiance nowcasting. Many approaches for such a nowcasting with the use of cloud motion vectors have been proposed [25][26][27][28][29]. A comparison with other studies is not always simple, as the region or the validation method can differ.…”

“…Schroedter-Homscheidt and Gesell for example proposed a nowcasting for DNI over Spain and validated it by calculating the bias. After approximately 5 h the bias equaled 130 W/m 2 [27]. Nonnenmacher and Coimbra as well as Gallucci et al used the RMSE as their validation method for the region of San Diego and Italy.…”

“…These vectors are derived from the infrared channel of MSG (Meteosat Second Generation) and since they do not only display the motion of the top of clouds they are called AMVs. In the field of solar radiation forecasts the use of cloud motion vectors is very common [25][26][27]. Comparable studies from Gallucci et al [28] or Sirch et al [29] used cloud motion vectors from MSG/SEVIRI (Spinning Enhanced Visible and Infrared Imager) to forecast solar surface radiation for up to 2 h. Due to different error measures and different study areas, a direct comparison of the quality reported in both studies is not possible.…”

Cloud height and tracking accuracy of three all sky imager systems for individual clouds

“…More accurate forecasts up to 8-9 h ahead can be achieved by satellite based systems (Schroedter-Homscheidt et al 2016). Due to the spatial and temporal resolution, satellite forecasting systems are not suitable for intra-hour forecasts with sub minute temporal resolution.…”

Determination of Cloud Motion Applying the Lucas-Kanade Method to Sky Cam Imagery

“…Furthermore, it is fundamentally important that plant operators can anticipate the amount of solar energy that will be generated in plants so that they can react in advance to future fluctuations in electricity generation. The successful integration of solar electricity from CSP plants into the existing electricity supply requires electricity production forecasting of between 24-48 h to cover the current day and the day ahead [5,6]. Cloud cover has been widely studied in different places around the world, above all to examine the role of clouds in solar radiation measurements [7][8][9][10].…”