Solar irradiance nowcasting system to optimize the yield in parabolic trough power plants = Solarstrahlungs-Kürzestfrist-Vorhersagesystem für die Ertragsoptimierung eines Parabolrinnenkraftwerks

Kurzfassung

One of the greatest challenges facing humanity in the 21st century is the transition to a fully decarbonized society. The most abundant energy resource available to mankind is the Sun. Concentrated solar thermal power (CSP) plants with thermal energy storages (TES), such as parabolic trough (PT) power plants, could provide a renewable source of dispatchable energy capable of balancing fluctuations in electrical grids caused by intermittent sources. Yet, whether PT power plants are going to play an important role in the future is mainly a question of cost. One of the ways to increase the competitiveness of PT power plants is the optimization of solar field controller. PT solar fields are complex spatially extended thermo-hydraulic facilities, which concentrate the direct normal irradiance (DNI) on receiver tubes. The operation of these solar fields is strongly affected by spatial and temporal variabilities of DNI, mainly caused by clouds. State of the art PT solar field controllers have only a limited awareness of the current spatial and temporal DNI variability, since these controllers have only access to irradiance measurements of one or a handful of pyrheliometers. All sky imager (ASI) based nowcasting systems can provide spatial DNI information with an adequate temporal and spatial resolution for PT solar field controller. These ASI-based nowcasting systems consist of cameras which take hemispherical images of the sky. The common working principle of ASI-based nowcasting systems includes cloud detection, -geolocation, -tracking and assessment of the corresponding current and immediate future solar irradiance. In the past years a manifold of distinct ASI-based nowcasting systems have been developed. The potential of nowcasting systems for control optimization is often highlighted in the literature. However, to the best of the author’s knowledge, there have been no investigations to date, which have analyzed the applicability of nowcasting systems with their corresponding uncertainties for the optimization of CSP power plants. The first objective of this thesis is the development of a real time capable ASI-based nowcasting system, qualified to describe complex but frequent multi-layer cloud conditions. Therefore, a system is developed which treats each detected cloud as an individual object with attributes such as geolocation, motion vector and transmittance. The processing steps 3-D cloud modeling, -tracking and the determination of the cloud transmittance are developed, benchmarked and combined to a modular nowcasting system, which creates DNI maps for lead times up to 15 minutes ahead. These DNI maps have a spatial extension up to 64 km² with a spatial resolution ≤ 20 m as well as an intra minute temporal resolution. The entire system is validated over two complete years with three spatially distributed reference DNI measurements. Furthermore, the same dataset is used to develop a real-time capable uncertainty analysis with an average coverage factor of 68.3%, taking into consideration spatial variations within the DNI maps. An additional dataset of one year is used to validate the uncertainty analysis. Both, the system validation as well as the uncertainty analysis indicate a strong dependency of the nowcast quality with the prevailing weather conditions. Secondly, this thesis investigates the applicability of DNI maps for the optimization of PT solar field controller, under consideration of the uncertainties. Solar field simulations are performed using the so called Virtual Solar Field (VSF), a detailed dynamic simulation tool. In a first step the DNI maps of the nowcasting system are classified in one of 7 combined spatial and temporal DNI variability classes. For each of the classes optimized control parameters are determined. Class depended control strategies with distinct objectives are benchmarked with a state of the art solar field controller. Results of detailed simulations over 22 days as well as performance estimations over two years indicate an overall significant benefit of roughly 2% in revenue for the novel class depended control strategies.