A Data-Driven Model for Predicting the Energy Interception in Solar Central Receiver System

Central receiver systems are one of the most promising large-scale solar energy generation technologies. The core component of these systems is a heliostat field comprising a set of sun-tracking mirrors to focus sunlight onto a receiver. The energy intercepted by the solar heliostat field is based on incident direct normal irradiance and mirror surface area. The amount of direct normal irradiance is very predictable using clear sky models however in real-world systems it is impacted by several factors including suspended dust, clouds and air pollution. This paper presents a data-driven method to determine the irradiance loss associated with the presence of clouds and its impacts on energy generation by the central receiver system. The data-driven mode is based on the random forest machine learning method which was trained on a one-year solar irradiance dataset. The results show that the data-driven model is better at predicting solar irradiance compared to conventional analytical methods. A comparative analysis is also presented at the end of this paper.