Evaluation of the performance of different models for predicting direct normal solar irradiance

Solar energy is considered as a clear and sustainable energy resource, and the application of solar energy includes electric power generation and solar concentrators. The precise estimation of solar irradiance plays an important role in evaluating the performance of active solar energy utilizations such as concentrator photovoltaic systems. While global solar irradiance received by a horizontal surface can be easily measured, the availability of direct normal irradiance (DNI) is quite limited. Many models for predicting DNI have been developed and a number of them provided a satisfactory performance. However, it may be difficult for users to efficiently pick up the appropriate models that can be applied to their projects. This study analyses the solar irradiance data in Hong Kong based on continuous measurements and evaluates the performance of three empirical and machine learning models. The accuracy of individual approaches was evaluated using measured Hong Kong data. The results would be helpful to select suitable DNI prediction models for various applications.