A solar resource classification algorithm for global horizontal irradiance time series based on frequency domain analysis

Abstract

A novel daily and 15-min period solar resource classification algorithm, based only on the Fourier transform of global horizontal irradiance (GHI) and calculable parameters, is presented. These methods can identify GHI series that are not attenuated by clouds, whether the clouds may or may not be visible. Additional sky condition classes are also adopted in which the representative conditions of the remaining days or periods are classified. Although the proposed day classification is too conservative to be widely applicable, the 15-min period classification is generally in good agreement with the Perez sky clearness index while providing greater sensitivity to turbid conditions and algorithm flexibility. It is shown, however, that the adoption of the clearness index in the proposed 15-min period method less successfully distinguished between clear and constant overcast conditions under winter continental conditions due to relatively large solar zenith angles. It is also shown that the proposed 15-min period and Reno and Hansen statistical methods are generally in good agreement on the detection of clear conditions under temperate conditions. The classification results of the proposed method are further shown in comparison to high-resolution sky images to elucidate the character of each class and the classification algorithm flexibility.