Solar irradiance nowcasting by LightGBM with cloud information from sky images

Abstract

In this paper, clouds were extracted from all-sky images. The states of the clouds, such as the direction from the sun and the thickness, were classified. And we clarified that this information is effective for highly accurate solar radiation prediction. It has used that LightGBM to make predictions based on the amount of solar radiation, the estimated amount of solar radiation in fine weather, and the positional information such as the distance and direction from the sun of clouds extracted from all-sky images, and the effectiveness of the information was confirmed. Overall, it was shown that cloud information considering the estimated value of solar radiation in clear weather and the distance and direction from the sun is effective.