We have been developing a short-term solar radiation prediction model using meteorological satellite images. In this study, we examined the method of creating confidence intervals using quantile regression as probabilistic information for the prediction. We investigated the impact of differences in variables and methods used in quantile regression on the coverage rate, interval width, and deviation width of the confidence intervals. As a result, it was found that using the clear sky index and ensemble forecast data from meteorological models as variables in quantile regression could reduce the interval and deviation width.
