통계적 기법을 이용한 일사량의 공간 및 시간적 상세화

Abstract

We evaluated the forecast performance of solar radiation on the Korean Peninsula from the solar radiation data provided by the UASIBS-KIER model of KIER with the solar radiation forecast data generated by the GFS model of the NCEP. In this study, the spatial and temporal downscaling of the GFS model, a cumulative solar radiation forecast model per 3 hours units, was conducted to measure forecast performance. First, in the case of spatial downscaling, K-means clustering of solar radiation values corresponding to all latitude and longitude values of the Korean Peninsula was performed, and seven clusters were formed. Second, in the case of temporal downscaling, the cumulative solar radiation forecast value over 3 h was converted to the cumulative solar radiation value over 1 h. To measure the downscaling performance, the downscaling results of the cumulative forecast over 1 h of the GFS model were compared with the observed solar radiation values of the UASIBS-KIER model. We performed downscaling by using the same average ratio as the cumulative value over 3 h. In addition, we performed downscaling by using the correlation between the cumulative value over 1 h by the UASIBS-KIER model. Based on the rRMSE and rMAE, the use of the correlation was found to be more efficient that the use of the same average ratio.