301. Probabilistic prediction of direct normal irradiance derived from global horizontal irradiance over the Korean Peninsula by using Monte-Carlo simulation.
- Author
-
Kim, Chang Ki, Kim, Hyun-Goo, Kang, Yong-Heack, Yun, Chang-Yeol, and Kim, Shin Young
- Subjects
- *
PROBABILISTIC databases , *MONTE Carlo method , *SPECTRAL irradiance , *MEAN square algorithms , *SOLAR power plants - Abstract
Highlights • Evaluation of Engerer Model for derving DNI from GHI at minute scale in the Korean Peninsula. • Producing the prediction interval of DNI estimates by using Monte-Carlo simulation. • Comparison of Monte-Carlo simulation with multi-model ensembles and bootstrap technique. Abstract Solar resource assessment is carried out in a feasibility study using reliable meteorological elements including solar irradiance. In concentrating solar power plants, the direct normal irradiance is the key variable in the system operation. However, direct normal irradiance is rarely measured as compared to global horizontal irradiance. There are several models that can be used to derive the direct normal irradiance from global horizontal irradiance. In this study, the Engerer model is used as a decomposition model, then evaluated against in situ observations at three ground stations: Seoul, Buan, and Jeju ground stations. The relative root mean square errors between the observed and direct normal irradiance estimated by the Engerer model are 15.0%, 19.4%, and 17.1% at Seoul, Buan, and Jeju ground stations, respectively. The uncertainty of estimates is represented by the prediction interval from probabilistic prediction through Monte-Carlo simulation that employs the bias between estimation and ground truth for training datasets. The prediction interval for 90% confidence level is 117.9 W m−2 at the Seoul station, resulting from Monte-Carlo simulation. The prediction interval coverage probability is 92.8%, implying that the probability that observed DNI is not included in the prediction interval is 7.2%. The error metrics for probabilistic prediction indicates that Monte-Carlo simulation provides both valid and more informative estimations. [ABSTRACT FROM AUTHOR]
- Published
- 2019
- Full Text
- View/download PDF