A Combination of HARMONIE Short Time Direct Normal Irradiance Forecasts and Machine Learning: The #hashtdim Procedure.
- Resource Type
- Article
- Authors
- Gastón, Martín; Fernández-Peruchena, Carlos; Körnich, Heiner; Landelius, Tomas
- Source
- AIP Conference Proceedings. 2017, Vol. 1850 Issue 1, p1-8. 8p. 1 Diagram, 2 Charts, 3 Graphs, 2 Maps.
- Subject
- *IRRADIATION
*MACHINE learning
*NUMERICAL weather forecasting
*SOLAR technology
*SOLAR energy
- Language
- ISSN
- 0094-243X
The present work describes the first approach of a new procedure to forecast Direct Normal Irradiance (DNI): the #hashtdim that treats to combine ground information and Numerical Weather Predictions. The system is centered in generate predictions for the very short time. It combines the outputs from the Numerical Weather Prediction Model HARMONIE with an adaptive methodology based on Machine Learning. The DNI predictions are generated with 15-minute and hourly temporal resolutions and presents 3-hourly updates. Each update offers forecasts to the next 12 hours, the first nine hours are generated with 15-minute temporal resolution meanwhile the last three hours present hourly temporal resolution. The system is proved over a Spanish emplacement with BSRN operative station in south of Spain (PSA station). The #hashtdim has been implemented in the framework of the Direct Normal Irradiance Nowcasting methods for optimized operation of concentrating solar technologies (DNICast) project, under the European Union's Seventh Programme for research, technological development and demonstration framework. [ABSTRACT FROM AUTHOR]