 |
 |
Conference papers
The Solar Forecast Similarity Method: a new method to compute solar radiation forecasts for the next day
Abstract : The need for PV plant owners to plan what they are injecting in the electricity grid is more and more stringent to avoid endangering the whole supply in electricity. A new solar forecast algorithm, named Solar Forecast Similarity Method, has been developed to predict irradiance for the next day based on a statistical study of the long term HelioClim-3 irradiation database. This algorithm searches in the past for the most similar days compared to the day of interest and uses their following days to produce a forecast. The model has been optimized against the database itself to compute the most adequate set of parameters over France and for the month of January 2014. With this configuration, the results are a null bias and a root mean square error of 48%. The algorithm outperforms the persistence by 20% and the error is similar to existing methods. An objective validation has then been carried out to compare the irradiance forecasts to high quality measurements from several Baseline Surface Radiation Network (BSRN) ground stations. The method is very promising since the comparison results are in line or lower than the one obtained with the first validation analysis performed on the HelioClim-3 database. For high frequencies, however, predictions have a high error for rapidly varying weather. This demonstrates that the method provides information for the averaged production the following day but requires another input to reliably predict high frequency irradiance.
Cited literature [9 references]
https://hal-mines-paristech.archives-ouvertes.fr/hal-01353949
Contributor : Lucien Wald <>
Submitted on : Tuesday, August 16, 2016 - 1:53:56 PM
Last modification on : Wednesday, October 14, 2020 - 4:02:36 AM
Long-term archiving on: : Thursday, November 17, 2016 - 10:50:08 AM
Contributor : Lucien Wald <>
Submitted on : Tuesday, August 16, 2016 - 1:53:56 PM
Last modification on : Wednesday, October 14, 2020 - 4:02:36 AM
Long-term archiving on: : Thursday, November 17, 2016 - 10:50:08 AM
File
2016_shc_forecast_boilley.pdf
Publication funded by an institution