Skip to Main content Skip to Navigation

Sizing of a short term wind forecasting system

Abstract : In a context of global warming and energy transition, the development of renewable energies is essential in order to ensure energy production that meets a constantly growing demand. French wind power producers benefit from a “obligation to purchase” from EDF for 15 years. After that, they have to sell their production in the competitive market. To do so, they must announce in advance the amount of energy they will inject into the grid. In case of imbalance, they are charged penalties. In France, the deadline for selling energy is 30 minutes. Thus, in this thesis, several downscaling approaches, parametric (linear regression) and non-parametric (random forests) are developed, calibrated and evaluated. The considered lead times range from 30 min to 3 h.The downscaling methods considered are rarely used for lead times lower than 1 h since numerical models are generally run every 6 to 12 hours. The use of in-situ measurements in downscaling methods to correct the numerical prediction at initialization, allows a significant performance gain. Compared to traditional statistical methods for short term forecasting, the improvement compared to the persistence method ranges from 1.5%, 10 min ahead, to more than 30%, 3 h ahead. In order to limit the accumulation of errors in the conversion from wind speed forecast to wind power forecast, an analysis of the error induced by different meteorological variables, such as wind direction or air density, is presented. First, the forecast at the farm scale is explored and then the spatial dimension is introduced. Finally, the economic value of such a short term forecasting model is explored. The different steps of the electricity market are studied and the different sources of uncertainty and variability, such as forecast errors and price volatility, are identified and assessed. For the two wind farms considered in this study, the results show that the short term forecasts allow an increase in annual income between 4 and 5%
Complete list of metadatas

Cited literature [116 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Friday, March 20, 2020 - 11:27:08 AM
Last modification on : Thursday, October 29, 2020 - 3:01:52 PM
Long-term archiving on: : Sunday, June 21, 2020 - 1:23:34 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02513065, version 1


Aurore Dupré. Sizing of a short term wind forecasting system. Geophysics [physics.geo-ph]. Institut Polytechnique de Paris, 2020. English. ⟨NNT : 2020IPPAX002⟩. ⟨tel-02513065⟩



Record views


Files downloads