Authors: Robin Girard*, Arthur Bossavy, George Kariniotakis
*MINES ParisTech, Center for Energy and Processes, B.P. No 207, F-06904, Sophia Antipolis, France. Tel: +33 (0) 4 93 67 8964 mail: robin 'dot' girard 'at' mines-paristech.fr
Highlight results
- Ramp forecast with ECMWF Ensemble prediction system
- New operational tools for reliable temporal prediction intervals
Background :
Today, there is a growing interest in developing short-term wind power forecasting tools able to provide reliable information about particular, so-called ”extreme” situations. One of them is the large and sharp variation of the production a wind farm can experience within a few hours called ramp event. Developing forecast information specially dedicated to ramps is of primary interest both because of the difficulties usual models have to predict them, and the potential risk they represent in the management of a power system.
Most existing wind power forecasting models are designed to provide point forecasts of expected future wind farm production. In recent years, research work has focused on associating uncertainty estimation with this type of point forecast.
Ramp forecast and temporal intervals:
Because forecasting the probability density function is not suitable for describing a future event that might not occur we propose to forecast ramps on the basis of a specific timing, plus the probability of observing a ramp within a set of time intervals (hereafter referred to as “prediction intervals”) centred on that timing. Thus, our approach aims at providing a suitable uncertainty estimation associated with the forecast of a ramp timing.
Figure 1:Temporal confidence intervals have different probabilities depending on the weather situation summarized by ensemble predictions.
Use of ECMWF ensemble forecast:
The proposed methodology relies on numerical weather prediction ensembles. Ensembles are alternative numerical weather predictions that are produced by perturbing the initial conditions, or a different parameterization of a numerical weather prediction model. The unperturbed prediction is referred to as control forecast and, in general, provides the most accurate wind production forecasts.
Prototype implementation :
The proposed methodology has been implemented on the ANEMOS platform and is running operationally for several TSO (e.g. PPC, RTE)
The forecast intervals and probabilities are evaluated according to several criteria including in particular:
- Reliability, to show that the forecast probabilities correspond to the observed frequencies
- Sharpness, to show that the forecast probability are condition specific and hence improve climatology.
Figure 2: Forecast probabilities are shown to be reliable, i.e. They correspond to observed frequencies
Further work:
Further ongoing work will include a systematic evaluation of the method under different climatic condition as well as well as a probabilistic optimisation tool allowing ramp forecast temporal intervals to be included into a quantitative decision process.
Bibliography:
[1] Bossavy, A. Girard, R. Kariniotakis, G.- Forecasting ramps of wind power production with numerical weather prediction ensembles. Wind Energ. (2012)
[2] Bossavy, A. Girard, R. Kariniotakis, G.Forecasting wind power uncertainty related toramp events In the scientific Procd of the European Wind energy Conference (EWEC), Warsaw,Poland, April 20-23 2010
[3] R. Girard, A. Bossavy and G. Kariniotakis. Forecasting ramps of wind power production atdifferent time scales, In Procd. of the 2011 European Wind energy Association Annual Event (EWEA 2011), Brussels, Belgium March 14-17, 2011
