Analog model for extrem wind and power forecasting

Authors: Pascual Álvaro, Valero Francisco, Martín Maria Luis

Universidad Complutense de Madrid

  This e-mail address is being protected from spambots. You need JavaScript enabled to view it
 
Facultad de Físicas.
Depto. de Física de la Tierra, Astronomía y Astrofísica II.
Av. Complutense S/N 28040, Madrid, Spain.
Tel: +34 91 394 50 88
mail: a 'dot' depascual 'at' fis.ucm.es

Highlight results

An analog model performs better estimations than climatology and persistence for forecasting extreme wind and power production events.

Introduction:

Many types of meteorological extreme events are strongly associated with synoptic structures. The analysis and recognition of particular synoptic patterns can provide statistic information about the probability of extreme events. The analog method is a methodology based on comparison of an input synoptic situation with situations stored in a historic data base.

Methodology:

The analog method is a methodology based on analysis and recognition of synoptic patterns. A basic assumption accepted on the analogs downscaling methodology is the existence of a strong relationships between large-scale meteorological field patterns and local data. For the application of the methodology two historic data bases are needed. The first corresponds to the Large-Scale Data Base (LSDB) and the second one for the Local Data Base (LDB).

Searching analog patterns is performed by comparison of an analyzed synoptic situation with the atmospheric situations stored in the LSDB. Thus, the most similar large-scale situations are obtained. Once the algorithm have found one (or more) analog situations, relationships of the local field with them are supposed in order to obtain the local analog data of LDB.

Fig: Flowchart of the analog methodology applied in the ANPAF model.

The analog methodology has been applied and validated including different LDBs:

Daily mean wind in 32 stations on Spain.
Wind Gust in 32 stations on Spain.
Power Outputs in 443 wind farms on Denmark.
Power Outputs of 15 aggregated areas on Denmark.
Power Outputs of 4 control areas on Germany.
Power Outputs of 105 wind farms on Ireland.

In the validation process of the model, many aspects have been considered analyzing the sensitivity of the model forecasting and taking into account:

Uncertainties associated to the number of analogs included in the forecasting.
Improvement of the forecasting associated with the historic records length.
Analysis of results applying deterministic point of view (estimation of BIAS, RMSE and correlations).
Analysis of results applying probabilistic point of view (estimation of BBS, rank histograms and reliability diagrams).

The probabilistic validation has been carried on evaluating the skills of the model in forecasting diverse thresholds from mean values to extreme events.The application of different thresholds in the model allows obtaining to the end user a probabilistic forecasting of a “customized” extreme event.

Results:

Figure: (left) BSS of wind power production forecasting over Denmark and Germany. Values have been evaluated for different thresholds associated with percentages of the days, from very low production to very high production events. (right) BSS of wind gust forecasting on Spain. Threshold corresponds with the estimation of the wind gust mean in each station.

The model has shown a very stable forecasting giving results that can improve up to 25% the reference models (climatology and persistence). This improvement is obtained for all thresholds applied, being small for very extreme events forecasting (events under the 10 percentile and over the 99 percentile). The probabilistic forecasting shows an adjusted spreading of estimations.

Perspectives:

There are two available lines in developing the methodology in future works:

The implementation of the model in real operative test cases.
Study of the calibration of a multivariable analogue methodology applying mathematical optimization algorithms.

Bibliography:

[1] Pascual A., Martín M.L., Valero F., Luna Y., Morata A. (2011): Wintertime connections between extreme wind patterns in Spain and large-scale geopotential height field. Atmospheric Research (submitted).

[2] Pascual A., Valero F., Martín M.L., Morata A., and Luna M.Y. (2012): Probabilistic and deterministic results of the ANPAF analog model for Spanish wind field estimations. Atmospheric Research, 108, 39-56.

[3] M. L. Martín, F. Valero, A. Pascual, A. Morata, and M. Y. Luna (2011): Springtime connections between the large-scale sea level pressure field and gust wind speed over Iberia. Natural Hazards and Earth Sciences, 11, 191-203.

[4] M.L. Martín, F. Valero, A. Morata, M.Y. Luna, A. Pascual, D. Santos-Muñoz (2011): Springtime coupled modes of regional wind in the Iberian Peninsula and large-scale variability patterns. International Journal of Climatology, 31(6), 880-895.

SafeWind