The assessment of extreme wind speeds is a crucial issue for securing structural safety of wind turbines and inquiring largest loads to which turbines must be prepared to undergo. International standards suggest applying the Gumbel method of fitting the annual maxima to their theoretic probability distribution. Yet, often, wind databases are too short to apply such methods with statistical significance, and other procedures are commonly adopted [such as peaks over threshold (POT) and independent storms], which involve dependency on arbitrary thresholds for filtering data and issues of sub-asymptocity, i.e. how well the selected dataset fits to density functions describing the distribution of peaks or extreme values. The present paper aims at contributing to such currently ongoing debate, providing a statistical analysis of the application of POT and independent storms methods on wind time series of various lengths from different geographical areas. The CERN data analysis framework ROOT has been employed for guaranteeing excellent standards of computational precision and wealth of statistical information. Analysis of uncertainties in the wind speeds estimates and tests of the goodness of fit of the datasets to the proper distributions have been carried on. An algorithm for choosing the optimum thresholds was developed, which encapsules and compromises the statistical complexity of the methods. A declustering procedure has been carried on for discriminating proper peaks in the POT method: it has been tested that such declustering provides a dramatic improvement of the statistical quality of the method.

Applied statistics for extreme wind estimate

CASTELLANI, Francesco
;
ASTOLFI, DAVIDE
2015

Abstract

The assessment of extreme wind speeds is a crucial issue for securing structural safety of wind turbines and inquiring largest loads to which turbines must be prepared to undergo. International standards suggest applying the Gumbel method of fitting the annual maxima to their theoretic probability distribution. Yet, often, wind databases are too short to apply such methods with statistical significance, and other procedures are commonly adopted [such as peaks over threshold (POT) and independent storms], which involve dependency on arbitrary thresholds for filtering data and issues of sub-asymptocity, i.e. how well the selected dataset fits to density functions describing the distribution of peaks or extreme values. The present paper aims at contributing to such currently ongoing debate, providing a statistical analysis of the application of POT and independent storms methods on wind time series of various lengths from different geographical areas. The CERN data analysis framework ROOT has been employed for guaranteeing excellent standards of computational precision and wealth of statistical information. Analysis of uncertainties in the wind speeds estimates and tests of the goodness of fit of the datasets to the proper distributions have been carried on. An algorithm for choosing the optimum thresholds was developed, which encapsules and compromises the statistical complexity of the methods. A declustering procedure has been carried on for discriminating proper peaks in the POT method: it has been tested that such declustering provides a dramatic improvement of the statistical quality of the method.
2015
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11391/1213678
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 4
  • ???jsp.display-item.citation.isi??? 3
social impact