Probabilistic maximum-value wind prediction for offshore environments

Andrea Staid, Pierre Pinson, Seth D. Guikema

Research output: Contribution to journalJournal articleResearchpeer-review

Abstract

High wind speeds can pose a great risk to structures and operations conducted in offshore environments. When forecasting wind speeds, most models focus on the average wind speeds over a given period, but this value alone represents only a small part of the true wind conditions. We present statistical models to predict the full distribution of the maximum-value wind speeds in a 3 h interval. We take a detailed look at the performance of linear models, generalized additive models and multivariate adaptive regression splines models using meteorological covariates such as gust speed, wind speed, convective available potential energy, Charnock, mean sea-level pressure and temperature, as given by the European Center for Medium-Range Weather Forecasts forecasts. The models are trained to predict the mean value of maximum wind speed, and the residuals from training the models are used to develop the full probabilistic distribution of maximum wind speed. Knowledge of the maximum wind speed for an offshore location within a given period can inform decision-making regarding turbine operations, planned maintenance operations and power grid scheduling in order to improve safety and reliability, and probabilistic forecasts result in greater value to the end-user. The models outperform traditional baseline forecast methods and achieve low predictive errors on the order of 1–2 m s−1. We show the results of their predictive accuracy for different lead times and different training methodologies.
Original languageEnglish
JournalWind Energy
Volume18
Issue number10
Pages (from-to)1725–1738
ISSN1095-4244
DOIs
Publication statusPublished - 2015

Keywords

  • Probabilistic prediction models
  • Maximum value winds
  • Offshore wind

Fingerprint

Dive into the research topics of 'Probabilistic maximum-value wind prediction for offshore environments'. Together they form a unique fingerprint.

Cite this