Wind forecasting using Principal Component Analysis

Christina Skittides, Wolf-Gerrit Fruh

Research output: Contribution to journalArticle

48 Citations (Scopus)
487 Downloads (Pure)

Abstract

We present a new statistical wind forecasting tool based on Principal Component Analysis (PCA), which is trained on past data to predict the wind speed using an ensemble of dynamically similar past events. At the same time the method provides a prediction of the likely forecasting error. The method is applied to Meteorological Office wind speed and direction data from a site in Edinburgh. For the training period, the years 2008–2009 were used, and the wind forecasting was tested for the data from 2010 for that site. Different parameter values were also used in the PCA analysis to explore the sensitivity analysis of the results.

The forecasting results demonstrated that the technique can be used to forecast the wind up to 24 h ahead with a consistent improvement over persistence for forecasting more than 10 h ahead. The comparison of the forecasting error with the uncertainty estimated from the error growth in the ensemble forecast showed that the forecasting error could be well predicted.
Original languageEnglish
Pages (from-to)365-374
JournalRenewable Energy
Volume69
Early online date22 Apr 2014
DOIs
Publication statusPublished - Sep 2014

Keywords

  • Wind energy resource;
  • Principal Component Analysis (PCA)
  • Forecasting

Fingerprint Dive into the research topics of 'Wind forecasting using Principal Component Analysis'. Together they form a unique fingerprint.

  • Cite this