Spatio-temporal modelling of short-term wind power prediction errors (02004/FU5766 - Improved wind power prediction)

Julija Vlasova, Ewelina Kotwa, Henrik Aalborg Nielsen, Henrik Madsen

AbstractForecasts of wind power generation are more and more frequently used in various management tasks related to integration of wind power generation in power systems. The quality of the forecast is very important, and a reliable estimate of the uncertainty of the forecast is known to be essential. Today the forecasts of wind power generation are provided without a proper consideration to the spatio-temporal dependencies observed in the wind power generation field. State-of-art prediction systems, like Wind Power Prediction Tool (WPPT 1 ), typically provide forecast for a single wind farm. Predictions for a larger region with wind farms are then obtained by an upscale of the individual farm predictions. This means that the spatio-temporal relations are not adequately considered.

The aim of this work is to investigate weather it is possible to improve the wind power forecasting system WPPT, developed in Denmark, by examining the spatio-temporal correlation of the prediction errors. The paper is organized in the following way: Section 2 presents the data set used in the work. Then in Section 3 a pre-modelling analysis of the data structure is provided. The results of the correlation study are presented and discussed. Further, the modelling step is performed: three types of the models for WPPT error analysis are discussed: ARX, Threshold and Conditional Parametric Models. This takes place in Section 4. The description of each model consists of 4 parts Modeling, Estimation, Application and Results. The first two are theoretical: Modeling is a general description of the model; Estimation deals with how the parameters can be estimated. The last two parts show respectively how the model was applied to the data and the results obtained. Section 5 describes validation methods used in this study for checking the adequacy of the performance of the fitted models. The paper concludes in Section 6 with a small discussion on the general results and possible future work.
TypeTechnical report
PublisherInformatics and Mathematical Modelling, Technical University of Denmark, DTU
AddressRichard Petersens Plads, Building 321, DK-2800 Kgs. Lyngby
SeriesIMM-Technical Report-2007-18
Electronic version(s)[pdf]
BibTeX data [bibtex]
IMM Group(s)Mathematical Statistics