Servizio Comunitario di Informazione in materia di Ricerca e Sviluppo - CORDIS

Wind power prediction tool in central despatch centres

By designing a tool to predict wind power generation across a large electricity network in Denmark, the project has shown that it is possible to achieve large savings in operating costs. The tool has important implications for the efficient operation of such integrated networks.

A major part of the research was spent finding the best mathematical model for on-line predictions of wind power. The consortium then developed a suitable model that was tested in off-line scenarios. The wind power prediction tool (WPPT) was then tested on a large electricity network in which up to 400 MW were generated by wind power in January 1995. In order to make real-time predictions of total wind power for the network, wind speeds were recorded every five minutes, from seven wind farms located across the region, and fed into an autoregressive computer model. Such a model continually determines the most likely outcome of a range of incoming data and learns to recognise consistent patterns in the data. The WPPT provided on-line calculations and predictions of wind power production for a large network. The autoregressive model was able to make forecasts with a time horizon of between 0.5 and 12 h, and estimate the margin of error within the calculation. Although the system has the potential to make predictions up to 36 h ahead, without incorporating meteorological forecasts this is not yet possible. Nonetheless, the current capability of the WPPT makes it far superior to neural networks, which are limited to a small forecast horizon of between 0.5 and 3 h . The on-line system has performed well since November 1994. Indeed, the uncertainty of the prediction errors was soon found to decline due to the self-learning ability of the system.

Reported by

ELSAM system
Fjordvejen 1-11
7000 Fredericia
See on map