Characterization of Wind Farm Energy Production in a Zone by Artificial Neuronal Networks

Abstract

This paper studies the effect of a number of close wind farms, where there is a relationship among the farms. First of all, data are statistically analysed to check main relationships between farms.

The second analysis goes further in the characterisation of wind farm relationships. Artificial Neural Networks (ANN) provide a powerful tool to classify data automatically. They are very useful in finding patterns of productions of a zone. The patterns are obtained only with data of the power of wind farms through competitive neural networks and self-organizing feature maps (SOFM).

Time variations of wind farm output have been studied in the third analysis. Only hourly data is available, so only slow variations can be analysed (such as down and dusk variations). Two tools have been used in the time study: correlogram and spectrogram of each wind farm.

In addition to this, those patterns found by ANN have been compared in the forth analysis with meteorological data of a nearby station.