Short-term prediction of wind power with a clustering approach

Abstract

A clustering approach is presented for short-term prediction of power produced by a wind turbine at low wind speeds. Increased prediction accuracy of wind power to be produced at future time periods is often bounded by the prediction model complexity and computational time involved. In this paper, a trade-off between the two conflicting objectives is addressed. First, a set of the most relevant parameters (predictors) is selected using the underlying physics and pattern immersed in data. Five scenarios of the input space are created with the k-means clustering algorithm. The most promising clustering scenario is applied to produce a model for each clustered subspace. Computational results are compared and the benefits of cluster–specific (customized) models are discussed. The results show that the prediction accuracy is improved the input space is clustered and customized prediction models are developed.