Wind Power Prediction with Cross-Correlation Weighted Nearest Neighbors

Abstract

A precise wind power prediction is important for the integration of wind energy into the power grid. Besides numerical weather models for short-term predictions, there is a trend towards the development of statistical data-driven models that can outperform the classical forecast models [1]. In this paper, we improve a statistical prediction model proposed by Kramer and Gieseke [5], by employing a cross-correlation weighted k-nearest neighbor regression model (x-kNN). We demonstrate its superior performance by the comparison with the standard u-kNN method. Even if different pre-processing steps are considered, our regression technique achieves a comparably high accuracy.