Seasonal Analysis and Prediction of Wind Energy Using Random Forests and ARX Model Structures

Abstract

To effectively utilize wind energy, many learning-based autoregressive models have been proposed in the literature. Improving their short-term prediction accuracy, however, is difficult, which mainly result from the stochastic nature of wind. Moreover, the incorporation of seasonal effects to improve their accuracies has not been considered, as most reported studies only relied on relatively short data sets. This brief examines meteorological data that were recorded over a six-year period and contrast various model structures and identification methods proposed in the literature. One focus of this brief is the prediction of wind speed and direction, which has not been extensively studied in the literature but is important for grid management. The reported results highlight that an increase in prediction accuracy can be obtained: 1) by incorporating seasonal effects into the model; 2) by including routinely measured variables, such as radiation and pressure; and 3) by separately predicting wind speed and direction.