Analyzing the correlation and predictability of wind speed series based on mutual information

Abstract

The stochastic nature of wind makes it difficult for wind speed forecast to achieve high accuracy. We suggest a mutual information analysis before wind speed forecast. This pre‐processing procedure reveals the correlation of wind speed data; therefore, helps to analyze the predictability of specific wind speed series and reduce the redundancy of forecast models, helping to select the most appropriate length of the input window of a neural network. To verify the feasibility of the proposed method, three kinds of time series and a forecast model are introduced as simulation studies. The results demonstrate that mutual information reflects the correlation and predictability of wind speed quantitatively. © 2018 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.