Evaluating the intrinsic predictability of wind speed time series via entropy-based approaches

Abstract

The intrinsic predictability of wind speed time series is pivotal for various wind engineering applications, such as optimizing wind energy resource assessments and enhancing the accuracy of forecasting models. This study employs entropy-based approaches, specifically Permutation Entropy (PermEn) and Sample Entropy (SampEn), to evaluate wind speed predictability across diverse conditions. We systematically investigate the variation of these entropy measures in relation to terrain complexity, mean wind speed, seasonal variation, sampling frequency, and window length. Our analysis reveals that terrain complexity significantly influences entropy values. We observed that a positive correlation between mean wind speed and entropy, where higher wind speeds are associated with increased predictability. Seasonal variations also demonstrate a clear impact on entropy measures. Such dependence, however, varies between different stations. Furthermore, the study highlights the sensitivity of entropy measures to sampling frequency and window length, indicating that higher sampling frequencies and longer window lengths result in larger values of PermEn, and lower values of SampEn.