Title: Application of EMD-Adaboost in wind speed prediction

Authors: Jun-Qi Yang; Hai-Zhong Liu

Addresses: School of Mathematics and Physics, Lanzhou Jiaotong University, Lanzhou, Gansu Province, 730000, China ' School of Mathematics and Physics, Lanzhou Jiaotong University, Lanzhou, Gansu Province, 730000, China

Abstract: Wind speed in advance can provide decision support for a wind farm operation. This paper attempts to propose an improved artificial neural network algorithm based on empirical mode decomposition combined with the ensemble learning model Adaboost to improve and optimise the wind speed prediction method. In the prediction process, a new hidden layer node selection method is proposed. After using empirical mode decomposition to obtain new input data, considering the relationship between each component and output after decomposition, a single hidden layer node method is adopted, which is confirmed by experiments. The root mean square error (RMSE) of the final forecasting model under two different volatility conditions (unstable volatility and stable volatility) reached 0.46 and 0.2809, and the R-squared was 0.84 and 0.96. These indicators are the optimal level in a comparison model, and are statistically significant (R-squared at the 0.05 significance level).

Keywords: new energy; wind prediction; machine learning; nonlinear prediction; empirical mode decomposition; ensemble learning; hidden layer; Adaboost; artificial neural networks; root mean square error.

DOI: 10.1504/IJDS.2022.126854

International Journal of Data Science, 2022 Vol.7 No.2, pp.164 - 180

Received: 29 Sep 2021
Received in revised form: 20 Dec 2021
Accepted: 18 Apr 2022
Published online: 09 Nov 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article