Short-Term Wind Speed Predicting Based on Legendre Multiwavelet Transform and GRU-ENN

• Published in: IEEE access 2025, Vol.13, p.4381-4397
• Main Authors: Zheng, Xiaoyang, Luo, Xiaoheng, Liu, Dezhi
• Format: Article
• Language: English
• Subjects: Accuracy; Autoregressive processes; discrete wavelet transform; Elman neural network; empirical mode decomposition; gated recurrent unit; Legendre multiwavelet transform; Neural networks; Prediction algorithms; Predictions; Predictive models; Stability; Transforms; Wind power; Wind speed
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2024.3435674

Wind energy has become a vital component of the power system. Due to the stochastic and intermittent characteristics of wind speed, thus enhancing accuracy and stability in short-term wind speed prediction is imperative and important for effectively harnessing wind energy. This paper proposes a novel hybrid model combing Legendre multiwavelet transform, Gated recurrent unit and Elman neural network (LMWT-GRU-ENN) for short-term wind speed prediction. More precisely, the rich properties, especial various regularities of LMW bases are utilized to effectively match the non-linearity and larger non-stationary features of short-term wind speed corresponding to multi-resolution level and multi-wavelet bases. GRU model is used to predict the low frequency components, and ENN model is implemented to predict the high frequency components obtained by LMWT, which can effectively improve the prediction performance by thoroughly making use of their advantages. Finally, massive experiments are conducted on two short-term wind speed datasets, and the experimental results demonstrate the proposed method attains the excellent performance of in both accuracy and stability compared with other state-of-the-art methods.