.png){kind=icon}
.png){kind=icon}
Return
Double-layer rolling decomposition and bidirectional temporal convolutional network for wind speed forecasting
.png){kind=icon}
PRE
AI
DOI：10.1080/15435075.2026.2675003.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Accurate and stable wind speed forecasting is fundamental for effective green energy management. This study proposes an enhanced anti-information leakage wind speed prediction framework that integrates a double-layer rolling decomposition method, the proposed new neural network model, and an intelligent optimizer. To address future information leakage caused by utilizing signal decomposition algorithms on random noise, this study proposes a double-layer rolling decomposition method to dynamically decompose and reconstruct the sequences from the data preprocessing phase and error correction phase, thereby avoiding information leakage and enhancing sequence learnability. This study proposes a novel hybrid neural network to extract forward and backward features from preprocessed wind speed sequences, thereby improving forecasting accuracy and robustness, which combines a bidirectional temporal convolutional network, attention mechanism, and intelligent optimizer. Comparative experiments are conducted on four wind speed datasets from China. The proposed wind speed prediction system achieved mean absolute percentage error value of approximately 1.5% and the prediction interval coverage probability exceeding 94%. Furthermore, this study verifies the effectiveness of multi-scale rolling decomposition integration with different rolling window lengths in improving prediction accuracy. The proposed system can provide valuable technical support for the development of wind power industry and the achievement of carbon neutrality goals.
Keywords:
Bidirectional temporal convolutional network
enhanced anti-information leakage framework
intelligent optimizer
wind speed prediction
Journal
IF:
3.1
Papers:
2.4K
Citations:
3.9K