Hybrid Wind Power Forecasting Model Improves Grid Stability and Prediction Accuracy

Researchers at Nanchang University in China have developed a hybrid wind power forecasting model that combines Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) and ModernTCN-Informer to enhance the accuracy and reliability of wind power predictions. This model addresses the challenges posed by the randomness and volatility of wind power generation, allowing for more precise predictions and improved grid stability.

Key Takeaways:

• The hybrid model reduces the Mean Absolute Error (MAE) of 12-step predictions by 3.1% compared to the Informer model and by 29.5% compared to the LSTM model.
• The MAE is further reduced by 64.1% after incorporating ICEEMDAN, and the R2 value reaches 0.969.
• The proposed model is more accurate in multi-step predictions compared to the comparison models.
• The hybrid model captures correlations among univariate patch sequences across multiple time steps, long-term dependencies within univariate patch sequences, and latent correlations across variables.
• The model effectively mitigates data fluctuations by obtaining several relatively stable subsequences through ICEEMDAN decomposition.

Statistics:

• 3.1% reduction in MAE compared to the Informer model.
• 29.5% reduction in MAE compared to the LSTM model.
• 64.1% reduction in MAE after incorporating ICEEMDAN.
• R2 value reaches 0.969.
• Multi-step prediction accuracy is superior to that of the comparison models.

Sources:

• Wind Power Prediction Based on a Hybrid Model of ICEEMDAN and ModernTCN-Informer. IEEE Access, 2025,13():145256-145270.
• IEEE Access - http://ieeexplore.ieee.org/servlet/opac?punumber=6287639.
• DOI: 10.1109/ACCESS.2025.3599609.
• Nanchang University, School of Information Engineering, Nanchang, People's Republic of China.
• Jun He, Zijian Cheng, Zijie Zhong, Lizhuo Liang, Jianhui Ye.