Adaptive singular spectral decomposition hybrid framework with quadratic error correction for wind power prediction

Area of Science:

• Renewable Energy Systems
• Artificial Intelligence in Energy
• Signal Processing for Power Systems

Background:

• Accurate wind power forecasting is crucial for grid stability and efficient renewable energy integration.
• The inherent nonlinear and stochastic nature of wind data presents significant prediction challenges.
• Existing forecasting models often struggle with noise reduction and hyperparameter optimization.

Purpose of the Study:

• To develop a robust and accurate hybrid model for high-precision wind power forecasting.
• To address the challenges posed by wind data's complexity and improve prediction reliability.
• To enhance the integration of wind energy into power grids through advanced forecasting techniques.

Main Methods:

• Adaptive Improved Singular Spectrum Analysis (ISSA) for parameter-free noise reduction and modal decomposition.
• Optimized Bidirectional Temporal Convolutional Network-Bidirectional Long Short-Term Memory (BiTCN-BiLSTM) networks.
• AdaBoost ensemble learning for dynamic error correction and enhanced robustness.
• Hybrid strategy-enhanced dung beetle optimization (OTDBO) for hyperparameter tuning.

Main Results:

• Significant performance improvements demonstrated on seasonal datasets from Dabancheng wind farm.
• Mean Absolute Error (MAE) reduced by 45.4% and Root-Mean-Square Error (RMSE) by 47.6% (p < 0.001).
• Training time was reduced by 12.1%-21.3%, indicating computational efficiency.

Conclusions:

• The proposed hybrid model offers a substantial advancement in wind power forecasting accuracy and robustness.
• This method provides a scalable solution for reliable renewable energy integration into power grids.
• The combination of ISSA, BiTCN-BiLSTM, and AdaBoost effectively handles complex wind data characteristics.