Short-Term Wind Power Prediction Based on Encoder-Decoder Network and Multi-Point Focused Linear Attention Mechanism.

Area of Science:

• Renewable Energy Systems
• Artificial Intelligence in Power Engineering
• Grid Integration of Wind Power

Background:

• Wind energy is a clean but unpredictable power source, posing challenges for grid stability.
• Accurate short-term wind power generation forecasting is essential to mitigate grid integration risks.
• Existing models often struggle to capture complex temporal and multidimensional features of wind power data.

Purpose of the Study:

• To propose a novel composite model, MLL-MPFLA, for enhanced short-term wind power generation prediction.
• To improve the accuracy and reliability of wind power forecasting for better grid management.
• To validate the proposed model's performance against established forecasting techniques.

Main Methods:

• A composite model (MLL-MPFLA) integrating a multilayer perceptron (MLP) for feature extraction and an LSTM-based encoder-decoder network for temporal analysis.
• Utilization of a multi-point focused linear attention mechanism during the decoding phase to refine predictions.
• Comparative performance evaluation against MLP, LSTM, LSTM-Attention-LSTM, LSTM-Self_Attention-LSTM, and CNN-LSTM-Attention models.

Main Results:

• The MLL-MPFLA model demonstrated superior predictive performance across key metrics: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and R-squared (R2).
• The combination of MLP for multidimensional feature extraction and LSTM for temporal dependency exploration proved effective.
• The multi-point focused linear attention mechanism significantly contributed to enhanced prediction accuracy.

Conclusions:

• The proposed MLL-MPFLA model offers a significant advancement in short-term wind power generation forecasting.
• The model's ability to integrate multidimensional and temporal features leads to more accurate predictions, crucial for grid stability.
• MLL-MPFLA provides a robust solution for optimizing the integration of wind energy into power grids.