Enhancing wind power forecasting accuracy through LSTM with adaptive wind speed calibration (C-LSTM)

Area of Science:

• Renewable Energy Systems
• Artificial Intelligence in Energy
• Climate Change Mitigation

Background:

• Wind power is crucial for carbon neutrality and requires accurate forecasting.
• Deep learning models like LSTM advance wind power prediction.
• Inaccurate forecasted wind speed limits the reliability of current wind power forecasting.

Purpose of the Study:

• To develop an advanced model for accurate wind power forecasting.
• To address the challenge of unreliable forecasted wind speed in existing models.
• To improve the efficiency of wind energy exploitation.

Main Methods:

• Proposed a novel model: LSTM with Adaptive Wind Speed Calibration (C-LSTM).
• C-LSTM integrates a mechanism for autonomous wind speed calibration during training and inference.
• Fuses historical and forecasted wind speed using adaptive weighting parameters and concurrent parameter updating.

Main Results:

• C-LSTM significantly outperforms standard LSTM in Mean Squared Error (MSE) and accuracy.
• Demonstrated improved performance across 25 distinct wind turbines.
• The adaptive wind speed calibration technique effectively coordinates discrepancies between forecasted and actual wind speeds.

Conclusions:

• C-LSTM enhances the accuracy and reliability of wind power forecasting.
• Adaptive wind speed calibration is an effective strategy for improving deep learning-based wind power prediction.
• The proposed method contributes to more efficient utilization of wind energy resources.