A novel wind speed prediction model based on neural networks, wavelet transformation, mutual information, and coot...

Area of Science:

• Renewable Energy Systems
• Computational Intelligence
• Time Series Analysis

Background:

• Wind energy is a crucial renewable resource, but its intermittent nature poses integration challenges for power grids.
• Fluctuations in wind speed require accurate prediction models to ensure grid stability and reliable energy supply.

Purpose of the Study:

• To develop and evaluate a novel hybrid model for precise wind speed prediction.
• To address the complexities of wind speed variability for enhanced grid integration of wind power.

Main Methods:

• A hybrid model combining Discrete Wavelet Transform (DWT) for signal smoothing, Mutual Information (MI) for feature selection, Coot Optimization Algorithm (COOT) for optimal feature selection, and Bidirectional Long Short-Term Memory (BiLSTM) for pattern recognition.
• Performance evaluation using standard error metrics: Mean Squared Error (MSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Coefficient of Determination (R²), and Median Absolute Error (MedAE).

Main Results:

• The proposed hybrid model demonstrated high prediction accuracy on two distinct wind speed datasets.
• The model significantly outperformed 14 benchmark models across various performance metrics.
• The integration of DWT, MI, COOT, and BiLSTM effectively captured complex patterns in wind speed data.

Conclusions:

• The developed hybrid model offers a superior solution for wind speed prediction compared to existing methods.
• This advanced prediction capability can facilitate more effective integration of wind energy into power grids.
• The findings highlight the potential of combining signal processing, feature selection, and deep learning for renewable energy forecasting.