A hybrid prediction model for forecasting wind energy resources.

Area of Science:

• Energy Science
• Artificial Intelligence
• Environmental Engineering

Background:

• Wind energy is crucial for global energy transition due to its clean and renewable nature.
• The inherent volatility, randomness, and uncertainty of wind speed limit wind power productivity.
• Accurate wind speed prediction is essential for optimizing wind farm operations and energy integration.

Purpose of the Study:

• To develop a hybrid prediction model for enhanced wind speed forecasting.
• To improve the productivity and reliability of wind power generation.
• To address the challenges of non-stationary and stochastic wind speed data.

Main Methods:

• Applying Variational Mode Decomposition (VMD) to decompose non-stationary wind speed data into intrinsic mode functions (IMFs).
• Utilizing sample entropy to determine the optimal number of decomposition components (K).
• Employing Elman neural networks for trend components and Radial Basis Function (RBF) networks for stochastic components.
• Integrating Lorenz disturbance to refine predictions by accounting for atmospheric effects.

Main Results:

• VMD effectively decomposes wind speed data, avoiding modal aliasing issues common in other methods.
• The hybrid Elman-RBF approach for different IMFs yields superior prediction accuracy compared to single models.
• The proposed hybrid model demonstrates higher accuracy than other neural network models in real-world wind farm data tests.

Conclusions:

• The hybrid VMD-Elman-RBF model with Lorenz disturbance significantly improves wind speed prediction accuracy.
• This enhanced prediction capability aids in formulating effective wind farm control strategies.
• The research contributes to optimizing wind farm self-regulation and advancing global energy innovation.