Multi-step wind speed forecasting based on a hybrid decomposition technique and an improved back-propagation neural...

Area of Science:

• Renewable Energy Systems
• Computational Intelligence
• Time Series Analysis

Background:

• Accurate wind speed forecasting (WSF) is crucial for stable power grid operation and the economic viability of wind energy.
• Existing forecasting methods often struggle with the inherent complexity and volatility of wind speed data.

Purpose of the Study:

• To propose a novel hybrid prediction model for enhanced wind speed forecasting (WSF).
• To evaluate the effectiveness of a secondary decomposition algorithm (SDA) in improving WSF accuracy and robustness.

Main Methods:

• A hybrid model combining Wavelet Transform (WT), Symplectic Geometry Mode Decomposition (SGMD), and Marine Predators Algorithm-optimized Back-Propagation Neural Network (BPNN).
• The proposed Secondary Decomposition Algorithm (SDA) involves a two-stage decomposition process: initial decomposition by WT, followed by SGMD on detailed components.
• Prediction of decomposed subsequences using the MPA-optimized BPNN.

Main Results:

• The hybrid WT-SGMD-MPA-BP model demonstrated superior prediction accuracy and robustness compared to existing models across 1-4 step predictions.
• The SDA significantly reduced the difficulty of WSF, yielding substantial improvements in Mean Absolute Percentage Error (MAPE).
• Re-decomposition of WT details via SGMD further enhanced prediction accuracy, notably reducing Root Mean Square Error (RMSE).

Conclusions:

• The proposed hybrid model with SDA offers a valuable advancement for wind speed forecasting.
• The secondary decomposition strategy effectively improves the performance of WSF models.
• Further research could explore the model's applicability to other nonlinear time series forecasting challenges.