Implementation of hybrid wind speed prediction model based on different data mining and signal processing approaches

Area of Science:

• Meteorology and Climatology
• Artificial Intelligence in Environmental Science
• Renewable Energy Systems

Background:

• Accurate wind speed (WS) estimation is vital for optimizing power systems and water resource management.
• Existing methods often struggle with the complex, non-linear nature of wind speed data.
• Advanced forecasting techniques are needed to improve prediction accuracy for renewable energy planning.

Purpose of the Study:

• To enhance wind speed prediction accuracy by integrating artificial intelligence (AI) with signal decomposition methods.
• To evaluate the performance of various AI models combined with discrete wavelet transform (DWT) and empirical mode decomposition (EMD).
• To identify the optimal model structure and input parameters for reliable long-term wind speed forecasting.

Main Methods:

• Utilized feed-forward back propagation neural network (FFBNN), support vector machine (SVM), and Gaussian processes regression (GPR) models.
• Applied discrete wavelet transform (DWT) and empirical mode decomposition (EMD) for signal preprocessing.
• Evaluated model performance using statistical criteria (Willmott's index, MSE, R², Taylor diagram) for 1-month-ahead WS forecasting.

Main Results:

• Both DWT and EMD signal processing techniques demonstrably improved the prediction performance of standalone AI models.
• The hybrid EMD-Matern 5/2 kernel GPR model achieved the best performance, with R² values of 0.802 (test) and 0.606 (validation).
• Optimal model structure involved using input variables with a time delay of up to 3 months.

Conclusions:

• Hybrid models combining signal decomposition (EMD) and AI (GPR) offer superior wind speed forecasting capabilities.
• The findings provide valuable insights for practical applications, planning, and management in the wind energy sector.
• This research underscores the potential of advanced signal processing and AI for improving meteorological predictions.