A combined model for short-term wind speed forecasting based on empirical mode decomposition, feature selection,...

Area of Science:

• Renewable Energy Systems
• Time Series Analysis
• Machine Learning Applications

Background:

• Accurate short-term wind speed forecasting is crucial for wind power production planning and control.
• Traditional forecasting models struggle with the inherent non-linearity and non-stationarity of wind speed data.
• Developing advanced models is necessary to overcome these challenges and improve prediction accuracy.

Purpose of the Study:

• To develop an improved wind speed forecasting model.
• To enhance the prediction performance of wind speed for wind power applications.
• To address the limitations of traditional models in handling complex wind data.

Main Methods:

• The proposed model integrates Empirical Mode Decomposition (EMD) to decompose wind speed time series into Intrinsic Mode Functions (IMFs), reducing non-stationarity.
• Feature Selection (FS) and Support Vector Regression (SVR) are employed to predict high-frequency IMFs.
• Cross-validated Lasso (LassoCV) is utilized for predicting low-frequency IMFs and trends.

Main Results:

• The combined model was tested using data from two wind stations in Michigan, USA.
• Experimental results demonstrated superior prediction performance in multi-step wind speed forecasting compared to traditional models.
• The proposed model outperformed individual methods and existing EMD-based combined models.

Conclusions:

• The developed combined model effectively improves wind speed forecasting accuracy.
• This approach offers a robust solution for enhancing wind power production planning and control.
• The integration of EMD, FS, SVR, and LassoCV provides a powerful tool for complex time series prediction.