Short-term wind speed prediction using hybrid machine learning techniques.

Area of Science:

• Renewable Energy Systems
• Machine Learning Applications
• Environmental Science

Background:

• Wind energy is a key renewable resource globally.
• Accurate wind speed prediction is essential for effective wind power estimation.
• Short-term forecasting is critical for grid integration and operational planning.

Purpose of the Study:

• To evaluate and compare the performance of various hybrid machine learning models for short-term wind speed prediction.
• To identify the most accurate and computationally efficient model for wind speed forecasting.
• To assess model applicability across different geographical locations in Tamil Nadu, India.

Main Methods:

• Employed hybrid machine learning models: Twin Support Vector Regression (TSVR), Primal Least Squares Twin Support Vector Regression (PLSTSVR), Iterative Lagrangian Twin Parametric Insensitive Support Vector Regression (ILTPISVR), Extreme Learning Machine (ELM), Random Vector Functional Link (RVFL), and Large-Margin Distribution Machine-based Regression (LDMR).
• Utilized short-term wind speed data from five stations in Tamil Nadu, India (Chennai, Coimbatore, Madurai, Salem, Tirunelveli).
• Performance evaluation using metrics: RMSE, MAPE, SMAPE, MASE, SSE/SST, SSR/SST, and R².

Main Results:

• Large-Margin Distribution Machine-based Regression (LDMR) demonstrated superior prediction accuracy compared to other models.
• Extreme Learning Machine (ELM) exhibited the fastest computational performance among the evaluated models.
• Model performance varied across different stations, highlighting the need for localized optimization.

Conclusions:

• LDMR is recommended for accurate short-term wind speed forecasting.
• ELM is a viable option when computational speed is a primary concern.
• Hybrid machine learning approaches offer significant potential for enhancing renewable energy management.