A widely applicable and robust LightGBM - Artificial neural network forecasting model for short-term wind power...

Area of Science:

• Renewable Energy Systems
• Machine Learning Applications in Power Grids

Background:

• Wind energy is a crucial clean energy source, but its intermittent nature poses challenges for grid stability.
• Accurate wind power generation prediction is essential for mitigating fluctuations and ensuring reliable power supply.
• Existing prediction models often neglect the nonlinear relationship between wind speed and power output and are limited by data scope.

Purpose of the Study:

• To develop a robust and universal wind power density prediction model.
• To address the limitations of existing models by incorporating nonlinear dynamics and diverse data sources.
• To improve the accuracy and reliability of wind power forecasting for grid integration and electricity trading.

Main Methods:

• A hybrid model combining LightGBM for feature extraction and artificial neural networks for prediction was developed.
• The model utilizes a data collection process that does not require meteorological measurement equipment, ensuring broad applicability.
• Model performance was validated using data from six diverse terrains spanning from 2020 to 2022.

Main Results:

• The developed model demonstrated high accuracy, with average prediction errors below 2% for 71.68% of cases and below 6% for 82.188% of cases.
• The model achieved an average R-squared value of 0.9755 and an average correlation coefficient of 0.9875.
• The results indicate superior performance and robustness across different terrains and time periods.

Conclusions:

• The hybrid LightGBM and artificial neural network model effectively captures the nonlinear relationship between wind speed and power generation.
• The model's universality, stability, and robustness make it a practical tool for real-world wind power forecasting.
• The validated accuracy and performance provide strong evidence for the model's utility in guiding electricity trade and enhancing grid management.