A novel interval prediction method in wind speed based on deep learning and combination prediction.

Area of Science:

• Renewable Energy Systems
• Data Science
• Time Series Analysis

Background:

• Accurate wind speed forecasting is crucial for efficient wind turbine operation and power grid stability.
• Existing methods often struggle with the chaotic nature and noise inherent in wind speed data.
• Quantifying wind speed uncertainty is essential for reliable energy management.

Purpose of the Study:

• To introduce the Combined Method for Interval Forecasting (CMIF) for enhanced real-time wind speed uncertainty prediction.
• To improve the accuracy and reliability of wind speed interval forecasts.
• To facilitate better wind turbine operation and power grid dispatching.

Main Methods:

• Utilized time-varying filtering for empirical mode decomposition and phase space reconstruction to address chaotic phenomena and noise.
• Considered and selected high-performing statistical and machine learning models.
• Employed a multi-objective optimizer to combine selected models for the final prediction.

Main Results:

• CMIF demonstrated significant improvements in predicted wind speed interval accuracy, ranging from 1.07% to 55.37% compared to single models.
• The method achieved narrow prediction intervals while maintaining a high coverage rate.
• Experimental validation was performed using data from the Gansu wind tower.

Conclusions:

• CMIF effectively enhances the accuracy of wind speed interval forecasting.
• The method provides a reliable way to quantify wind speed uncertainty.
• CMIF supports improved operational efficiency for wind turbines and power grids.