An integrated method with adaptive decomposition and machine learning for renewable energy power generation...

Area of Science:

• Energy Science
• Environmental Science
• Data Science

Background:

• Traditional energy sources negatively impact the environment, necessitating a shift to renewables.
• Renewable energy integration, like wind and photovoltaics, faces challenges due to generation randomness.
• Accurate short-term forecasting is vital for managing renewable energy uncertainty and ensuring grid security.

Purpose of the Study:

• To develop and validate an integrated forecasting system for renewable energy sources.
• To address the challenges posed by the inherent randomness of renewable energy generation.
• To enhance the accuracy and operational efficiency of ultra-short-term renewable energy forecasting.

Main Methods:

• Ensemble empirical mode decomposition for data preprocessing and modal identification.
• Support vector regression optimized by sparrow search algorithm for time series forecasting.
• Combination of statistical methods tailored to different data series characteristics.

Main Results:

• The proposed integrated system significantly improves accuracy in ultra-short-term renewable energy forecasting.
• Experimental validation confirms the model's effectiveness across various forecasting scenarios and data characteristics.
• The method demonstrates strong applicability and competitiveness in real-world renewable energy prediction.

Conclusions:

• The developed forecasting system provides a robust technical basis for effective renewable energy utilization.
• The integrated approach successfully mitigates risks associated with renewable energy generation variability.
• The system meets operational efficiency and accuracy requirements for practical renewable energy management.