Electric load forecasting based on kernel extreme learning machine optimized by improved sparrow search algorithm.

Area of Science:

• Electrical Engineering
• Artificial Intelligence
• Computational Science

Background:

• Accurate electric load forecasting is crucial for power system efficiency and renewable energy integration.
• Existing methods face challenges in optimizing forecasting models and handling complex load patterns.

Purpose of the Study:

• To develop a novel and accurate electric load forecasting framework.
• To improve the optimization algorithm for enhanced decomposition and forecasting accuracy.

Main Methods:

• A multi-strategy improved Sparrow Search Algorithm (WHFSSA) was developed to enhance search capabilities.
• Variational Mode Decomposition (VMD) parameters were optimized using WHFSSA for precise load sequence decomposition.
• A forecasting model, WHFSSA-KELM, integrated decomposed subsequences with Kernel Extreme Learning Machine (KELM).

Main Results:

• The WHFSSA-KELM model demonstrated significant improvements on two electric load datasets.
• Dataset 1 showed a 5.7% average R2 improvement for temperature and historical load forecasting.
• Dataset 2 achieved average R2 improvements of 5.6%, 7.6%, and 10.9% for three-step-ahead forecasts.

Conclusions:

• The proposed WHFSSA-KELM framework provides more accurate electric load forecasts.
• The enhanced forecasting accuracy contributes to the safe and stable operation of power systems.
• The study highlights the effectiveness of the improved optimization algorithm and integrated forecasting approach.