Machine learning-based energy management and power forecasting in grid-connected microgrids with multiple distributed...

Area of Science:

• Electrical Engineering
• Computer Science
• Renewable Energy Systems

Background:

• Integrating renewable energy sources into microgrids presents forecasting and management challenges.
• Accurate power generation prediction is crucial for grid stability and efficiency.

Purpose of the Study:

• To enhance microgrid efficiency and reliability using advanced machine learning for power generation forecasting.
• To evaluate the performance of Support Vector Regression (SVR) against traditional models.

Main Methods:

• Developed and applied a Support Vector Regression (SVR) model for power generation forecasting.
• Utilized historical energy production data, weather patterns, and grid conditions.
• Compared SVR model performance against linear regression models using error metrics.

Main Results:

• SVR model achieved significantly lower error metrics: MSE (2.002 solar, 3.059 wind), MAE (0.547 solar, 0.825 wind), RMSE (1.415 solar, 1.749 wind).
• Resulted in an 8.4% reduction in operating costs and a 12% increase in renewable energy utilization.
• Improved supply-demand balance by 10% and reduced peak load demand by 15%.

Conclusions:

• SVR model offers superior accuracy for renewable energy forecasting in microgrids.
• The approach enhances energy management, reduces costs, and improves grid stability.
• Machine learning, particularly SVR, shows significant potential for revolutionizing renewable energy integration and management.