Optimizing solar power efficiency in smart grids using hybrid machine learning models for accurate energy generation...

Area of Science:

• Energy Systems Engineering
• Artificial Intelligence
• Renewable Energy Technologies

Background:

• The fourth energy revolution integrates renewable energy into intelligent, or smart, grids.
• Variable renewable energy output, dependent on weather, poses integration challenges.
• Smart grids leverage AI and real-time data for optimized energy production and distribution.

Purpose of the Study:

• To enhance solar power generation efficiency within smart grids.
• To evaluate hybrid machine learning models for predicting solar plant performance.

Main Methods:

• Developed and tested hybrid machine learning models: Hybrid Convolutional-Recurrence Net (HCRN), Hybrid Convolutional-LSTM Net (HCLN), and Hybrid Convolutional-GRU Net (HCGRN).
• Utilized solar plant data including power production (MWh), plane of array irradiance (POA), and performance ratio (PR).

Main Results:

• The Hybrid Convolutional-LSTM Net (HCLN) model achieved superior accuracy.
• HCLN demonstrated the lowest Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) for MWh, POA, and PR predictions.
• Specific RMSE values: 0.012027 (MWh), 0.013734 (POA), 0.003055 (PR). Specific MAE values: 0.069523 (MWh), 0.082813 (POA), 0.042815 (PR).

Conclusions:

• Proposed hybrid machine learning models effectively improve solar power generation system efficiency.
• Accurate prediction of key solar plant measurements is achievable with these models.
• This research supports the integration of renewables into smart grids.