Optimizing photovoltaic power plant forecasting with dynamic neural network structure refinement

Area of Science:

• Renewable Energy Systems
• Artificial Intelligence in Energy
• Photovoltaic Power Generation

Background:

• Efficient management of energy systems relies on accurate photovoltaic (PV) power generation predictions.
• Despite advancements in weather forecasting, precise PV power prediction remains a significant challenge.
• The inherent uncertainty of renewable energy sources necessitates reliable forecasting methods.

Purpose of the Study:

• To develop and validate a novel approach for optimizing photovoltaic power prediction.
• To enhance the accuracy of solar power forecasting by dynamically refining neural network structures.
• To minimize prediction errors using a combination of genetic algorithms and neural network optimization.

Main Methods:

• A novel methodology combining genetic algorithms and dynamic neural network structure refinement was employed.
• Neural network parameters, including neurons, transfer functions, weights, and biases, were dynamically adjusted during training.
• The approach was evaluated using annual, monthly, and seasonal data over twelve representative days, with comparisons to multiple linear regression and nonlinear autoregressive neural network models.
• MATLAB was utilized for modeling, training, and testing, with validation on a real 4.2 kW PV plant.

Main Results:

• The proposed method demonstrated significant improvements in prediction accuracy compared to existing models.
• Evaluation metrics including mean square error (MSE), R-value, and mean percentage error indicated promising results.
• Achieved low mean square errors of 20 W on cloudy days and 175 W on sunny days.
• High prediction versus target regression consistency was observed, with R values ranging from 0.95824 to 0.99980.

Conclusions:

• The novel approach effectively enhances the reliability and accuracy of photovoltaic power generation predictions.
• Dynamic adjustment of neural network parameters is a viable strategy for optimizing solar power forecasting.
• The methodology offers a significant advancement in managing energy systems with high renewable energy penetration.