Optimized Random Forest for Solar Radiation Prediction Using Sunshine Hours.

Area of Science:

• Renewable Energy Systems
• Photovoltaic Technology
• Solar Radiation Modeling

Background:

• Increasing demand for renewable energy sources, particularly solar photovoltaics.
• Essential need for accurate solar radiation prediction mechanisms for effective installation planning.
• Queretaro, Mexico, offers favorable conditions with abundant direct solar radiation.

Purpose of the Study:

• To present solar radiation prediction results using an optimized Random Forest (RF) algorithm.
• To compare the performance of the optimized RF model against other machine learning models.
• To validate the developed prediction method in different scenarios and time periods.

Main Methods:

• Optimization of hyperparameters for the Random Forest (RF) and Adaboost algorithms.
• Comparative analysis with conventional methods like linear regression and recurrent networks.
• Application and validation of the prediction model in Juriquilla, Queretaro, for 2020 and 2021 data.

Main Results:

• Optimized RF and Adaboost models achieved a 95.98% accuracy improvement in solar radiation prediction.
• Demonstrated superior performance compared to linear regression (54.19%) and recurrent networks (53.96%).
• Achieved high accuracy without increased computational time or performance demands.

Conclusions:

• The optimized Random Forest algorithm provides a highly effective and accurate method for solar radiation prediction.
• The developed approach offers robust performance and confirms its validity for solar energy planning.
• The method is efficient, maintaining performance without escalating computational requirements.