Estimating Global Horizontal Irradiance of Solar Photovoltaic System from Satellite Data

Global Horizontal Irradiance (GHI) is a key factor in determining solar energy potential and photovoltaic (PV) system efficiency. Accurate estimation of GHI is essential for optimizing solar power generation, yet ground-based measurements are often limited by high costs and geographic constraints. This paper presents a comprehensive framework for predicting GHI using satellite-based data combined with machine learning techniques. The study incorporates multiple data preprocessing steps, including Fourier transformation and time series decomposition, to enhance the learnability of the features. Several machine learning models, such as Linear Regression, Ridge Regression, Random Forest, Elastic Net, and XGBoost, were employed to map satellite data to actual ground-based GHI values. Performance was evaluated using metrics such as Mean Absolute Error (MAE) and Root Mean Squared Deviation (RMSD), with XGBoost and Elastic Net demonstrating superior results. The findings show that combining satellite data with advanced machine learning methods provides a robust and cost-effective solution for accurate GHI prediction. Future work aims to extend the model's applicability by incorporating more diverse datasets and investigating the impact of climate change on solar irradiance patterns. The proposed framework can also be used to optimize solar power plant design and enhance renewable energy infrastructure planning.