A novel approach to solar radiation forecasting

In this study, an hourly solar radiation forecasting approach was developed to combat climate change, ensure the stable and reliable use of solar energy, and conduct feasibility studies for solar energy projects. For this purpose, a time series was constructed using hourly solar radiation measurements recorded between 2018 and 2022, obtained from the Afyonkarahisar Meteorology Service. The proposed forecasting model employs a hybrid approach based on Seasonal-Trend Decomposition (STL), combining the SARIMA method, which effectively captures linear relationships in the solar radiation time series, and the Long Short-Term Memory (LSTM) neural network, known for its ability to model nonlinear dependencies. The STL method, particularly effective for time series with high seasonal variability, enables the decomposition of the solar radiation series into distinct components for separate modeling. The stability and predictive performance of the developed hybrid model were evaluated across all seasonal periods using 10-fold cross-validation applied to the solar radiation data.