A Hybrid Model For Prediction Of Solar Insolation

The rapid shift toward renewable energy will require very high accuracy predictions in solar insolation to efficiently implement solar panels in the right regions. This study presents a hybrid machine learning strategy for predicting hourly minute-scale insolation of sun radiation for any given location, unlike other common models based solely on monthly averages or daily mean values. The methodology utilizes a GRU model to predict weather parameters (temperature, humidity, pressure, etc.) for a specific time. These weather features are then fed into a Random Forest model trained on ISRO's monthly solar insolation data and weather data from OpenWeather. This framework effectively bridges the resolution gap between monthly and hourly forecasts, enabling granular insights for solar energy planning. The model is tested over Ahmedabad, India; results showed high predictability accuracy with error metrics like MAE, RMSE, and R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> to validate the robustness of the model. This workflow allows users to estimate the solar energy potential for any city at a given time in order to aid in decisions on placing solar panels, conducting feasibility studies, and estimating energy yield. With GRU, RF, and the logical division approach combined, there is a highly efficient, computationally explainable, and scalable solution for energy planning. There is potential in expanding this hybrid framework to more cities, especially about the promotion of sustainable energy projects and smart grid management.