Understanding Solar Irradiance Predictions Through Explainable AI: Perspectives from Gwangju

초록

Accurate and interpretable solar irradiance forecasts are critical to maximizing photovoltaic energyproduction, especially in areas with significant solar potential. Gwangju Metropolitan City, located inthe southwestern part of South Korea, enjoys abundant solar irradiance, making it an ideal location forsolar energy projects. As Gwangju increasingly focuses on renewable energy, accurate solarirradiance forecasts are essential for efficient planning and management of energy resources. Thispaper uses explainable artificial intelligence techniques, in particular Shapley additive explanations(SHAP), to interpret solar irradiance prediction models built using data from Gwangju. By employingboosting ensemble learning methods such as extreme gradient boosting and light gradient boostingmachine, which are known for their efficiency and accuracy, this research aims to improve theprediction accuracy while increasing the transparency and reliability of the models. SHAP helpsidentify the most influential features that affect solar irradiance, thereby providing betterinterpretability. The study aims to improve the accuracy of solar irradiance predictions, contributing tothe effective deployment and integration of renewable energy in urban environments.

키워드