Akademik Veri Yönetim Sistemi
ELECTRICAL ENGINEERING, cilt.107, sa.6, ss.7329-7350, 2025 (SCI-Expanded, Scopus)
Machine learning (ML) algorithms can provide highly accurate predictions, but their complexity often makes them difficult to interpret due to their black-box nature. Combining ML and Explainable Artificial Intelligence (XAI) makes these models more transparent and enables users to understand the key factors behind the predictions. This paper presents a variety of ML approaches combined with XAI to predict solar power generation, aiming to optimize energy management in smart grids. For this purpose, five different ML algorithms, namely Random Forest, Gradient Boosting, eXtreme Gradient Boosting, Light Gradient Boosting Machine, and Decision Three, and three different XAI models, namely Shapley Additive Explanation, Local Interpretable Model-agnostic Explanations, and Permutation Feature Importance were investigated. According to the findings, the Light Gradient Boosting Machine achieved the best performance, showing the lowest root mean squared error of 0.088, mean squared error of 0.007749, mean absolute error of 0.052896, and the highest R-squared value of 0.842645. XAI techniques provided detailed, actionable insights into the model behavior, helping to identify key features influencing predictions. While Permutation Feature Importance identified key features, Shapley Additive Explanations showed how this feature interacts with others, and Local Interpretable Model-Agnostic Explanations clarified individual predictions. This research provides the most comprehensive understanding in the literature on feature impact and model interpretability for solar power forecasting and contributes to sustainable energy applications by combining ML prediction with XAI interpretability for effective smart grid integration.