This work investigates the integration of solar radiation (Rs) into the power grid—a very crucial energy source in sustaining life on Earth. In fact, Rs is a pivotal energy source. However, due to the intermittent features of this energy source, the challenge of effectively integrating it into the power grid arises. In this work, Machine Learning (ML) and feature selection (FS) methods are applied in a pioneering way to improve the accuracy of Rs forecasting. Principal Component Analysis (PCA) has been applied alongside various ML models, including Random Forest (RF), Gradient Boosting Models (GBM), Logistic Regression (LR), Classification and Regression Tree (CART), and Decision Tree (DT), in combination with Recursive Feature Elimination (RFE). Quantitative analysis indicates that LR and RF achieve the lowest normalized Mean Absolute Error (nMAE) under PCA, although the difference is minimal. Notably, GBM stands out in performance. RFE has demonstrated powerful predictive precision throughout the three models. CART, LR, and GBM. Comparatively, RF has similar performance under both PCA and RFE, but in the case of LR, its predictive accuracy comes out with RFE. For GBM, nMAE is lower under PCA. Nuanced dynamics of both DT and CART models are discussed, along with their adaptability to RFE. The emphasis of this study is on model-specific considerations in the optimization of solar energy (SE) integration and provides valuable insights into power grid enhancements.