Advanced Prediction of Solar Radiation Using Machine Learning and Principal Component Analysis

Solar radiation (Rs) is a crucial energy source, vital for illumination, warmth, and life sustenance on Earth. However, its intermittent nature poses integration challenges into power grids. This research introduces an innovative approach leveraging Machine Learning (ML) models for accurate Rs forecasting. Principal Component Analysis (PCA) combines with ML models, including random forest (RF), Gradient Boosting Models (GBM), Logistic Regression (LR), Classification and Regression Tree (CART), and Decision Tree (DT). Our goal is to enhance solar radiation forecasting, making solar energy more reliable and cost-effective thereby advancing renewable energy integration into power grids. This research offers innovative perspectives on model integration, feature selection using PCA, and model suitability, contributing to the advancement of Multi-Variate Time Series analysis. Our findings demonstrate promising outcomes across models, with the following negative Mean Absolute Errors (nMAE): LR and RF achieve the lowest nMAE, around − 0.144 (0.014) and − 0.151 (0.015), respectively. GBM slightly outperforms them with an nMAE of about − 0.154 (0.017), while CART records the highest nMAE of approximately − 0.209 (0.026). Small standard deviations (in parentheses) suggest stable performance during repeated cross-validation.