<p>The recent outbreaks solar energy generation, some challenges have popped up owing to solar intermittent behaviour, require an accurate estimation of the worldwide solar radiation. In this reference AI-ML algorithms are gaining popularity and effectiveness in calculating solar radiation. Since collinearity might cause troubles such as unstable parameter estimation, inaccurate models, and poor predictive performance. However, multiple linearity tends to become a major issue in linear models. To address these issues, the spearman rank Correlation (SRC) and variance inflation factor (VIF) was introduced, and a new variable selection method called Spearman Rank Correlation-Variance Inflation Factor (SRC-VIF) was proposed, in addition to this article presents a better stack ensemble with GRU (SE-GRU) regressor include the machine learning models like, Random Forest Regressor (RF), Support Vector Regressor (SVR), Multilayer perceptron Regressor (MLP), K-Nearest Neighbor Regressor <b>(</b>KNN), Light Gradient Boost Machine (LGMs), CatBoost Regressor (CB), and Extreme Gradient Boost (XGBost), while the ensemble stacking with GRU regressor is used as meta learners. The ensemble technique includes stacking, blending, bagging and boosting methodology. In this paper use above all technique for the prediction of solar radiation of Agra (India) location. With the help of this technique we precise the result and able to minimise the errors in predicted value such as Mean Absolute Error (MAE) 15–56%, Root Mean Squared error (RMSE) 22–60% and R squared error (R2) 26–45%. Ultimately, we can say that the ensemble technique helps to decrease prediction errors and improves planning for intermittent solar resources.</p>