Advancing solar distilled water yield prediction using hybrid machine learning and weighted average techniques

This study explores predictive modeling for solar still productivity, crucial for sustainable water purification. It examines various machine learning models, including Random Forest (RF) and Multilayer Perceptron (MLP), and introduces advanced hybrid models. The standout, a hybrid MMLPs-RF Weighted Average model, demonstrates superior predictive accuracy, achieving exemplary metrics like a Coefficient of Determination (COD) of 0.99249, Mean Squared Error (MSE) of 0.00002, Mean Absolute Error (MAE) of 0.00403, Root Mean Squared Error (RMSE) of 0.00494, and Explained Variance Score (EVS) of 0.99252. The research highlights the suitability of the Weighted Average method for the complex, non-linear solar energy data, enhancing accuracy through optimized model contributions and reduced variance. This approach is particularly effective for addressing the dynamic and seasonal nature of solar productivity. The findings underscore the potential of hybrid models, especially those using weighted averaging, to improve solar still efficiency and sustainability. By providing a robust framework for solar still optimization and a critical analysis of machine learning methods' applicability, this research contributes significantly to renewable energy and environmental science, paving the way for future predictive modeling advancements in renewable energy systems. [Display omitted] •Weighted Average enhances accuracy in solar forecasts.•MLP surpasses RF in predicting solar productivity.•MMLPs-RF model leads with exceptional accuracy.•Adjustments for seasonal variations improve forecasts.•Hybrids excel over single-method approaches in solar studies.