Performance prediction of heat pipe evacuated tube solar collectors: Analytical modeling and data-driven machine learning/ANN approach with developing web application

This study predicts the performance of evacuated tube collectors by examining diverse working fluids in heat pipe and manifold, as well as geometric, optical, thermodynamic, and hydrodynamic factors. A comprehensive dataset was created by integrating data from thermal network modeling, validated with previous studies, and experimental results based on literature findings across the field, supporting the development and application of data-driven machine learning methods. Random Forest, Extra Trees, XGBoost, LightGBM, and ANNs were evaluated, with Extra Trees delivering the best performance, followed by LightGBM. Additionally, XGBoost excelled in terms of speed and stability. Extra Trees achieved an average Mean Absolute Percentage Error (MAPE) of 1.1450% and an average Root Mean Square Percentage Error (RMSPE) of 4.0179%, then validated with test datasets and experimental data from other studies, giving an RMSPE of 4.06% and a correlation coefficient of 0.995. In addition to analytical and machine learning modeling, a graphical user interface developed in this study enables users to predict outputs by inputting specific parameter values, bypassing complex heat transfer calculations. The platform integrates visualization and machine learning predictions, offering an efficient all-in-one solution for evaluating collector performance. This application improves access to performance predictions, filling a critical gap in the literature.