Modeling of performance of fixed and moving type air solar collectors with machine learning algorithms

In this thesis study, in order to determine the thermal performances of the fixed and the moving (with programmable control system solar tracking feature) solar air collector (SAC) with same feature in (0.0232, 0.0322, 0.0386 kg/s) mass flow rate values in Elazig province climate conditions experimentally tested. The parameters measured in the experiments are solar radiation (I), collector outlet air velocity, collector outlet, inlet, glass and floor temperatures. Energy and exergy efficiency values were calculated using experimental data according to the I. and II. laws of thermodynamics. In the moving SAC system, the mean energy and exergy efficiencies for mass flow rates (0.0232, 0.0322, 0.0386 kg / s) were calculated as (%50.3, %64.5, %78.1), (%1.9, %2.28, %2.68), respectively. The mean energy and exergy efficiencies for the fixed SAC system (0.0232, 0.0322, 0.0386 kg/s) were calculated as (%32.8%, %43.43%, %53.03%), (%0.8, %0.96, %1.1), respectively. The continuous vertical solar radiation has increased the efficiency of the moving SAC system. It has been determined that as the flow rate increases, energy and exergy efficiency increases. Artificial Neural Network (ANN) and Decision Tree (DT) machine learning algorithms were used to estimate the calculated exergy yield results of the fixed SAC system. Statistical error analyzes such as absolute error (MAE), root mean square error (RMSE), relative absolute error (RAE) and root relative absolute error (RRAE) were used for the validity of the predictive models. It has been observed that the results of predictive models obtained by machine algorithms are similar to experimental results. The method that creates the predictive model with the least error (RMSE 0.0564) is the ANN algorithm.