Empirical data-driven multi-layer perceptron and radial basis function techniques in predicting the performance of nanofluid-based modified tubular solar collectors

In the present study, the modified evacuated tube solar collector (METSC) with a bypass pipe utilizing copper oxide/distilled water (Cu ₂O/DW) nanofluid is experimented. Then, the performance of METSC was predicted through Artificial Neural Networks (ANNs) techniques. The input variables were different volumes of the storage tank from 5 to 8 l, various diameters of the bypass pipe from 6 to 10 mm, and various volumetric concentration of the nanofluid from 0 to 0.04. Also, the output variables were the temperature difference of fluid in 1-h period and the energetic efficiency of METSC. The results demonstrated that the METSC performance was mostly impacted by the tank volume alteration. Moreover, the optimum bypass tube diameter value was obtained, and it was denoted that using the Cu ₂O/DW nanofluid enhances the daily energy efficiency of METSC up to 4%. Furthermore, it was shown that both MLP and RBF techniques are two reliable algorithms to predict the thermal characteristics of an METSC. The maximum amounts of mean relative percentage error for MLP and RBF algorithms were reported as 0.576 and 0.907, respectively. Hence, two mathematical models were reported for formulating the output variables in terms of the input variables using the MLP technique.