Application of artificial neural network for performance prediction of a nanofluid-based direct absorption solar collector

Recently, artificial neural network techniques have been widely used for the performance prediction of the renewable energy systems, in which solar collectors are one of the most used mechanical equipment. In this paper, the thermal performance of a nanofluid-based direct absorption solar collector is predicted using an artificial neural network based Multi-Layer Perceptron system. In the experimental part of study, nine collector prototypes with different geometries were tested at different conditions to investigate the effect of the collector depth and length on the collector thermal performance and also, to provide the required data for the network evaluation. The collector depth and length, the working fluid flowrate and concentration and the reduced temperature difference are selected as input parameters of the network to estimate the collector efficiency and Nusselt number. The proposed artificial neural network approach proved that the variation of the collector depth of 5–15 mm increases the collector efficiency about 9%, while the collector length has an insignificant effect on the collector efficiency. The Nusselt number of the collector increases considerably by the collector depth and nanofluid flowrate. The proposed network has the best performance for predicting the collector efficiency with nanofluid concentration of 1000 ppm as the input parameter by achieving the MAPE of 1.470%. In the case of predicting Nusselt number, the best performance with MAPE of 2.576% is obtained with collector length of 300 mm. Using the forward stepwise regression selection method, the best combination of input parameters for predicting the Nusselt number is obtained using all input parameters. The consistency of the experimental and predicted results confirms the great ability of artificial neural network to predicting the thermal performance of direct absorption solar collectors.