An adaptive neuro-fuzzy approach to predict the thermal efficiency of differently configured solar flat plate water collector systems

In this study, the thermal efficiency delivered by the solar flat plate water collector system (SFPWCS) is predicted using an adaptive neuro-fuzzy inference system (ANFIS). Thermal power output and the incident solar irradiance are considered as inputs to predict the thermal efficiency. Initial experiments are conducted with three differently configured solar flat plate water SFPWCS such as stand-alone (SA), series (SC)- and parallel (PC)-connected rigs. For all the cases of experimental observations, the mass of flow rate of the fluid is fixed as 0.56 kg/min. Under the same atmospheric conditions, PC-SFPWCS registered the highest thermal efficiency as 26.51%. Three individual neuro-fuzzy models are designed, individually trained and tested using the real-time experimental data sets. The predicted results by all three models are then validated by comparing them once again with the experimental results. All three proposed architectures are capable of predicting the real-time experimental results with an accuracy of 99.89%.