Application of Generalized Regression Neural Network in Predicting the Thermal Performance of Solar Flat Plate Collector Systems

Abstract

Abstract This study presents a smart neural network (NN) model for estimating the thermal performance of a transient nature solar flat plate collector system (SFPCS). For this purpose, a series of experimental studies are conducted through four successive days with three different arrangements of SFPCS (standalone, series, and parallel). Experimental results of such arrangements are then used for designing a generalized regression neural network (GRNN) model. The GRNN architecture proposed in this study consists of four inputs (mass flowrate, solar irradiance, fluid temperature difference, and collector area) and two dependent outputs (power output and efficiency of SFPCS). Such GRNN architecture is trained, tested, and validated with real-time experimental transient datasets for each arrangement individually. The results of the GRNN model are in good agreement with experimental datasets. The overall accuracy of the developed GRNN model in predicting the performance of standalone, series, and parallel connected SFPCS is 98%.