Accurate prediction of exergetic efficiency of solar air heaters using various predicting methods

Abstract In this article, three different types of predicting techniques have been implemented for estimating exergy efficiency of solar air heaters (SAHs). For the purpose, group method of data handling (GMDH), artificial neural network (ANN) and multi linear regression (MLR) models were applied. For these models, a total of 210 data sets were collected from various experiments. These experiments were performed with two different solar collectors with smooth and roughened surfaces, using mass flow rate from 0.007 to 0.0222 kg/s. Ten different types of variables, which are ambient temperature, wind speed, relative humidity, fluid inlet temperature, fluid mean temperature, mass flow rate, plate temperature, wind direction, solar intensity and solar elevation, were selected as independent variables in all models. The experimentally obtained exergy efficiency is selected as output or dependent variable. In this study, 168 sets of data were picked-up for training purpose and 42 datasets were selected for testing. GMDH, ANN and MLR techniques accurately performed with values of correlation coefficient (R) as 0.98977, 0.99981 and 0.97693 respectively. The comparative study of all models reveals that out of these three techniques, ANN performs the best. In the ANN model, the values obtained were 8.50E-03, 0.00584 and 0.99981 of RMSE, MAE and R respectively, which are the optimal results when compared to those of other models. After ANN, GMDH performed better than MLR. The above analysis reveals that the exergy efficiency of SAH predicted using ANN technique gives the highest accuracy.