Artificial neural network approach for evaluation of temperature and density profiles of salt gradient solar pond

Abstract

The purpose of this study is to evaluate temperature and density profiles of an experimentally investigated salt gradient solar pond (SGSP) by using artificial neural network ( ANN). The input parameters of the ANN are solar pond depth, ambient temperature, radiation absorption coefficient of salty solution in the pond, initial density values of the pond and time of day. The output parameters of the ANN are temperature and density profiles in the pond. The experimental data set consists of 168 values. These divided into two groups, of which the 134 values were used for training/learning of the network and the rest of data ( 34 values) for testing/validation of the network performance. According to the ANN predicted results compared to the experimental results, the mean relative error (MRE) is 2.30% for temperature and 0.63% for density. The correlation coefficients (R-2) between the experimentally measured and the ANN predicted results are 0.9632 for temperature and 0.9855 for density in the test/validation data set. The calculated errors of proposed ANN model are in acceptable ranges. These results indicated that the ANN approach could be considered as an alternative and practical technique to evaluate the temperature and density profiles of a SGSP.