Prediction of Efficiency for a Passive Flat Plate Collector for Water Desalination using Artificial Neural Network

Abstract

Artificial neural network was used for modeling and prediction of the efficiency of a  passive flat plate collector for water desalination. An extensive experimental program design was undertaken on the collector to obtain the parameters required for the modeling. The neural model to predict the efficiency was developed based on groups of experiments carried out. Five (5) parameters: ambient, inlet fluid and outlet fluid temperatures, radiation, and aperture area of the collector were used as inputs into the network architecture of 5 [5]₁ 1 in predicting the efficiency. After series of network architectures were trained using different training algorithms such as Levenberg-Marquardt, Bayesian Regulation, Resilient Backpropagation using MATLAB 7.9.0 (R20096), the LM 5 [5]₁ 1 was selected as the most appropriate model. Prediction of the neural model exhibited reasonable correlation with the experimental collector efficiency. The predicted collector efficiency gave minimal MSE errors and higher correlation coefficients and Nash-Scutcliffe efficiency (NSE) indicating that the model was robust for predicting the efficiency of a passive flat plate collector for desalination of water.

Keywords: Collector efficiency, desalination, passive solar collector, artificial neural network, Nash-Scutcliffe efficiency, MSE error, modeling.