Combination of Taguchi Method and Artificial Intelligence Techniques for the Optimal Design of Flat-plate Collectors

In this paper, artificial neural networks (ANNs) and genetic algorithms (GAs) are used for the design of solar flat-plate collectors. It is proved in this paper that by using the Taguchi method for selecting the data required for training the ANN is very effective in allowing the network to learn the behavior of the system satisfactorily. The parameters on which the flat-plate collector design depends are the collector tube material, the type of collector absorbing plate material, the number of collector riser tubes, the collector riser tube diameter, the type of absorber coating and the thickness of the bottom insulating material. By using the method of Taguchi experiments three levels of six variables were used together with three levels of available solar radiation intensity (Gt) and collector inlet minus ambient temperature difference to estimate the collector thermal efficiency. Thus a total of 162 patterns were collected from these combinations from which 130 were used for the training of the ANN and the rest 32, selected randomly, were used to validate the training accuracy. The input parameters are the factors on which the collector performance depends, listed above, and the output parameters are the collector optical efficiency and the loss coefficient. The trained ANN was then used with a genetic algorithm to find the optimum combination of the values of the input parameters, which maximizes the collector efficiency estimated from the optical efficiency and the loss coefficient. The results obtained are very similar to the results achieved by other researchers using much complicated optimization methods, whereas the present method not only is very accurate but it is also very quick.