Performance analysis of CO2/NH3 cascade refrigeration system using artificial neural networks
In this study, artificial neural networks (ANNs) have been used for performance analysis of a CO2/NH3 cascade refrigeration system. Energy and exergy analysis of the system is firstly investigated using a computer code implemented in EES. It is well known that four main variables, including condensing temperature of ammonia, evaporating temperature of carbon dioxide, condensing temperature of carbon dioxide and temperature difference in cascade condenser affect the coefficient of performance (COP) and the exergetic efficiency. In this study, these two parameters in addition to certain useful values such as mass flow rate of high and low temperature circuits, power consumption of each compressor and also total exergy destruction are estimated in terms of the above temperatures. Feed-forward backpropagation learning algorithm was used in the network. A set of calculated data obtained from EES was used as training and test data. The computer program has been performed under MATLAB environment using neural network toolbox. New formulation obtained from ANN for this couple of refrigerants is presented for the calculation of abovementioned target values. The R value obtained when unknown data were used to the networks was 0.999992 which is very satisfactory. As an alternative method, it can be easily implemented in all programming languages with the aims of simulation or optimization. It can also be used where a very accurate and fast estimation of the system performance is of interest to engineers.