Intelligent simulation on refrigeration system using artificial neural network

Because of the dynamic, nonlinear and multi-parameter characteristic of the refrigeration system, it is difficult to keep the system operating under the optimal state. Based on the improved back-propagation (BP) of artificial neural network (ANN) with the momentum factor, the program to predict the performance of refrigeration system at part-load condition is established by Visual C++ 6.0. The training or testing data is from a refrigeration experiment system with HCFC22. The input layer includes 3 neurons, i.e. the indoor and outdoor air temperature and compressor frequency. The prediction result indicates that the artificial neural network method is a kind of effective way to analyze the performance of refrigeration system. This work can provide guidance on the saving-energy control method of refrigeration system at part-load condition.