Prediction Of The Refrigeration And Air Conditioning System Performance By Neural Networks Based On Genetic Algorithms

There are many disadvantages of traditional ways to design refrigeration and air-conditioning systems,such as high development cost,long development period and low efficiency.With traditional design ways are ameliorated by computer simulation means,it could not only reduce development cost,shorten development period but also improve development effieieney and is taking an important significance for promoting the modemization of product design process.This essay takes single-stage steam compression refrigeration system under steady state as researeh object. By using computer stimulation technology,with necessary and reasonable assumption and simplification,establish model of four parts: the compressor thermal parameters model,the condenser distributed parameter model,the capillary distributed parameter model and the evaporator distributed parameter model. According to the results calculated from the compressor thermal parameters model,using the sample data by GA-BP neural network model predicting the refrigerating capacity and the input power,also comparison with the coefficient model. According to the results calculated from the condenser distributed parameter model,using the sample data by GA-BP neural network model predicting the heat transfer and the refrigerant outlet temperature. According to the results calculated from the capillary distributed parameter model,using the sample data by GA-BP neural network model predicting the refrigerant flow and the refrigerant critical temperature. According to the results calculated from the evaporator distributed parameter model,using the sample data by GA-BP neural network model predicting the heat transfer and pressure drop.Finally,combining the components to establish a steady state simulation model of the whole system. According to the results calculated from the system model,using the sample data by GA-BP neural network model predicting the per unit mass cooling capacity and the COP of the system. The prediction output of the neural network established according to the results of sample training is accurate,with the error from-5% to 5%. Meanwhile,the model calculation rate much faster than the distribution model. The research and analysis shows that the neural network model can quickly and accurately predict the performance of the various components and the whole system.