制冷压缩机热力性能的神经网络模拟

压缩机热力性能的准确计算，对于使用压缩机制冷空调装置的优化设计起到很关键的作用，而单纯的理论模型难以反映实际的复杂因素，影响计算精度．采用人工神经网络与传统理论模型相结合的方式，建立智能型的压缩机热力计算模型，利用人工神经网络的自学习和泛化功能改善压缩机容积效率和电效率的计算模型精度．神经网络采用多层前向网络（ＭＬＰ），网络训练采用同伦ＢＰ算法．对房间空调器用滚动转子式压缩机启动过程的输入功率变化，以及汽车空调器用变转速往复式压缩机的容积效率进行仿真，并与实验结果对照．结果表明，智能型压缩机模型很好地改善了传统计算模型的精度，而且适应能力更强

Neural Network Method for Predicting Refrigeration Compressor Performance

A novel intelligent compressor model which combines artificial neutral network (ANN) with traditional theoretical model was presented. The functions of self learning and generalization of ANN were used to improve the traditional model. Multi layer perceptron (MLP) network was adopted and homotopic BP algorithm was used to train the MLP efficiently. Input power curve of a rolling piston rotary compressor installed in a room air conditioner in the process of start up and volumetric efficiency of a reciprocating inverter compressor used in an automotive air conditioner were calculated and compared with the experimental data. It shows that the new model based neural model reaches more precise results than the traditional ones. Furthermore, the new compressor model is of better flexibility in a large scale.