基于人工神经网络的蒸汽压缩制冷系统建模方法

Modeling Method of Vapor Compression Refrigeration System Based on Artificial Neural Network

针对传统蒸汽压缩制冷系统建模方法运算量大且建模精度难以保证的问题,研究了基于人工神经网络的建模方法。结合蒸汽压缩制冷系统实际特点,研究并确定了训练数据、网络结构等要素,分析了网络训练算法。利用Levenberg-Marquardt(LM)法和最速下降法2种典型网络训练算法分别对神经网络参数进行寻优,比较了2种算法的优劣。最后,将所建网络模型应用于蒸汽压缩制冷系统输出变量的预测,结果表明:该网络模型预测精度较高,说明建模方法有效。

Aiming at the problems that the traditional modeling method of Vapor Compression Refrigeration System( VCRS) has a large amount of computation and the accuracy of modeling is difficult to guarantee,the modeling method based on Artificial Neural Network( ANN) is researched. Combined with the practical characteristics of VCRS,the training data and network structure are studied and determined,and the network training algorithm is analyzed. By using Levenberg-Marquardt( LM) method and the steepest descent method,2 typical network training algorithms are used to optimize the parameters of neural network,and the advantages and disadvantages of the 2 algorithms are compared. Finally,the network model is applied to the prediction of the output variables of VCRS. The results show that the network model has higher prediction accuracy,and the modeling method is effective.