基于ANN的空气处理单元换热器的动态性能预测

应用人工神经网络技术对安装在空气处理单元（Air Handler Unit，AHU）中的换热器建立了动态神经网络模型。该模型的所有预测值基本都在实际测量值的95%~105%之间。实际测量值和神经网络模型预测值之间的平均相对误差（Mean Relative Error，MRE）小于2.5%，充分说明模型逼近实际对象的精确性。实验结果显示，神经网络，尤其是BP神经网络，可以作为一个很容易的建模工具来获得空气处理单元中换热器的神经网络模型，同时也很容易应用在类似的热系统中。

Prediction of dynamic performance for heat exchanger mounted in AHU-based on ANN

This study uses the neural network technique to obtain the dynamic models of the heat exchanger mounted in the Air Handler Unit（AHU）.Almost all of the predicted values are within 95%~105% of the actual measured values.And the consistence between measured and ANNs predicted results is achieved by a Mean Relative Error（MRE）<2.5%.The results show that neural networks，especially the BP network，can be an easily modeling tool for obtaining the models of heat exchangers and similar thermal system.