基于不同神经网络模型的冷凝器两相换热量的研究

在混合工质下利用4种神经网络模型(反馈神经网络模型(BP)、遗传神经网络模型(GA - BP)、极限学习机网络模型(ELM)和递归神经网络模型(RNN))预测了板式换热器的换热量(含相变换热).结果显示:热源温度为30、40、50 ℃时,GA - BP神经网络模型的平均绝对误差(MAE)、平均相对误差(MAPE)和均方根误差(RMSE)均小于其他3种神经网络模型,且与实际值接近.该结果表明,GA - BP神经网络模型比其他3种神经网络模型更适用于预测板式冷凝器的换热量(含相变换热).

Research on two - phase heat exchange of condenser based on different neural networks

Four neural network models(back propagation neural network model(BP), genetic algorithm -neural network model(GA - BP), extreme learning machine neural network model(ELM)and recurrent neural network model(RNN))were used to predict the heat exchange volume(including phase change heat exchange)of plate heat exchangers with mixed refrigerants.The results show that when the heat source temperature is 30 ℃, 40 ℃ and 50 ℃, mean absolute error(MAE), mean absolute percentage error(MAPE), root mean square error(RMSE)of GA - BP neural network model are smaller than those of the other three neural network models and close to the actual values.The results show that GA - BP neural network model is more suitable than other three neural network models for predicting heat exchange(including phase change heat exchange)in plate condensers.