基于BP算法的中深层同轴套管换热量预测

Heat transfer prediction of medium and deep coaxial casing based on BP algorithm

通过搭建热泵测试试验台,进行多工况试验获取试验数据来建立BP神经网络同轴套管换热量预测模型,并进行仿真模拟,对同轴套管换热量进行预测。结果表明:(1)热泵系统在水流量28 m^(3)/h、回水温度10℃的工况下稳定运行能效最高,同轴套管有效换热量为563 kW。(2)隐含层节点数为12时,BP神经网络预测模型最优,训练最大均方误差MSE为0.023%,最优模型基本结构为9-12-1。(3)对比同轴套管换热量预测值与检验值仿真结果,BP神经网络同轴套管换热量预测平均百分比误差MAPE为0.235%,预测准确率为99.765%。该预测模型具有较高的精度和可靠性,能够准确预测同轴套管换热量的变化趋势,对于提高热泵系统的能效和性能具有广泛的应用价值。

By building a heat pump test bench and obtaining test data under multiple working conditions,a BP neural network coaxial casing heat transfer prediction model was established,and the simulation was carried out to predict the heat transfer of the coaxial casing.The results show that:(1)The heat pump system has the highest energy efficiency during stable operation at a water flow rate of 28 m^(3)/h and a return water temperature of 10℃,and the effective heat transfer of coaxial casing is 563 kW.(2)When the number of hidden layer nodes is 12,the BP neural network prediction model is optimal,with a maximum training mean square error(MSE)is 0.023%,and the basic structure of the optimal model is 9-12-1.(3)Comparing the simulation results of the predicted value and the test value of the coaxial casing,the average percentage error of the prediction of heat transfer of the coaxial casing of the BP neural network is 0.235%,and the prediction accuracy is 99.765%.The prediction model has high ac-curacy and reliability,and can accurately predict the change trend of heat transfer of coaxial casing,which has a wide range of application value for improving the energy efficiency and performance of heat pump systems.