基于BP神经网络的横流式蒸发冷凝器鼓泡式板片传热性能预测  

Prediction of heat transfer performance of bubbling plates in cross-flow evaporative condensers based on BP neural network

作　　者：席鹏飞 章立新[1] 张坤龙 陈权 周庆权 高明[1] 刘婧楠[1] 陈永保[1] 潘旭光 陈婷婷 Xi Pengfei;Zhang Lixin;Zhang Kunlong;Chen Quan;Zhou Qingquan;Gao Ming;Liu Jingnan;Chen Yongbao;Pan Xuguang;Chen Tingting(University of Shanghai for Science and Technology,Shanghai,China)

机构地区：[1]上海理工大学上海市动力工程多相流动与传热重点实验室 [2]浙江三新科技有限公司

出　　处：《暖通空调》2021年第4期136-140,129,共6页Heating Ventilating & Air Conditioning

基　　金：国家自然科学基金资助项目(编号:51976127)。

摘　　要：为预测蒸发冷凝器中鼓泡式板片空气侧的复合换热系数,搭建了一个由2块鼓泡式板片组成的传热性能测试实验系统,在一定工况下通过调节电加热功率以保持板片壁温为60℃。实验期间环境条件变化范围为:大气压98.8~99.3 kPa,进口空气干球温度26~37℃,进口空气湿球温度23~32℃。可调节参数的范围为:喷淋水流量100~400 L/h,截面风速1.0~3.7 m/s,板片间距20~30 mm。计算了板片与空气间的复合换热系数。利用3层BP神经网络处理实验数据,输入参数为进口空气干球温度和湿球温度、喷淋水流量、截面风速及板片间距,输出参数为板片与空气间的复合换热系数。预测结果的相关系数为0.999 2,平均相对误差为0.355 94%,均方根误差为0.508 01 W/(m^(2)·K),表明BP神经网络对蒸发冷凝器中鼓泡式板片空气侧复合换热系数的预测有较高的准确度。In order to predict the composite heat transfer coefficient of bubbling plates on the air side of evaporative condensers,establishes a heat transfer performance test system consisting of two bubbling plates,and keeps the wall temperature of the plates at 60℃by adjusting the electric heating power under certain working conditions.During the experiment,the environmental conditions vary from 98.8 to 99.3 kPa in atmospheric pressure,26 to 37℃in inlet air dry bulb temperature and 23 to 32℃in inlet air wet bulb temperature.The range of adjustable parameters is as follows:spray water flow rate 100 to 400 L/h,cross-section wind speed 1.0 to 3.7 m/s,plate spacing 20 to 30 mm.Calculates the composite heat transfer coefficient between plate and air.Processes the experimental data by three-layer BP neural network.The input parameters are inlet air dry and wet bulb temperatures,spray water flow rate,cross-section wind speed and plate spacing,and the output parameter is the composite heat transfer coefficient between plate and air.The correlation coefficient of prediction results is 0.9992,the average relative error is 0.35594%,and the root mean square error is 0.50801 W/(m^(2)·K),which indicates that the BP neural network has high accuracy in predicting the composite heat transfer coefficient of bubbling plate air side in evaporative condensers.

关 键 词：蒸发冷凝器 鼓泡式板片 BP神经网络 传热性能 复合换热系数 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TK124[自动化与计算机技术—控制科学与工程]