基于机器学习预测建筑用空调性能的研究  

Research on predicting the performance of air conditioning for buildings based on machine learning

作　　者：焦焕淞 肖鑫 JIAO Huan-song;XIAO Xin

机构地区：[1]东华大学环境学院空气环境与建筑节能研究所,上海201620

出　　处：《节能》2024年第1期55-58,共4页Energy Conservation

基　　金：中央高校基本科研业务费专项基金(项目编号:2232021D-11);上海市引进海外高层次人才计划;东华大学青年教师科研启动基金;东华大学高层次人才专项基金。

摘　　要：露点蒸发冷却空调利用水蒸发吸热实现制冷。由于气候不同,不同地区露点蒸发冷却空调的性能存在差异。基于BP神经网络法和随机森林法两种机器学习方法,将环境温度、相对湿度、空气流速、工作空气比例、热质换热器高度、间隔、层数等7种特征作为自变量,将露点蒸发冷却空调的制冷量、制冷COP、露点效率和湿球效率作为因变量,对露点蒸发冷却空调的性能进行机器学习并预测。结果显示:两种机器学习方法均可对露点蒸发冷却空调的性能进行预测,随机森林法的预测误差小于10%,BP神经网络法的预测误差低于2%。BP神经网络法的预测效果更佳。Dew point evaporative cooling(DPEC)air conditioning achieves cooling through the water evaporation and heat absorption.Due to climatic variations,the performance of DPEC air conditioning differs across different regions.With two machine learning methods,namely the Back Propagation(BP)neural network and the Random Forest algorithm.This study utilizes seven features,including environmental temperature,relative humidity,air velocity,working air ratio,height of the heat exchanger,spacing,and number of layers,as independent variables.The dependent variables include the refrigeration capacity,Coefficient of Performance,dew point efficiency,and wet bulb efficiency of the DPEC air conditioning.Both machine learning methods were employed to predict the performance of the DPEC air conditioning.The results indicate that both machine learning methods can effectively predict the performance,with the prediction error of the Random Forest algorithm being less than 10%and the BP neural network below 2%.The BP neural network method demonstrates superior prediction performance.

关 键 词：露点蒸发冷却 空调性能 机器学习 BP神经网络法 随机森林法 

分 类 号：TU831[建筑科学—供热、供燃气、通风及空调工程]