基于用户画像的暖通空调智能调控  

Research on HVAC intelligent control strategy based on user behavior pattern

作　　者：胡锐[1] 袁海峰[1] 芮忠 HU Rui;YUAN Haifeng;RUI Zhong(Information Construction and Management Center,Suzhou University of Science and Technology,Suzhou 215009,China)

机构地区：[1]苏州科技大学信息化建设与管理中心,江苏苏州215009

出　　处：《现代电子技术》2024年第1期134-139,共6页Modern Electronics Technique

基　　金：2019年校级教学改革与研究项目(2019JGMY-323);国家自然科学基金资助项目(61472267)。

摘　　要：目前,建筑物中的暖通中央空调系统基本上是根据行业准则设置的,然而,多项研究表明,由于用户的偏好、姿态和需求各异,这种传统做法不太可能满足大多数用户的热量需求。为了更精准地满足用户热量需求,使得基于用户行为模式来针对性提高热舒适性成为可能。文中采用深度强化学习的方法实现智能化暖通空调智能送风策略,该策略能够根据环境和用户行为模式动态地确定最佳暖通空调设置(温度设置和送风设置),从而极大提高用户的热舒适度,同时,实验结果还表明该智能调控策略相比传统固定值控制策略具有一定的节能效果。At present,HVAC(heating,ventilation and air⁃conditioning)systems in buildings are basically set according to industry standards.However,a number of studies have shown that the traditional approach is unlikely to satisfy the heat demand of most users because of their different preferences,attitudes and needs.In order to meet the heat demand of users more accurately,it is possible to improve thermal comfort based on user behavior patterns.In this paper,a deep reinforcement learning is adopted to realize HVAC intelligent air supply strategy,which can dynamically determine the best HVAC setting,that is,temperature setting and supply air setting,according to the environment and user behavior patterns,so as to greatly enhance the thermal comfort of the users.The experimental results also show that the intelligent control strategy has a certain energy⁃saving effect in comparison with the traditional fixed value control strategy.

关 键 词：深度强化学习 用户行为模式 热舒适度 深度学习 建筑节能 服装热阻值 DQN 神经网络 

分 类 号：TN99-34[电子电信—信号与信息处理] TP391.41[电子电信—信息与通信工程]