基于深度强化学习的地面辐射供暖系统优化控制  

The Optimal Control of Floor Radiant Heating System Based on Deep Reinforcement Learning

作　　者：唐明武 张津育 刘依然 吴小舟 TANG Mingwu;ZHANG Jinyu;LIU Yiran;WU Xiaozhou(Dalian University of Technology,School of Civil Engineering,Dalian Liaoning 116024;Dalian University of Technologv,School of Economics and Management,Dalian Liaoning 116024)

机构地区：[1]大连理工大学土木工程学院,辽宁大连116024 [2]大连理工大学经济管理学院,辽宁大连116024

出　　处：《东北电力大学学报》2022年第2期14-25,共12页Journal of Northeast Electric Power University

基　　金：中央高校基本科研业务费,智慧建筑(DUT20TD113)。

摘　　要：暖通空调系统运行能耗是建筑能耗的主要方面,为了降低能耗节约资源,暖通空调系统优化控制就显得尤为重要.随着深度强化学习(DRL)的发展,其在工程优化控制中的表现良好,更多的暖通空调系统开始采用基于DRL的优化控制方法,但针对类似地暖的热惰性大的系统应用研究相对较少.因此,文中将DRL算法应用到地面辐射供暖系统优化控制中,并综合考虑人体热舒适、工作效率及系统能耗,分别与传统的地暖通断控制和PID控制效果进行对比分析.结果表明,DRL控制在室内热舒适度略优于传统通断控制和PID控制,但工作效率及能耗方面相对略差一些.该研究对象是针对于单房间区域模型,但包括了三个不同的地区,验证了DRL控制策略在地板辐射供暖系统的可行性,为进一步多房间模型研究提供基础.The energy consumption of HVAC system is the main aspect of building energy consumption.In order to reduce energy consumption and save resources, the optimal control of HVAC system is particularly important.With the development of Deep Reinforcement Learning(DRL),optimal controlbased on DRL has been extensively applied in HVAC systems due to its good performance in engineering optimization control.However, there are relatively few studies on the application oflarge thermal inertia system, such as floor heating system.Therefore, this paper focused on theoptimal control of floor radiant heating system using the DRL algorithm, and comprehensively compared the human thermal comfort, human performance, system energy consumption with the traditional on-off control and PID control.The results show that DRL control is slightly better than traditional on-off control and PID control regarding to indoor thermal comfort, and slightly worse than those with regard to human performance andsystem energy consumption.The research object is aimed at the single room area model, but includes three different areas, which verifies the feasibility of DRL control strategy in floor radiant heating system, and provides a basis for further multi room model research.

关 键 词：地面辐射供暖系统 控制策略 深度强化学习 通断控制 变水温控 

分 类 号：TU962[建筑科学]