摘要
为了降低暖通空调系统(Heating,Ventilation and Air Conditioning,HVAC)能耗,并解决现有研究大多集中于制冷时段设备的局部优化而忽略了末端工况的重要性,以及传统的控制方法可能无法有效的面对复杂或者是动态的系统环境的问题。本文将专家知识与Deep Q-Network和L-BFGS-B相结合,提出了1种混合无模型控制(Hybrid model-free control,HMFC)策略,在考虑末端工况的情况下,对严寒地区地源热泵系统进行制冷与制热时段的热泵冷冻水/热水出水温度、空调水循环泵频率、地埋管循环泵频率以及组合空调机组新风比的全局优化,并将结果与基于规则的控制(Rule-based control,RBC)和模型预测控制(Model predictive control,MPC)进行对比分析。结果表明,HMFC可以快速收敛,并且收敛后的结果在制热时段较RBC节能5.21%;在制冷时段,相较于RBC与MPC分别节能6.88%与2.91%,说明所提出的HMFC策略具有良好的优化控制性能,对于没有大量历史数据的建筑而言,该策略可以作为可行优化方案。最后,将该策略投入至实际工程中,实践结果表明,在专家知识的约束下,该策略可以实现有效节能并具有良好的鲁棒性。
The energy consumption of heating,ventilation and air conditioning(HVAC)systems is an important part of building energy consumption,so it is very important to optimize the energy management of HVAC.In terms of optimization objectives,most of the existing research focuses on the local optimization of equipment during refrigeration but ignores the importance of the terminal condition,and the traditional control method may not be effective in the face of a complex or dynamic system environment.Therefore,in order to solve the above problems,this paper combined expert knowledge with Deep Q-Network and L-BFGS-B,and proposed a hybrid model-free control strategy(HMFC)to globally optimize the heat pump chilled water/hot water discharge temperature,HVAC water circulation pump frequency,buried pipe circulation pump frequency,and the fresh air ratio of combined airconditioning units in the cooling and heating periods of the ground source heat pump system in the cold region.The results were compared with rule-based control(RBC)and model predictive control(MPC),to find that HMFC can converge quickly,and the energy saving of the converged result was 5.21%more than that of RBC in the heating period.In the cooling period,compared with RBC and MPC,the energy saving was 6.88%and 2.91%higher respectively,which indicated that the proposed HMFC strategy has good optimization control performance,and can be used as a feasible optimization scheme for buildings without a large amount of historical data.Finally,the strategy was put into real engineering,and the practical results showed that the strategy can realize effective energy saving and good robustness under the constraint of expert knowledge.
作者
张晓明
王馨慰
张昊天
王晨铮
陈启立
ZHANG Xiaoming;WANG Xinwei;ZHANG Haotian;WANG Chenzheng;CHEN Qili(School of Municipal and Environmental Engineering,Shenyang Jianzhu University,Shenyang 110168,China;Shenzhen Funeng New energy Technology Company Limited,Shenzhen 518109,Guangdong,China)
出处
《建筑科学》
北大核心
2025年第8期86-99,共14页
Building Science
关键词
深度强化学习
混合无模型控制策略
暖通空调系统
全局优化
deep reinforcement learning
hybrid model-free control strategy
HVAC system
global optimization
作者简介
张晓明(1963-),男,硕士,教授级高级工程师;通讯作者:王馨慰(1999-),女,硕士。
