面向需求侧响应的热泵式建筑供暖系统研究  

Buildings heat pump heating systems for demand response

作　　者：蒋沛宏 徐宝萍[1] 李前岗 王锡[1] JIANG Peihong;XU Baoping;LI Qiangang;WANG Xi(North China Electric Power University,Beijing 100096,China)

机构地区：[1]华北电力大学,北京100096

出　　处：《区域供热》2023年第2期100-111,共12页District Heating

基　　金：国家自然科学基金资助项目(52278106);国家重点研发计划项目政府间国际科技创新合作(2021YFE0194500)。

摘　　要：通过电动热泵技术与建筑供暖系统被动蓄放热的结合,实现用户侧需求响应,是低成本且有效地促进可再生能源消纳的技术手段。本文基于建筑供暖系统各部件的详细物理模型和MATLAB编程,实现了融合末端-建筑热过程-源端的建筑供暖系统综合动态热模拟,并实验验证其可靠性。通过该模拟平台获得仿真数据集,采用深度学习算法建立了建筑及末端散热器的神经网络模型。基于上述模型,利用模拟退火优化算法,实现了需求响应模式下供暖系统模型预测控制。并将该控制策略与规则控制、比例控制应用于北京某典型建筑的供暖期运行仿真。结果表明供暖初末寒期,模型预测控制较比例控制降低成本约66.1%,较规则控制降低成本约48.3%,负荷转移持续时间约20 h;严寒期模型预测控制较比例控制降低成本约61.0%,较规则控制降低成本约9.6%,负荷转移持续时间约8 h。It is a cost-effective solution by utilizing passive thermal storage in building heating systems,combined with electrical heat pump technology,for demand response operation,to achieve the goals of integrating large share of renewables.Firstly,based on the detailed physical model of each component of the building heating system and MATLAB programming,an integrated dynamic thermal simulation platform combined from the sub-models for rooms,radiators and the heat pump is developed,and its reliability is verified through experiments.Then the training data set is obtained through the simulation platform,and the neural network model of the building and the radiator is established by using the depth learning algorithm.Based on the above model,a model predictive control strategy for cost optimization of heating system is proposed by using simulated annealing algorithm.Finally,the proposed control strategy,rule control and proportional control are applied to a modeling analysis case of a typical building in Beijing during the heating period.The results show that,the model predictive control can reduce the cost by 66.1%compared with the proportional control,can reduce the cost by 48.3%compared with the rule control at the beginning and end of the heating period,and the delayed operation time is about 20 h;in the severe cold period,the model predictive control can reduce the cost by 61.0%compared with the proportional control,can reduce the cost by 9.6%compared with the rule control,and the delayed operation time is about 8 h.

关 键 词：需求响应 模型预测控制 综合热模拟 神经网络 退火算法 

分 类 号：TU832[建筑科学—供热、供燃气、通风及空调工程]