时变多参数不确定性下室内环境的节能控制  

Energy efficient control of indoor environments under time-varying multi-parameter uncertainty

作　　者：赵建豪 宋华 南新元[2] 蔡鑫[2] ZHAO Jianhao;SONG Hua;NAN Xinyuan;CAI Xin(No.1 Department of Polyolefin of Dushanzi Petrochemical Company,Karamay,Xinjiang 833699,China;School of Electrical Engineering,Xinjiang University,Urumqi,Xinjiang 830017,China;Xinjiang Architectural Design and Research Institute,Urumqi,Xinjiang 830002,China)

机构地区：[1]中国石油独山子石化分公司聚烯烃一部,新疆克拉码依833699 [2]新疆大学电气工程学院,新疆乌鲁木齐830017 [3]新疆建筑设计研究院,新疆乌鲁木齐830002

出　　处：《河北科技大学学报》2024年第6期597-608,共12页Journal of Hebei University of Science and Technology

基　　金：国家自然科学基金(52065064,62303394);新疆维吾尔自治区自然科学基金(2022D01C694)。

摘　　要：为解决当前室内环境受多种时变参数影响不确定性较大,且现有控制设备无法根据室内环境对工作功率进行自适应调节造成的能源浪费问题,采用在深度确定性策略梯度(deep deterministic policy gradient,DDPG)中融入优先经验重放(prioritized experience replay,PER)的方法,合理优化控制设备功率,对室内空气质量(indoor air quality,IAQ)和热舒适度进行控制。结果表明:DDPG-PER算法能够在冬夏两季不同室外空气质量条件下,结合多时变参数,将室内环境控制在要求范围之内;与定风量控制系统相比,能源成本降低了13.30%,用户全年可节约2000元左右电费。研究结果对中国“碳中和”策略和绿色低碳建筑行业的发展具有一定的参考价值。To solve the problem that the current indoor environment is affected by a variety of time-varying parameters with large uncertainty and the existing control equipment cannot adaptively adjust the operating power according to the indoor environment,which has caused a great waste of energy,the method of integrating the prioritized experience replay(PER)into the deep deterministic policy gradient(DDPG)is adopted.Prioritized experience replay(PER)in DDPG is used to rationally and optimally control the power of the equipment for indoor air quality(IAQ)and thermal comfort.Experiments show that the proposed DDPG-PER algorithm can control the indoor environment within the required range by combining multiple time-varying parameters under different outdoor air quality conditions in winter and summer seasons.Moreover,compared with the fixed-air-volume control system,it reduces the energy cost by 13.30%,and saves about 2000 RMB of electricity cost in a whole year,which is valuable for China’s"carbon-neutral"strategy and the development of green and low-carbon buildings.

关 键 词：环境质量监测与评价 强化学习 深度确定性策略梯度 优先经验重放 节能控制 

分 类 号：TP393[自动化与计算机技术—计算机应用技术] X510.2[自动化与计算机技术—计算机科学与技术]