多参数建筑环境的高效率建模、控制与优化方法研究

Due to the increasing energy-use intensity of the large/medium-sized public buildings, the advanced control and optimization problems for indoor environment have been paid more and more attention in both engineering and academia fields. Addressing the contradiction between the accuracy and complexity of the current building environment model, and consequently difficulty of the controller design, the rapid modeling, low-dimensional control and optimization problems for the building environment are investigated by introducing the model reduction method. Four parts are mainly included: 1. Considering the multiphase structure and multi-parameter coupling of the building environment system, a multi-scale and low-dimensional environment model for control issue is established by combining the singular perturbation theory with the model reduction method; 2. An adaptive disturbance rejection controller is designed based on the developed low dimensional model; Taking advantage of the air stratification caused by natural convection, an efficient and low-dimensional environmental optimal control law is constructed; 3. The multi-parameter optimization scheme of building environment is built with low cost and high spatial resolution; 4. An experimental platform is set up to verify the effectiveness and applicability of the reduced environmental modeling and control method. Through alternate promotions of theory, simulation and experiment, this project aims to provide an efficient and precise environmental modeling, control and optimization strategy for large/medium-sized buildings, which will promote the application and development of advanced building environment control technologies.

随着大中型公共建筑用能强度的持续增长，建筑室内环境的先进控制与优化技术日益受到学界与工程界重视。本项目针对当前建筑环境模型精度与复杂度之间的矛盾以及由此带来的控制器设计困难，通过引入模型降阶方法，围绕建筑环境的快速建模、低维控制与优化问题展开研究。具体包括：一、针对建筑环境系统多相结构、多参数耦合等特点，将奇异摄动理论与降维建模方法结合，建立面向控制的多尺度低维环境模型；二、研究基于低维模型的环境自适应抗扰控制器，并利用自然对流引起的垂直空气分层，设计高效率环境最优控制律；三、研究计算代价低、空间分辨率高的建筑环境多参数优化策略；四、搭建实验平台对低维环境建模与控制器设计方法的有效性和实用性进行验证。最终，通过理论、仿真和实验的交替推进，本项目旨在为大中型建筑提供一种高效、精确的环境控制和优化方案，促进先进控制技术在建筑环境与能量领域的实际应用与发展。

随着国内大中型建筑规模的日益增长，其室内环境调节和用能相关的控制与优化技术日益受到学界与工程界重视。本项目针对多参数复杂建筑环境的高效建模和优化控制问题，围绕多参数建筑环境低维建模与优化控制、基于有限传感器的室内温度场低维重构和估计、建筑室内微气候优化调控与补偿策略、建筑能量系统需求建模与预测等四个方面展开研究工作。具体包括：一、研究建筑室内多种环境参数时空变化规律和内部耦合关系，运用本征正交分解（POD）方法进行快速低维环境重构。以温室建筑为背景，运用计算流体力学方法获取温室环境系统关键参数的变化空间信息，运用POD法进行最优子空间投影。利用重构出的POD子空间，结合群智能进化算法对温室作物生长环境实现快速多目标优化。结合作物生长模型，设计温室系统的有限时域滚动优化控制策略；二、利用实际大空间建筑，对低维环境建模的有效性与稳定性进行实验验证。在此基础上研究得到环境传感器位置/数量的启发式优化方法，并基于有限传感器研究室内温度场低维重构和估计方案；三、面向建筑环境控制的人体热舒适度识别和微气候调控。结合实验和机器学习等手段研究人体舒适度与生理信息的相关性和识别方法。采用双目摄像头、热成像仪等非侵入式设备结合机器学习算法快速识别人体热舒适状态。在此基础上，利用珀尔贴热电效应，在人体周围补偿出一个局部热环境，通过优化调控策略，提高个体热舒适度同时降低建筑整体能耗。四、运用无监督/有监督聚类法、主元分析法以及递归特征消除法等多种数据分析手段研究建筑能量系统中的数据不平衡、特征提取等问题。运用多个子学习器建立集成学习（Ensemble learning）框架,结合线性回归和贝叶斯回归进行结果融合。运用迁移学习技术研究建筑能量建模中的数据共享及数据高效利用问题，设计基于实例的迁移学习模型用于数据稀疏的建筑能量需求精确建模。