根据美国教育部的数据，美国大约有37100所中学。教室建筑具有特殊的居住行为，例如突然的供暖和制冷需求，以及随着课程安排频繁切换房间功能。由于学校的负荷特点，使用大容量的制冷机和锅炉提供供暖和制冷。此外，这些学校的办公室和教室混合在一起，使得空气循环变得复杂。实际上，与所有其他商业建筑一样，由于大型复杂暖通空调系统的维护不当，这些供暖、通风和空调（HVAC）系统中的大多数都处于故障或非最佳运行状态。 此类暖通空调系统中存在数百个可能的故障点，尤其是老化系统。与无故障系统相比，单个故障可能会导致100%以上的能量损失。然而，就美国不同地区的多处故障对中学建筑能耗的影响而言，故障影响分析仍处于初级阶段。本研究采用了大规模故障分析框架来调查故障对中学建筑能耗的影响，使用DOE原型商业建筑作为4个气候区（即芝加哥、亚特兰大、迈阿密和旧金山）的基线。在每个基线中注入553种故障模式，通过改变故障相关参数来构建故障案例。 通过3000个样本的随机方法，每个故障都基于特定的故障发生分布。在高性能集群中，使用了完整的建筑模拟程序EnergyPlus进行模拟。基于深度学习的响应面被用于建立故障相关参数与建筑能耗/热舒适性之间的联系。使用Sobol指数进行灵敏度分析，以获得能量惩罚方面最关键的故障。模拟结果表明，在所有4个气候区中，最关键的故障是中学的冷却器污垢和锅炉污垢。这表明，对于拥有大型暖通空调系统的中学来说，制冷机和锅炉的日常维护对于节能非常重要。 此外，除冷水机组污垢和锅炉污垢之外的关键故障在区域级别的4个气候区域中各不相同。引用:2019年年度会议，密苏里州堪萨斯城，会议论文

According to the US Department of Education, there are about 37,100 secondary schools in the United States. The classroom buildings have special occupant behaviors, such as sudden heating and cooling demands, and frequent switches of room function with the class schedules. Due to the load characteristics of schools, the chillers and boilers with large capacities are used for providing heating and cooling. Furthermore, the mixing of offices and classrooms in those schools, makes the air loops complicated. In reality, like all other commercial buildings, the majority of those heating, ventilation, and air-conditioning (HVAC) systems are in faulty or non-optimal operating conditions due to the improper maintenances for large and complex HVAC systems. Hundreds of possible fault points exist in such HVAC systems, especially the aging systems. It is possible that a single fault could cause 100% more energy penalty compared to fault-free systems. However, the fault impact analysis is still in the infant stage in terms of the impacts on building energy consumption of the secondary school under multiple faults in different regions of the U. S. This study applied a large-scale fault analysis framework to investigate the fault impacts on the secondary school, using the DOE prototype commercial buildings as the baselines in 4 climate zones (i.e., Chicago, Atlanta, Miami, and San Francisco). There are 553 fault modes injected into each baseline to build the faulty case by changing the fault associated parameters. Each fault is based on a certain fault occurrence distribution through a stochastic approach of 3,000 samples. A whole building simulation program, EnergyPlus, was used for the simulation in the high-performance cluster. The deep learning based response surface was applied to build the connections between the fault associated parameters and building energy consumption/thermal comfort. The sensitivity analysis with a Sobol index was performed to obtain the most critical faults in terms of energy penalty. The simulation results showed the most critical faults are the chiller-fouling and boiler-fouling for the secondary schools in all 4 climate zones. This indicates the routine maintenance for chillers and boilers are important for energy savings for the secondary schools with large HVAC systems. In addition, the critical faults excluding the chiller-fouling and boiler-fouling are varied among the 4 climate zones at the zone level.