Thermal comfort modeling when personalized comfort systems are in use: Comparison of sensing and learning methods
Thermal comfort modeling when personalized comfort systems are in use: Comparison of sensing and learning methods
复制标题
DOI：
10.1016/j.buildenv.2020.107316
复制
发表时间：
2020-11-01
影响因子：
7.4
通讯作者：
Becerik-Gerber, Burcin
中科院分区：
文献类型：
--
作者：
Aryal, Ashrant;Becerik-Gerber, Burcin
当同一区域有多个居住者时，集中式 HVAC 系统通常无法满足个人需求。个性化舒适系统（PCS），例如局部风扇和加热器、加热/冷却椅子、局部通风系统，已被证明有助于通过在每个居住者周围创造微气候来维持舒适的热条件。以往的研究主要集中在常规HVAC操作下的个性化热舒适度建模，缺乏针对PCS设备使用时的个性化热舒适度建模的工作。在这项研究中，我们比较了不同的传感和机器学习方法，以在使用本地风扇或加热器时建立个人舒适度模型。该实验在受控环境中进行，分为三个部分：常规（无风扇/加热器）、风扇打开和加热器打开。我们的结果表明，与使用可穿戴设备监测手腕皮肤温度或热成像监测面部不同区域皮肤温度相比，来自环境传感器的数据的预测精度提高了 2%-5%。此外，与生理传感器相比，环境传感器更便宜，并且具有相对较少的隐私问题。总的来说，这项研究的结果支持使用环境传感器来构建个性化的热舒适模型，无论是否有 PCS。此外，结果还强调了在使用 PCS 设备和不使用 PCS 设备时建立单独的个性化热舒适模型的必要性。
Centralized HVAC systems are usually unable to cater to individual requirements when multiple occupants are present in the same zone. Personalized Comfort Systems (PCS) such as local fans and heaters, heated/cooled chairs, local ventilation systems, have shown to be useful for maintaining comfortable thermal conditions by creating a microclimate around each occupant. Previous studies have mostly focused on personalized thermal comfort modeling under regular HVAC operations, and there is a lack of work that focuses on personalized thermal comfort modeling when PCS devices are in use. In this study, we compare different sensing and machine learning methods to build personal comfort models when a local fan or heater is in use. The experiment was conducted in a controlled environment with three segments: regular (no fan/heater), fan on, and heater on. Our results indicate that the data from environmental sensors results in 2%-5% higher prediction accuracy compared to using a wearable device to monitor wrist skin temperature or thermal imaging to monitor skin temperature from different regions of the face. Furthermore, environmental sensors are more affordable and have relatively fewer privacy concerns compared to the physiological sensors. Overall, the results of this study support the use of environmental sensors for building personalized thermal comfort models with or without PCS. Furthermore, the results also highlight the need for building separate personalized thermal comfort models when PCS devices are in use, and when PCS devices are not in use.