Occupancy patterns obtained by heuristic approaches: Cluster analysis and logical flowcharts. A case study in a university office
Occupancy patterns obtained by heuristic approaches: Cluster analysis and logical flowcharts. A case study in a university office
复制标题
DOI：
10.1016/j.enbuild.2019.01.023
复制
发表时间：
2019-03-01
影响因子：
6.7
通讯作者：
De Simone, Marilena
中科院分区：
文献类型：
--
作者：
Mora, Dafni;Fajilla, Gianmarco;De Simone, Marilena
在办公室内建立了一个实验装置，目的是通过监测占用状态、空气温度、相对湿度、CO2、VOC、门窗开度和用电量来获得有关占用模式的信息。启发式的方法：聚类分析和模型的基础上，逻辑流程图。在地面实况占用数据中实施聚类分析，以通过考虑不同的时间步长来识别每日占用模式。确定了每日入住率低于和高于40%的集群。此外，在高入住率的聚类中，分析区分了最高入住率低于或大于40%的组，并采用连续参数验证了传感器复制每个识别聚类特征的能力。CO2和电力集群在集群数量、每个集群中的天数和占用率方面表现出相似性。此外，在基于逻辑流程图的模型中使用连续变量和二元变量来描述每小时的占用情况。使用一个参数的最佳解决方案返回的误差为12%，使用两个参数的误差为10%。具有三个参数的模型显示误差小于10%，添加第四个参数并没有显着提高准确性。(C)2019 Elsevier B. V.版权所有。
An experimental set-up was built in an office with the aim of obtaining information regarding occupancy patterns by monitoring occupancy state, air temperature, relative humidity, CO2, VOC, door and window opening, and electricity usage. Heuristic approaches were applied: cluster analysis and models based on logical flowcharts. Cluster analysis was implemented in the ground truth occupancy data to identify daily occupancy patterns by considering different time steps. Clusters marked by daily occupancy lower and greater than 40% were identified. Furthermore, in high occupancy clusters, the analysis distinguished groups in which the day with the highest occupancy was lower or greater than 40%.The same approach was applied with continuous parameters to verify the ability of sensors to replicate the characteristics of each identified cluster. CO2 and power clusters showed similarities in the number of clusters, days in each cluster, and occupancy percentage.In addition, both continuous and binary variables were used in models based on logical flowcharts to describe hourly occupancy profiles.The best solution with one parameter returned an error of 12%, by using two parameters an error of 10%. Models with three parameters showed errors of less than 10%, accuracy did not improve significantly by adding the fourth parameter. (C) 2019 Elsevier B.V. All rights reserved.