Keywords

1 Introduction

Smart buildings with the introduction of indoor environment management systems are state of the art (Manic et al. 2016). They have integrated monitoring and communication infrastructure that consists of smart devices such as sensors, which are connected to the Internet of Things (IoT). Building management systems (BMS) aim to optimize the energy consumption in buildings, and they manage the critical components such as heating, ventilation, and air conditioning (HVAC), gas, lighting, security system, electricity, and fire system, in communication with IoT distributed devices. However, according to (ASHRAE 2016) studies, the people who work in offices spend about 60–90% of their time indoors in buildings; thus, their comfort and health effects are strongly related to the indoor environment performance and quality evaluations. Since they focus on doing their job and usually don’t reflect their energy consumption (Lo et al. 2012; Jia et al. 2019), they may influence the evaluation’s performance systems both actively as they adjust thermostats or operate windows and doors and in a passive sense as they optimize indirectly the use of the buildings services, such as HVAC systems, lighting systems, security systems, elevators, water systems, and other building services.

This research provides a concept and design of an ongoing work of monitoring management tool concept and design based on an indoor building performance model, which incorporates fundamental characteristics of the indoor environment, such as indoor temperature and the working peoples’ productivity, adopting the agent-based modeling approach. The core of the monitoring tool is the IoT, which consists of sensors collecting a huge amount of data. At present, such IoT monitoring tools are used for improving the energy consumption of buildings via smart HVAC control. In this respect, our challenge is to achieve monitoring workers’ behaviors in a minimally intrusive way. That’s why we use existing infrastructure in the buildings, where the sensors are integrated and distributed separately, not requiring any disturbance to the working ones, and to develop effective data evaluations improving monitoring management system.

In the following section, we present various strategies and some observations improving our survey. In Sect. 20.3, we introduce the evaluation part by presenting the three actual characteristics of our tool: (1) data visualization, (2) analysis of indoor temperature collected data and predict system behaviors to react beforehand analysis, and (3) thermal sensation analysis in order to improve working peoples’ (hereinafter referred to as workers) productivity. Finally, we draw conclusions and recommendations for practice and future research.

This work is realized within the context of the iCity project. It is sponsored by the German Federal Ministry of Education and Research “FH-Impuls 2016” under Contract 13FH9I01IA.

2 Related Work

Many recent attempts have focused on studying the relation between indoor environment and workers’ behaviors in order to improve energy consumption, building’s quality, workers’ comfort, etc. Thus, various strategies were adopted to represent the essential elements effecting the occupant comfort and the energy consumption performance in the indoor environment. Indeed, researches are distributed in different axes. Some researches (Yang et al. 2013) are based on deterministic data technologies, wherein the data are collected via distributed sensors, sent through networks, and logged. With the support of sensor technologies, other research teams use modeling methodologies based on collected data of the distributed sensors. The modeling methodology is represented, for example, by statistical analysis, which consists of calculating the relationship between occupant behavior and indoor/outdoor environment conditions, electricity usage, or time period, whose results functions return the occupancy state or the probability of studied behaviors. For example, by using weather stations, occupancy sensors and digital photography collect data related to the indoor environment, occupant presence, and position of shading and windows in five office buildings of Austria (Lenoir et al. 2011). They analyzed the relationship between these parameters and deduced that such interactions are difficult to predict at the level of an individual person. However, they concluded that long-term general trends for groups of building occupants can be expressed as a function of indoor and outdoor environmental parameters. Peng et al. (2011) describe occupant behaviors by a quantitative method. They divided occupant behaviors into three types according to the usage with time dedicated, environment dedicated, and random modes. They used environment and user feedback, or probability and time step, to track the behaviors as a function of the above parameters. Finally, they assumed the effect of human behavior on building and energy use. Moreover, others (Jia et al. 2019; Fabi et al. 2014; Klein et al. 2012) integrated methods of statistics and identified indoor environmental factors, and not real-time environment data, that may change the occupant behavior, e.g., window opening or light switch; however, such a behavior may strongly relate to psychological or social conditions. In this sense, today, researchers integrate artificial intelligence (AI) approaches that range from machine learning that correlates behavioral inputs with buildings’ collected data to agent-based modeling (ABM). ABM is a computational model for simulation of object interaction with each other and the indoor environment. In our IoT monitoring management tool research, we adopt ABM.

Observations

According to the existing literature, there are many methodologies used to model the workers’ behavior toward improving the energy efficiency of buildings. Therefore, the authors offer here a discussion reflecting their opinion and fortify their line of research.

Some research analyzed the effect of workers’ behavior on building energy consumption based only on some hypothetical constraints and not on real measured data. For instance, Peng et al. (2011) supposed the existence of three typical lifestyles of workers without real data source. This research efficiency may be needed to combine occupants’ environment real-time data and behavioral tendency data that may be obtained through a monitoring tool. Other researchers, to optimize the energy performance of buildings, have realized the importance of tracking workers’ status, i.e., the status of space being occupied or unoccupied. Their aim is to create a worker behavior pattern that integrates the actual operation schedule optimizing indirectly the building services, such as HVAC and lighting systems. However, in addition to the indirect effect, Jia et al. (2019) added the direct effect of the worker behavior on the energy performance of buildings, such as adjusting thermostats or operating windows and doors. In the literature, we found more research that treats the indirect effect, while we believe that a more detailed direct effect study may potentially reduce the energy consumption of buildings. The aforementioned observations show that there is a relationship between workers’ behavior and the factors that initiate their actions (or behavior), needless to say, that the collected data are insufficient to offer a complete evaluation. Fabi et al. (2014) aim to show that indoor environment factors that initiate the worker’s behavior may result in a high probability of worker behavior change, for example, window opening; however, people mostly seem to act according to an individual decision, sometimes psychological, and rarely according to an environmental factor. To mention that, logically, if a worker behaves according to an environmental factor, this behavior is efficient and performs one or more building services; therefore, it affects and optimizes directly the monitored factors of performance in real time.

3 Setup and Overview of the IoT Monitoring Tool

The IoT monitoring tool is an analytic platform that converts the IoT collected data into insights and simplifies the decision-making process. It uses agent-based modeling (ABM). The IoT monitoring tool should be able to (1) analyze and predict system behavior to react beforehand, (2) share the analysis with the other users by embedding visualizations into a dashboard, (3) and analyze the productivity of office work.

Data were collected using sensor monitoring of a building of Applied Sciences’ university in Stuttgart, Germany. Monitoring concerned indoor and outdoor temperature. Parameters were recorded at 5-min intervals and then sent to a local server. Data concerns the outdoor temperature, as well as the indoor temperature average to a different location in the room. Data were collected for the period between January and July 2016. During this monitored period, the outdoor temperature varied between −5 °C and 38 °C, while the indoor temperature varied between 21 °C and 33 °C. The monitoring agents filter data according to conditions, for example, excluding days such as holidays, and/or the greatest similar days.

To visualize into dashboards, and share data analysis with other users, ThingsBoard IoT platform (ThingsBoard 2016) was adopted as IoT technology in our research. ThingsBoard is an open source for data collection, processing, visualization, and device management. IoT monitoring implements few methods for time-series prediction as linear regression used for forecasting the temperature. Figure 20.1 shows a comparison of predicted and recorded indoor temperatures. A good agreement is observed between recorded temperature and linear regression prediction. It has previously been established that indoor environmental quality influences human performance; feeling comfortable at the work office is one of our requirements as workers to improve productivity performance; this performance factor is influenced by many psychological, social, or environmental parameters, and whatever the parameters that can be taken in consideration, all possible parameters vary with time.

Fig. 20.1
figure 1

Simple predicted method

Full size image

Through several studies on performance related to office work, Seppänen et al. (2006) identified a relation between performance and temperature. Various metrics of performance were used in these studies. Field studies used a work task as a metric of performance, in call centers the talk time or using command “regress” in Stata for windows (a program that selects the best fitting linear model of dependent variable on explanatory variables); Seppänen et al. (2006) fit quadratic model to the data for normalized percentage change in performance vs temperature unweighted, weighted by simple size, and weighted by combined final weight separately. They deduced this equation for the curve with composite weighting factors:

$$ p=0.1647424\ T-0.0058274\ {T}^2+0.0000623\ {T}^3-0.4685328 $$

where p is productivity relative to maximum value presenting the worker thermal sensation and T is indoor temperature, °C. This equation is used for forecasting the workers’ productivity considering only the indoor temperature as an input parameter.

Figure 20.2 shows a comparison of real workers’ productivity and predicted productivity, (a), according to the collected indoor temperature, and (b), according to regression-based predicted indoor temperature values. A good agreement is observed between recorded and predicted values. Both cases show a consistent decrease in performance of typical office tasks when temperature increases above 26 °C.

Fig. 20.2
figure 2

(a) Normalized performance (thermal sensation) vs room temperature, (b) predicted thermal sensation vs predicted room temperature

Full size image

This analysis did not include different indoor environment parameters, such as the number of workers, devices, windows, and doors in a room. These parameters could have a significant impact. However, as the next step, the activities that have a significant impact could be monitored and included in the analysis model.

4 Conclusion

This paper proposes a methodology for the development of a simplified regression-based model for forecasting indoor temperature and predicting the workers’ thermal sensation. It shows relevance analysis and leads to a simplified forecasting model with restricted input parameters. Data included outdoor and indoor temperature and humidity. Analyses showed that the indoor temperature and the thermal sensation forecasting could be conducted with good precision using only the indoor temperature history. This result could not be generalized. However, the proposed methodology could be used for improving the indoor quality and the workers’ performance. Available data did not include indoor activities. The presence of significant activities should be considered to improve the forecasting model analysis.