A comprehensive review of approaches to building occupancy detection

Highlights

• A comprehensive review of the techniques that have been exploited to estimate building occupancy is presented.

• Building occupancy detection techniques were categorized as analytical, data-driven, and knowledge-based methods.

• A comparison framework to help readers to better benchmark occupancy detection systems is presented.

• Some potential future research directions are discussed.

Abstract

Detailed occupancy information in buildings is useful to improve the performance of energy management systems in order to enable energy consumption savings and maintain occupants' comfort. Different technologies employed to provide occupancy information account for high-precision devices such as optical and thermal cameras, and environmental or specialized sensors like carbon dioxide (CO₂) and passive infrared (PIR). Although the latter systems have lower accuracy, they have received significant interest due to their affordable and less-intrusive nature. Accordingly, various studies have been conducted to explore the various elements of these technologies. Nevertheless, the algorithmic aspect of the occupancy detection process has not been adequately taken into consideration. This paper presents an extensive review of the techniques that have been exploited to process the information provided by the sensors and carry out occupancy information detection. In this study, a complete set of comparison criteria, comprising the performance, the occupancy resolution, the type of sensors used, the type of buildings, and the energy saving potentials has been considered in order to perform an in-depth analysis of the occupancy detection systems. Through its examination, this paper elaborates significant remarks on occupancy detection algorithms in order to realize a method that is not only efficient in processing sensors’ data but also effective in providing accurate occupancy information.

Introduction

Smart grid is the key enabler of revolutionary power grids with a flexible, efficient, durable, and safe infrastructure. The emerging trends in communication, sensing, and metering technologies under smart grid archetype can enable an optimal management of all the units constituting power systems. Among the power grid sectors, the building sector receives significant attention due to its highest share of electricity consumption and potential of energy saving. Several researchers have shown the importance of occupancy information to improve building energy performance and reduce building energy consumption [1,2]. Agarwal et al. [3] proposed a heating, ventilating, air-conditioning (HVAC) control strategy that turns ON or OFF the system based on the occupancy detection in offices. The simulation results indicated that by using the proposed approach, the HVAC energy consumption was reduced between 10% and 15%. Jin et al. [4,5] reported a saving potential of 55% of the energy consumed by the ventilation system, by using a control ventilation strategy based on occupancy information. Leephakpreeda [6] proposed an occupancy-based lighting control, and showed that the energy consumption of the system can be reduced between 35% and 75%. Yokoishi et al. [7] showed that it was possible to save 3.5 h of lighting power per day in a campus room, by using a network of PIR and illumination sensors for occupancy detection. Scott et al. [8] used information collected by PIR and RFID sensors to predict home occupancy patterns for heating control. The authors analyzed three different control techniques (based on scheduled, always-on, and PreHeat algorithms), and consequently obtained an energy saving potential up to 35%. Pend et al. [9] proposed a strategy with learning capacity for the control of the cooling system of six offices. The results showed that 20.3% of energy could be reduced by using the proposed control strategy. In this regard, Building Energy Management Systems (BEMS) are promoted in order to facilitate the utilization of occupancy information to minimize the power grid stress and maintain users’ comfort [10,11].

Occupancy is not the only factor that influences electricity consumption, there are physical factors including building characteristics, equipment efficiency and weather conditions that can change the electricity consumption behavior of a building [12,13]. However, these factors cannot be easily handled or modified by humans during the building utilization. Actually, occupancy is an important part of human factors that describe the occupants' presence, their consumption patterns, and the indoor environmental conditions. The concept of ‘‘occupancy’’ is regarded as the primary level of occupant behavior modeling [14]. Occupancy information can account for three main features that are explained in different resolution levels regarding the targeted applications [15]:

• Temporal resolution: it indicates the frequency at which events occur (e.g., hours, minutes, and seconds).

• Spatial resolution: it expresses the building properties in terms of number of floors, rooms, and other relevant data.

• Occupancy resolution: it explains the presence/absence status, the number of people, as well as their identification and activity in a building's zone.

The technologies and methodologies used to estimate buildings’ occupancy information have been explored in literature. However, the algorithmic aspect of the building occupancy detection has not been completely analyzed in these works. Zhang et al. [2], Jia et al. [14], Kjaergaard et al. [16] and Balvedi et al. [17] summarized and compared different techniques for the acquisition of occupancy information. Likewise, these reviews presented a general analysis of the methods used to model the presence and behavior of occupants, as well as their influence on the energy consumption of buildings. Yang et al. [18] presented the opportunities and challenges of the occupancy sensing systems and occupancy modeling methodologies applied to institutional buildings. Mane et el [19]. briefly discussed the research trends and gaps in occupancy sensing. Labeodan et al. [20] published a survey of the occupancy measurement systems that are used in office buildings. The authors make a classification of the occupancy detection systems, using their spatial-temporal properties as comparison criteria. Saha et al. [21] presented a brief description of popular algorithms used for the detection, counting and tracking of occupants in buildings. Chen et al. [22] and Sun et al. [23] conducted an analysis of occupancy estimation and detection systems based on the type of the utilizing sensor. Likewise, they compared the sensors in order to identify their advantages and limitations for occupancy detection.

As presented above, it can be comprehended that the review papers about occupancy detection and estimation systems have explored the elements such as sensor type, cost, privacy issues, performance, energy savings, limitations, and infrastructure, as the essential prerequisite for occupancy detection. However, the analysis of the type of occupancy detection methods and their influence on building occupancy detection performance has not been adequately into consideration. Accordingly, this paper presents a categorization of the techniques that have been exploited to estimate building occupancy information. Additionally, a complete set of comparison criteria, comprising the performance, the occupancy resolution, the types of sensors used, the type of building, and the energy saving potential has been employed in order to perform an in-depth analysis of the occupancy detection methods.

The rest of this study is organized as follows: Section 2 presents the search method that has been used for the selection of the papers, discussed in the review. Section 3 presents the classification and analysis of the occupancy detection algorithms presented in the state of the art. Section 4 describes the advantages and disadvantages of sensors and algorithms, and benchmarks the performance of occupancy detection systems. Section 5 describes the existing trends and presents future research directions to encourage more improvements to this field. Finally, the conclusions are presented in Section 6.