Loading...
Please wait, while we are loading the content...
Similar Documents
Real-time Building Occupancy Sensing for Supporting Demand Driven HVAC Operations
| Content Provider | Semantic Scholar |
|---|---|
| Author | Ekwevigbe, Tobore Brown, Neil Pakka, Vijayanarasimha Fan, Denis |
| Copyright Year | 2013 |
| Abstract | Accurate knowledge of localised and real-time occupancy numbers can have compelling control applications for Heating, Ventilation and Air-conditioning (HVAC) systems. However, a precise and reliable measurement of occupancy still remains difficult. Existing technologies are plagued with a number of issues ranging from unreliable data, maintaining privacy and sensor drift. More effective control of HVAC systems may be possible using a smart sensing network for occupancy detection. A low-cost and non-intrusive sensor network is deployed in an open-plan office, combining information such as sound level and motion, to estimate occupancy numbers, while an infrared camera is implemented to establish ground truth occupancy levels. Symmetrical uncertainty analysis is used for feature selection, and selected multi-sensory features are fused using a neuralnetwork model, with occupancy estimation accuracy reaching up to 84.59%. The proposed system offers promising opportunities for reliable occupancy sensing, capable of supporting demand driven HVAC operations. INTRODUCTION Global warming is one of the most disturbing concerns facing humanity today due to the accelerated release of carbon dioxide (CO2) and other greenhouse gases into the atmosphere as a result of human activities. The problem is compounded by decreasing availability of fossil fuels, increasing population, environmental and economic concerns regarding energy use. These all constitute drivers for the adoption of more sustainable ways of securing our energy needs (Shuai et al., 2011). Approximately about 40% of the world’s energy is consumed by buildings (ASHRAE, 2007), of which roughly about half of this energy is consumed by Heating, Ventilation, and Air conditioning (HVAC) systems (Perez-Lombard et al., 2008). Reductions in HVAC related energy will go a long way in contributing to efforts aimed at delivering sustainable building energy use. Previous research have proposed up to 56% HVAC related energy savings with improvements in operation and management of HVAC systems (Sun et al., 2011, Tachwali et al., 2007). Realtime building occupancy sensing is useful for efficient control of building services such as lighting and ventilation, enabling energy savings, whilst maintaining a comfortable environment. Occupancy information can be used for determination of HVAC heat loads (Chenda and Barooah, 2010), system running time, required heating, cooling and distribution of conditioned air, and optimal selection of temperature set points (Li et al., 2012). Ideally, building controls should automatically respond to dynamic occupancy loads. However, current building energy management system (BEMS) often lacks this capacity, as such they usually rely on fixed assumptions (such as peak occupancy loads as opposed to the optimal) to operate HVAC and electrical systems, leading to possible energy waste (Erickson et al., 2011). One possible solution for achieving energy efficiency in buildings is to couple real-time occupancy information to building controls, such that services are provided only when needed (i.e during occupied instances), and to optimize HVAC operations such that the flow rate of conditioned air into a space is adjusted based on optimal occupancy numbers. Many occupancy detection systems in the literature have certain drawbacks with respect to accuracy, cost, intrusiveness, and privacy. This study attempts to address these limitations by fusing information from a network of low-cost sensors for building occupancy detection. This study is distinguished from previous research in that it introduces the use of symmetrical uncertainty analysis for feature selection, and a genetic based search to evaluate an optimal sensor combination for occupancy estimation. It goes further to investigate a new method of occupancy sensing: the use of case temperature. To the best of the authors' knowledge, these tools have not been examined for occupancy detection. BACKGROUND Conventional occupancy detection systems have several short-comings; Passive infrared (PIR) sensor is the most commonly used technology for occupancy sensing in non-domestic buildings especially for lighting control (Delaney et al., 2009), however it fails to detect stationary occupants, thus switching off services falsely. To address this problem, PIR sensors are coupled with other sensors. For instance, Padmanabh et al. (2009) used a combination of microphones and PIR sensors to gather occupancy information for efficient scheduling of a conference room. The room was considered occupied (meeting ongoing) if a microphone value exceeded a threshold twice in a 5-minute interval; otherwise it is classified as unoccupied (no meeting). Jianfeng et al. (2005) built a novel bathroom activity recognition system, consisting of microphone and PIR sensors. The system detected real-time sound events with a hidden Markov model to an accuracy of 87%. Both systems highlight the usefulness of sound sensing for activity monitoring, and hence occupancy presence detection. However, system functionality is prone to external interference which may limit their performance. Dodier et al. (2006) proposed a Bayesian belief network comprising of three PIR sensors and a telephone sensor to probabilistically infer occupancy. Occupied state of individual offices room was modelled with a Markov chain. Their system had a detection accuracy of 76%, but was unable to count the number of occupants. In an attempt to improve the robustness of occupancy numbers detection, Dong et al. proposed a system that used information from CO2, acoustic and PIR sensors to estimate the number of occupants in an open-plan office space (Dong et al., 2010, Lam et al., 2009b, Lam et al., 2009a). Using information theory, the most relevant information for occupancy prediction was extracted from sensor data, and fused with three machine learning algorithms (support vector machine, artificial neural networks, and hidden Markov model). An average reported accuracy of 73% was achieved by the hidden Markov model. Meyn et al. (2009) improved occupancy detection accuracy, by using a sensor network comprising CO2 sensors, digital video cameras and PIR detectors as well as historic building utilization data for occupancy estimation at the building level. The system used a receding-horizon convex optimization algorithm to infer occupancy numbers. Their system detection accuracy reached 89%. However, it was not able to estimate occupancy numbers at the room level. Other studies have highlighted the feasibility of occupancy detection in offices by monitoring office equipment usage. For example, Brown et al. (2011) proposed a useful method for establishing the usage pattern of electrical appliances (such as desktop PCs), from which occupancy can also be inferred. Using portable temperature sensors attached to the case of PCs, and a pinging software routine that runs on the local network, appliance duty cycles were detected to a precision in excess of 97%. Martani et al. (2012) studied the relationship between building occupancy and energy consumption, using number of WiFi connections as proxies for occupancy estimation. Overall, at the building level, occupancy accounted for between 63% and 69% variation of the total electricity consumption. Both systems were unable to detect occupants not using a computer. The use of wearable sensors for monitoring occupants have also been reported; Li et al. (2012) proposed an occupancy detection system based on RFID tags, which reported real-time occupancy numbers and their localized thermal zones in an office building. A K-nearest neighbour algorithm was used for occupancy tracking to average zone detection accuracy of 88% for stationary occupants and 62% for moving occupants. The system offers promising potentials for occupancy monitoring. However, occupants’ willingness to wear the devices may be a critical factor for its uptake, especially in office buildings. Vision-based systems have also been used (Benezeth et al., 2011, Tomastik et al., 2008), although occupants' privacy is a concern. Besides, their applicability is limited in heavily partitioned spaces. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://oaktrust.library.tamu.edu/bitstream/handle/1969.1/151431/ESL-IC-13-10-24.pdf?isAllowed=y&sequence=1 |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
