Researchers are working towards incorporating occupancy detection into smart thermostats by determining the number of people in each HVAC zone. Occupancy detection allows the HVAC system to begin temperature adjustments before actual air temperature changes occur which allows a deeper setback temperature while maintaining comfort.
Real time data collection can use occupancy data to generate an input value for smart thermostat programming. Research on developing occupancy detection systems for high occupancy buildings is often simulation based after the data collection portion. Program trials are significantly easier and more cost efficient to complete in simulations than in physical environments. Completed results appear promising; several studies reported estimated energy savings between 10% and 60% by incorporating occupancy detection. The wide range in savings is due to variances in the original HVAC programming and the type of occupancy data that was utilized [16, 17, & 18].
Real time data collection can use occupancy data to generate an input value for smart thermostat programming. Research on developing occupancy detection systems for high occupancy buildings is often simulation based after the data collection portion. Program trials are significantly easier and more cost efficient to complete in simulations than in physical environments. Completed results appear promising; several studies reported estimated energy savings between 10% and 60% by incorporating occupancy detection. The wide range in savings is due to variances in the original HVAC programming and the type of occupancy data that was utilized [16, 17, & 18].
SENSOR SYSTEMS
Many of the systems that have been made available for real life trials use PIR sensors in conjunction with other sensors and environment sampling technology. The systems are designed to continuously extrapolate the number of people within each area from the combination of multiple data points. These systems are more accurate than systems that rely solely on PIR sensors, but they are very expensive, running much higher than potential savings for the average company. They also require individual designs specific to each installation which requires a lengthy and complex installation process. The multitude of data sampling techniques and data processing equipment results in a higher manufacturing footprint and higher energy consumption than a PIR sensor array.
Because of the listed factors, this project focuses on research that uses only PIR sensors. Research projects have made progress on improving system arrangement and accuracy which has proven positive enough to continue the subject. Systems that only use PIR sensors use a different type of occupancy detection than multiple sensor systems. PIR sensors only detect moving heat sources, so the Gatekeeper method is used. Rather than constantly determining stand-alone occupancy values, the system calculates occupancy values based off the previous value. Sensors are placed on the boundaries of HVAC zones and +1 value is generated every time a person's entrance is detected and a -1 value is generated every time a person's exit is detected in a manner analogous to how turnstiles keep track of the number of people inside an amusement park. This setup requires the sensors to have suitable ranges for their locations. PIR sensors are manufactured in a variety of ranges, but range of only a few feet is optimal to avoid false detection. Doorways are often "gates" between HVAC zones so sensors placed above them can detect occupancy changes. Short ranges help ensure that only people passing through the doorway are detected and people walking past the door remain outside the range. Short ranges also minimize the effects of masking time [footnote 1]. PIR sensors are easily mass produced and they involve have a small manufacturing footprint due to their small size. Figure 4 shows a common size. The sensors must be mounted on a wall, and the cases are also small and unobtrusive. Figure 5 shows an operating sensor installed in a wall.
Because of the listed factors, this project focuses on research that uses only PIR sensors. Research projects have made progress on improving system arrangement and accuracy which has proven positive enough to continue the subject. Systems that only use PIR sensors use a different type of occupancy detection than multiple sensor systems. PIR sensors only detect moving heat sources, so the Gatekeeper method is used. Rather than constantly determining stand-alone occupancy values, the system calculates occupancy values based off the previous value. Sensors are placed on the boundaries of HVAC zones and +1 value is generated every time a person's entrance is detected and a -1 value is generated every time a person's exit is detected in a manner analogous to how turnstiles keep track of the number of people inside an amusement park. This setup requires the sensors to have suitable ranges for their locations. PIR sensors are manufactured in a variety of ranges, but range of only a few feet is optimal to avoid false detection. Doorways are often "gates" between HVAC zones so sensors placed above them can detect occupancy changes. Short ranges help ensure that only people passing through the doorway are detected and people walking past the door remain outside the range. Short ranges also minimize the effects of masking time [footnote 1]. PIR sensors are easily mass produced and they involve have a small manufacturing footprint due to their small size. Figure 4 shows a common size. The sensors must be mounted on a wall, and the cases are also small and unobtrusive. Figure 5 shows an operating sensor installed in a wall.
SENSOR ARRAY
A simplified model of a buildings sensor array is displayed in Figure 6 and will be referenced throughout this description. As previously stated, sensors are placed at zone "gates". Each gate has a pair of sensors to determine objects' direction of movement based on the order the two sensors are triggered [15]. For example, if the blue sensor in the doorway from the Hallway into one of the Office is triggered before the green sensor, 1 is added to the Office's occupancy and 1 is subtracted from the Hallway's occupancy.
The sensors are ideally embedded in the wall above the doorframe. The fresnel lenses provide a large detection area, and the wall limits the detection area to one side of the doorway. The sensors can also be placed on the ceiling with a deflector in the middle to keep their detection areas separate [15]. The sensors must be placed out of range of any sources of variable heat output, such as a heating vent, since they detect changes in heat. Sources of constant heat, such as windows and light fixtures, are not a problem as long as their heat output is lower than that of a human. |
PIR sensors very rarely have false signals, but one persistent issue overcoming signal blindness during the masking time. Adding a second set of sensors in transition regions, such as hallways, provides an added measure of accuracy. Hallways work well for secondary sensors because they are used to move between areas with few instances individual's rapidly doubling back on their paths.
PREDICTIVE MODELING
Another option to increase accuracy is incorporating predictive modeling. Predictive modeling uses data patterns to determine the probability of each possible path for a person to take and their target destination zone once they are detected in an initial zone. Predictive modeling increases accuracy but also has the consequence of addition cost and design difficulties. The added challenge of designing predictive models means that high accuracy PIR occupancy detection systems still need intensive research before they are a viable technology in the future. Predictive models act as a corrective factor for the programming portion of the system by decreasing the frequency of rapid adjustments due to events such as a person entering a zone and then exiting before they release enough heat to change the air temperature [18]. Predictive modeling is currently designed individually for each building's specific environment. Further research aims to at least partially standardize the design process. Standardized predictive models offer the possibility of easily finding a multiplier. Once accuracy rates are known, a multiplier can be determined to correct for incorrect counting. The multiplier is applied to the occupancy value as a corrective factor to increase accuracy.
One test analyzed the worth of predictive modeling by comparing a program that utilized only real-time data from PIR sensors with a program that utilized both PIR sensors and predictive modeling. The results showed an 16% increase in efficiency, from 70 to 89% [19]. Another study's predictive modeling achieved accuracy with less than 1% error when compared with ground truth data [15] [footnote 2]. This study created a system for an office with 3 employees always within the building. Higher occupancy causes increased error because the likelihood of missed signal during masking times increases. Until masking times decrease, PIR sensor systems are best suited for buildings where 10 or fewer frequently cross each border. One possibility to get around this issue is integrating occupancy detection into only a portion of the buildings HVAC programming. A third study designed a predictive model, named SCOPES, that had high accuracy when compared to ground-truth data, as shown in Figure 7. The high accuracy data was then fed to a simulation that showed the potential for an 8.8% energy savings for typical office buildings. Buildings that had underutilized zones with temperature regulation showed increased savings of 20% [3]. Collecting ground-truth data, as done in this study, is a large time commitment and is difficult to coordinate but is vital for determining accuracy. Some studies use simulation software to generate preliminary data and never move beyond this stage due to a lack of funding.
One test analyzed the worth of predictive modeling by comparing a program that utilized only real-time data from PIR sensors with a program that utilized both PIR sensors and predictive modeling. The results showed an 16% increase in efficiency, from 70 to 89% [19]. Another study's predictive modeling achieved accuracy with less than 1% error when compared with ground truth data [15] [footnote 2]. This study created a system for an office with 3 employees always within the building. Higher occupancy causes increased error because the likelihood of missed signal during masking times increases. Until masking times decrease, PIR sensor systems are best suited for buildings where 10 or fewer frequently cross each border. One possibility to get around this issue is integrating occupancy detection into only a portion of the buildings HVAC programming. A third study designed a predictive model, named SCOPES, that had high accuracy when compared to ground-truth data, as shown in Figure 7. The high accuracy data was then fed to a simulation that showed the potential for an 8.8% energy savings for typical office buildings. Buildings that had underutilized zones with temperature regulation showed increased savings of 20% [3]. Collecting ground-truth data, as done in this study, is a large time commitment and is difficult to coordinate but is vital for determining accuracy. Some studies use simulation software to generate preliminary data and never move beyond this stage due to a lack of funding.
FUTURE RESEARCH GOALS
The majority of research done on developing PIR sensor high occupancy detection was completed by collecting real-time data which was used as input for simulations. The trials tested technology and various sensor arrays without testing real-life smart thermostat programming or HVAC response behavior. Much progress needs to be made in a variety of areas before these systems can be made available on the market. PIR sensor technology is still advancing, especially regarding masking time. In order to be practical, predictive modeling programs must have either wider applicability to varying environments or a streamlined design process.
Once accurate data collection and interpretation are developed, the process of their integration into smart thermostat programming is simple. Occupancy detection allows deeper setback temperatures because occupancy is a predictor or air temperature's rate of change. PIR sensors are very accurate when determining binary occupancy without knowing the exact number of people present, so the system is nearly guaranteed to have a rough estimate of the occupancy. Air temperature will still be the dominant data point for determining HVAC output if the temperature falls outside a specified range based on a zone's maximum occupancy. Air temperature also changes from factors not always correlated to occupancy, such as light fixtures and windows. Higher accuracy leads to greater efficiency, but 100% accuracy is not required to create a system that results in energy savings. The interactive code shown in Figure 8 is a simplified model of smart thermostat programming. It takes the input of an occupancy value and outputs the additional power consumption needed to offset the body heat impact. Although the average consumer will never see code like this, it shows how occupancy detection can be integrated into existing systems. A similar version of this code is also displayed on the Calculator Page.
Once accurate data collection and interpretation are developed, the process of their integration into smart thermostat programming is simple. Occupancy detection allows deeper setback temperatures because occupancy is a predictor or air temperature's rate of change. PIR sensors are very accurate when determining binary occupancy without knowing the exact number of people present, so the system is nearly guaranteed to have a rough estimate of the occupancy. Air temperature will still be the dominant data point for determining HVAC output if the temperature falls outside a specified range based on a zone's maximum occupancy. Air temperature also changes from factors not always correlated to occupancy, such as light fixtures and windows. Higher accuracy leads to greater efficiency, but 100% accuracy is not required to create a system that results in energy savings. The interactive code shown in Figure 8 is a simplified model of smart thermostat programming. It takes the input of an occupancy value and outputs the additional power consumption needed to offset the body heat impact. Although the average consumer will never see code like this, it shows how occupancy detection can be integrated into existing systems. A similar version of this code is also displayed on the Calculator Page.
ADDITIONAL OPTIONS
PIR sensors occupancy detection can be supplemented by other forms of occupancy detection. One viable option is opportunistic factors. Follow the link below to learn more.
|
Footnotes:
[1] Masking Time is the time period during which the sensor rests before it is able to detect consecutive objects.
[2] Ground Truth Data is occupancy data that is collected in person for the purpose of determining accuracy.
[1] Masking Time is the time period during which the sensor rests before it is able to detect consecutive objects.
[2] Ground Truth Data is occupancy data that is collected in person for the purpose of determining accuracy.
Sources:
[16] Tachwali, Y., Refai, H., and Fagan, J. E., "Minimizing HVAC energy consumption using a wireless sensor network," presented at the 33rd Annual Conference of the IEEE Industrial Electronics, Piscataway, NJ, (2007).
[17] Warren, B. F., and Harper, N. C., "Demand controlled ventilation by room CO2 concentration: a comparison of simulated energy savings in an auditorium space," Energy and Buildings, 17, pp. 87-96, (1991).
[18] Li, N., Calis, G., and Becerik-Gerber, B., "Measuring and monitoring occupancy with an RFID based system for demand-driven HVAC operations," Automation in Construction, 24, pp. 89-99, (2012).
[19] Meyn, S., et al, "A sensor-utility-network method for estimation of occupancy in buildings," presented at the 48th IEEE Conference on Decision and Control held jointly with 2009 28th Chinese Control Conference, Shanghai, China, (2009)
Sources for the code:
[1] "U-Values for common materials," [Online], 2016 December 3rd, Combustion Research Corperation, from: http://www.combustionresearch.com/U-Values_for_common_materials.html
[2] U.S. Department of Energy, "Selecting Windows for Energy Efficiency," Whats New in Building Energy Efficiency, https://windows.lbl.gov/pub/selectingwindows/window.pdf, DOE, Pittsburg, PA (1997)
[16] Tachwali, Y., Refai, H., and Fagan, J. E., "Minimizing HVAC energy consumption using a wireless sensor network," presented at the 33rd Annual Conference of the IEEE Industrial Electronics, Piscataway, NJ, (2007).
[17] Warren, B. F., and Harper, N. C., "Demand controlled ventilation by room CO2 concentration: a comparison of simulated energy savings in an auditorium space," Energy and Buildings, 17, pp. 87-96, (1991).
[18] Li, N., Calis, G., and Becerik-Gerber, B., "Measuring and monitoring occupancy with an RFID based system for demand-driven HVAC operations," Automation in Construction, 24, pp. 89-99, (2012).
[19] Meyn, S., et al, "A sensor-utility-network method for estimation of occupancy in buildings," presented at the 48th IEEE Conference on Decision and Control held jointly with 2009 28th Chinese Control Conference, Shanghai, China, (2009)
Sources for the code:
[1] "U-Values for common materials," [Online], 2016 December 3rd, Combustion Research Corperation, from: http://www.combustionresearch.com/U-Values_for_common_materials.html
[2] U.S. Department of Energy, "Selecting Windows for Energy Efficiency," Whats New in Building Energy Efficiency, https://windows.lbl.gov/pub/selectingwindows/window.pdf, DOE, Pittsburg, PA (1997)