by Gordy Slack
If building managers could count the occupants of the buildings they run in real time, and if they could track where those occupants go, it would allow for new degrees of efficiency in air conditioning buildings. But if managers could also predict where those occupants would be an hour in advance, it would turn building management on its head, save billions of dollars in energy costs and drive down greenhouse gas emissions.
Professor Alberto Cerpa and his colleagues at UC Merced say they have taken a big step in that direction with a simple-to-deploy “optical turnstile” system that tracks how many people are in each of a building’s rooms as well as the application of mathematical algorithms that accurately forecast how many people will be using those same rooms hours in advance.
“Right now the people who manage the HVAC systems are pretty much blind to how people actually use their buildings,” says Cerpa, assistant professor of Electrical Engineering and Computer Science at UC Merced. “And since they cannot track occupancy, they condition the air assuming the maximum number of people is using all the rooms all day long. But many of those conditioned rooms are too often empty or are used by too few people.”
The system to track a building’s dwellers, the optical turnstile, employs a network of tiny, inexpensive, wireless optical sensors known as Smart Cameras Object Position Estimation System (SCOPES), which is strategically placed on ceilings throughout a building’s hallways. The micro-cameras count how many people are in each room and track where they are going. Beyond that, the system employs machine-learning techniques to predict, based on continually updated data analysis, where people are likely to go in the building hours into the future.
Cerpa and his colleagues used UC Merced’s Science and Engineering One (SE1) building as a testbed for their prototype optical turnstile system. Cameras have been attached to the ceilings in hallways of one entire wing of the building and are monitoring and analyzing the comings and goings on two floors and in 60 different rooms. The SCOPES system allows building managers to tailor CO2 ventilation to the number of inhabitants, says engineering graduate student Varick Erickson.
While some buildings already use Passive Infrared (PIR) sensors to detect occupancy, these only give a binary reading: either the room is occupied or it is not. The Merced system, on the other hand, allows building managers to tailor CO2 ventilation to the precise number of people in a room, says Erickson.
In the current study, the amount of energy used to condition the air for SE1 has been reduced by 40 percent.
“The importance of ventilation is something that people working on our area just have not been able to see,” says Erickson. “But it is key to improving energy use.”
If one person is sitting alone in an auditorium, the ventilation requirements are low even if the entire room must be cooled. But if hundreds of people are gathered there, the ventilation must kick into overdrive to ensure sufficient oxygen for everyone. And the more ventilation there is, the greater the volume of air that needs to be heated or cooled.
Currently, there is “not a lot of science behind the temperature and ventilation schedules buildings keep,” says Erickson. Most building managers just turn up the temperature and ventilation in the morning and turn them down again at night. In a building like SE1, though, where grad students sometimes work late into the night, and where conference rooms and auditoriums may often be empty for long periods during the day, that regimen is very inefficient. Not only will big, empty spaces be using lots of wasted energy during the day, but graduate students working late into the night will work in rooms that may be uncomfortably cold or hot.
Cerpa’s optical turnstile system has three main parts. First, the network of matchbook-sized cameras is able to recognize something moving through their field as a generic person and to determine what direction he or she is moving. The optical turnstiles don’t identify people, they only count them.
The cameras are restricted to the building’s hallways to skirt privacy issues, says Erickson. They cover every exit and entry point to all other rooms—except restrooms—so the system always knows how many people are inside each room without having to “look” inside the rooms themselves.
Because it takes time to condition a room to a comfortable temperature, starting the heating or cooling before users arrive is important. So part two of the system is software that predicts, based on data from the cameras, where people are likely to go at any given time.
“Every ten minutes we look at occupancy and, based on the estimated pattern distribution, project an hour into the future,” says Erickson. The system, employing a blended Markov chain, continues checking its predictions against reality and improving them. Over time, it comes to know a building, and to predict the behavior of the people in it, very well.
Finally, the third part of the system feeds that prediction into a control algorithm module that interfaces with the building management system and sets the optimal temperature and ventilation for all of the rooms.
The optical turnstile system will be easy to deploy in older buildings, Cerpa says. The micro-cameras are battery-operated, attached the ceilings with Velcro, wirelessly networked, self-computing, and are connected wirelessly to the back-end server. The only requirement is a building management system that allows for control of temperature and ventilation in a building’s different zones. The granularity of a building’s HVAC management system will determine the degree of control the system delivers. More independently operated air conditioning vents allow for more precise control.
Cerpa plans to commercialize the optical turnstile system in the next couple of years. “Even if you consider the material and operational costs (including installation, equipment, batteries, etc.) a building manager can amortize the whole thing on the savings of three weeks to a couple of months of operations depending on the building type and usage patterns,” says Cerpa.
“Cooling, heating, and ventilation in commercial and residential buildings use about 20 percent of U.S. energy,” says Cerpa. “If smarter buildings can reduce that by 40 to 50 percent…while making those buildings more responsive and more comfortable, the savings would be enormous!”