The Examined Life of Room 464

by Gordy Slack

Room 464, on the fourth floor of Sutardja Dai Hall, looks like a typical office. Humming away are desktop computers, a small fan, a dorm-room-style fridge, a laser printer, and uninterruptible backup power supply units. Illuminating the room are twelve florescent lighting fixtures built into the ceiling and a standing lamp equipped with a 60-watt incandescent bulb.

QR tags go on all of the electrical devices and outlets.

Look closer though, and you will see how it differs from an ordinary office. For one thing, each appliance is tagged with a quick response (QR) code, (a matrix barcode that is readable by cell phone applications) as is each electrical outlet. Between each outlet and its appliance is a communicating plug-level power meter (CPLPM), an instrument that can measure, and control, the electricity passing into each appliance. One of the computer monitors shows an image of Room 464 and a graphic interface showing how much electricity each appliance, and therefore the entire room, is consuming.

While most offices live an unexamined life, this one seeks the workplace equivalent of self-awareness: the researchers who occupy it are trying to teach Room 464 when it is best to do different electrical tasks and, most importantly, how to reduce its energy consumption by 30 percent at the drop of a hat…without losing its cool. 

At the end of last summer, CITRIS researchers and Professors David Culler, David Auslander, and Paul Wright won a $2M grant from the US Department of Energy to create a system that would allow Sutardja Dai Hall to respond to a high load on the California power grid by automatically instituting a temporary 30 percent reduction of electricity use. Such a capability, known as Demand Response (DR), if instituted widely in commercial buildings around the state, could save California utilities, taxpayers, and power customers $2.18 billion dollars over the next 20 years.

On a few hot days each summer the demand on the California grid can increase by 50 percent mainly due to increases in the use of air conditioning.  To meet this demand, state utilities fire up older, dirtier, and more expensive fossil-fuel power plants. And on extreme occasions, as in 2001, even those extra plants could not meet the demand resulting in the now-notorious Californian rolling blackouts.

Preserving California’s rarely used “peaker” plants for only a few hundred hours of service is extravagantly expensive. A better strategy would be to reduce the demand side of the equation for just a few hours on those very hot days, says Jay Taneja, a computer science graduate student.

A graphic interface shows how much electricity each appliance in Room 464 is consuming.

The DR approach—using communications and IT—is to signal devices and consumers that the grid is stressed and have them reduce their electric load. The project unfolding in Room 464, Distributed Intelligent Automated Demand Response (DIADR), does not depend on consumers manually responding to a plea from desperate utilities, however. Rather, this team is creating a “smart” and automated system that can figure out how to conserve the most energy for the grid while causing the least disruption for energy consumers, says Jason Trager, one of Professor Paul Wright’s graduate students. It will accomplish this by creating a network of appliances and building systems programmed to make sophisticated real-time decisions about what can be turned off—or turned down—when.

The HVAC is simple enough to turn down, says Domenico Caramagno, facilities manager at CITRIS. But the distributed load is trickier. For example, the system will have to recognize that the room’s uninterrupted power supplies can run computers for six to eight hours each,  says Trager, so power to computers plugged into them can be shut down for that long with no  adverse effect. The system also knows that the refrigerator can be turned off, as long as it is turned back on for five minutes every 15 minutes or so,  again with no adverse effect. The system knows that while the overhead ambient lighting is strong, workers at their desks can get by happily with desk lamps for a while.

The software needed to track and facilitate these negotiations is being developed by Professors Culler and Auslander.  “Gateway” software will work on a very local room-by-room or floor-by-floor level and a broader building-wide level. Eventually, other, more system-wide software will also track and coordinate the network of participating buildings on, say, the entire UCB campus.

Trager and his team of undergraduates have tagged every computer, every lamp, and every fan that they could find: 730 QR tags in all. All of these go into StreamFS (, which is like a file-system for building sensors and resources and also rates each appliance and describes its basic constraints: How much electricity does it use? How essential is it? How long can it be off without causing problems? Does it have a backup power supply? And so on. Each tag will connect the project’s database and management software via StreamFS to the appliance so the computer’s that manage the system can both read from and influence the appliances.

The immediate aim of the DOE grant is to allow a quick and automated reduction in use by 30 percent to avoid blackouts. But the ramifications go far beyond that, says Trager. Once the building and its occupants “know” what it is capable of, it should boost everyday efficiency as well. 

The same network of sensors and management software that enables a 30 percent reduction in a crisis will also make possible constant and more subtle adjustments. When the grid has energy to spare, recharge the back-up power supplies and turn off AC controls. When the grid is stretched, it can turn down the AC and dim the lights a little, turn off unused computers and fans, and stop charging batteries unless necessary.

“Our ultimate vision is that the building is continuously breathing electricity in and out, responding to whatever the grid presents,” says Trager. “You will always have the capacity to move around your load, depending on the slack in the system. Currently, load follows supply. In the future, supply will follow load.”

This will be especially important as California shifts to cleaner but less controllable energy sources. You can always fire up the extra coal or natural gas plant, but you can’t turn on the sun or wind. Today, highly variable but renewable energy sources generate about 18 percent of California’s electricity. By 2020 that number is expected to be about 33 percent.

Creating a communicative, nimble and controllable Room 464 is straightforward enough, but multiply that job by the 123 rooms in the building, and then the hundred or so buildings on the UC campus, and the complexity skyrockets. But so do the benefits, to both the environment and the economy. 

The automated adjustments to grid demand will certainly increase efficiency. But so could human behavior, if, that is, the building’s occupants have information to work with.

Toward that end, Trager is developing a smaller-scale software interface that will let occupants of Room 464, and eventually the rest of SDH, know just how much electricity they are using. “If people get feedback, they change. They will use less,” he says.