Lab Makes Gains on Web-Based
Demand Response System
by David Jones

Researchers at Lawrence Berkeley National Laboratory said last week they have wrapped up a successful trial run of an Internet-based system designed to avoid blackouts and help large electricity users cut their bills -- without utility operators or building managers having to lift a finger.

The test was designed to "explore how a utility, an ISO [independent system operator] or the grid might communicate in the future" with customers to manage electricity use in large buildings whenever blackouts or excessive electricity demand threaten the grid, or when high prices might encourage large power users to reduce their energy use, Berkeley lab scientist and project principal investigator Mary Ann Piette said.

The experiment involved five large buildings -- all of which use different computer programs to handle their energy operations -- that were tied into the Berkeley lab's system for triggering load shedding, also known as load control and demand response, prices or requests to curtail use in response to power emergencies. "The basic idea was to see whether we could fully automate load shedding," Piette said in an interview Tuesday.

Working with the company Infotility, the lab used a two-way communications language known as XML to send fictitious price signals through the Web to "gateway boxes" attached to the different energy-management systems in the buildings -- an Albertsons grocery store, a Bank of America office building, a Roche biotechnology facility in Palo Alto, Calif., the University of California at Santa Barbara library and the Ronald Dellums Federal Building in Oakland, Calif.

After some initial glitches, researchers succeeded in shedding load at all five buildings in response to price increases. "We've demonstrated that many different types of systems can listen to a common XML signal and initiate coordinated load control using the Internet," said Piette, who works in the Berkeley lab's Environmental Energy Technologies Division.

The California Energy Commission funded the tests, which were conducted Nov. 12 and 19. The five buildings previously had received CEC funding during California's energy crisis in 2000-01 to upgrade their automated energy controls.

Other systems have tested demand response systems, Piette noted. New York's transmission grid operator, for instance, conducted an experiment in 2002 that sent signals to electricity users encouraging them to curtail use in response to price increases. But a recent Energy Department evaluation of the program found that few customers took advantage of automated systems to shed load.

The XML project was the first to test a technology providing fully automated demand response in large buildings through two-way communications, Piette said. XML sent consistent, fictitious price signals of $.10 per hour to the five buildings for most of the two-week test period. It also checked every five minutes to make sure the signals were being received.

When XML sent a mock price spike to the buildings on Nov. 12, some bugs appeared. At two sites gateway boxes went offline and had to be reset, and three of the five buildings failed to shed load in response to the price increase. But a second trial run a week later went off smoothly.

"All the sites were successful the second time. All shed their load without anyone touching anything" after the price was boosted to $.30 per hour, and they further reduced demand when the price hit $.75 per hour, Piette said. The XML system allowed LBNL researchers to follow the load shedding as it occurred. Preliminary figures from the test showed that, altogether, the buildings dropped about 10% of their power loads.

Also significant was the experiment's success in sending load-shedding signals to a variety of energy-management systems, Piette said. Building energy management typically is handled by systems manufactured by industry giants like Honeywell, Johnson Controls and Trane, she noted, and the test showed that demand-control technology is compatible with existing machinery.

The test illustrated the "potential for flattening overall system load," she said, which could prove especially useful on hot days when buildings with large air-conditioning systems tend to tax the grid and prices soar. By moderating demand during these "peak" periods, "annual costs could go down on everybody's bill," Piette said.

In addition, automated demand response avoids the need for companies and agencies to train workers and send them out to shut off power manually during an energy crisis. "If a power plant or transmission line goes down, [demand response] signals can quickly reduce power demand as well as the likelihood of a full-scale outage," according to a Berkeley lab statement. "Some experts on the electrical grid believe that a system with automatic demand response could have avoided the blackout in the eastern U.S. and Canada on August 14, 2003."

There are potential risks. An automated demand-response system without a human fail-safe device might curtail essential services at exactly the wrong time, though the Berkeley lab technology gives operators the chance to opt out.

Berkeley lab scientists are now sifting through the data generated from the November test. The next step will be to recruit new participants to conduct the experiment on a larger scale with more buildings, preferably during warm weather this summer, Piette said.

Back to "Today at Berkeley Lab"