Share
Blog
Cite
Print
Email
BookmarkScienceDaily (Sep. 15, 1999) — WEST LAFAYETTE, Ind. -- Brownouts like those plaguing cities this summer may soon be relegated to the history books if engineers meeting today at Purdue University have their way.
See also:
A national engineering consortium is developing a "self-healing" computerized system designed to prevent power failures by anticipating the quickly changing demands of industrial, commercial and residential electric customers. The system would accurately predict power needs for the coming day and then automatically meet those demands by better managing the distribution of electricity and supplementing the grid with power from small, neighborhood generators, which would kick on when needed.
Such an automated system would enable a sophisticated control of small sections within a service area, which is a sorely needed innovation because changes in electrical demand can vary drastically from one part of town to another.
"We want to endow the grid with certain self-regulation capabilities," says Lefteri Tsoukalas, an associate professor in the Purdue School of Nuclear Engineering.
Recent findings show that the method would be effective in better controlling the flow of electricity during times of peak demand, such as this summer's heat wave. The findings are based partially on the analyses of energy-consumption profiles of customers in the Chicago-area city of DeKalb, Ill., which was chosen for the research because it contains a wide range of users, from a university to industrial and residential neighborhoods. Engineers will discuss the work during a meeting at Purdue on Monday (9/13).
The system will be called TELOS, for Transmission Entities with Learning-Capabilities and On-Line Self-Healing. "Telos" is a Greek word meaning purpose, or the where and why of things, says Tsoukalas, who specializes in "neurofuzzy systems," computer software designed to think more like people by learning from experience and using the "fuzzy logic" of human reasoning.
Because electricity cannot be stored in large amounts, it is extremely difficult to maintain a smooth flow throughout a power grid that is made up of diverse types of customers, some of whom vary their consumption considerably from week to week, day to day or even hour to hour. The supply must be balanced constantly to meet the changing demands, which can fluctuate much differently in various sections of the service area, says Tsoukalas, who heads up a team of Purdue engineers and graduate students participating in the Consortium for the Intelligent Management of the Electric Power Grid. The consortium's other members are Commonwealth Edison Co., the University of Tennessee, the Tennessee Valley Authority and the Electric Power Research Institute, an organization of electric utilities (http://www.epri.com/).
Fuzzy logic systems work by evaluating the overall accuracy, instead of the fine precision, of a solution to a problem. The human brain uses the same sort of approach to make effective decisions.
"We adjust the thermostat in the room not by calculating precisely what the perfect temperature would be, but basically by deciding that it's comfortable or uncomfortable," Tsoukalas says. Another example of fuzzy logic can be found in language. Although human vision can literally distinguish more than a million shades of colors, languages do not contain a million words for colors.
"We would be overwhelmed if we didn't have the ability to summarize," he says.
In a similar way, mathematical models can be used to predict future changes in electricity demands by evaluating the present usage in the context of environmental factors and historical patterns. Such a system might reason along the following lines: "The weather is getting warmer, it's the middle of the summer and the humidity is increasing. Therefore, when historical consumption patterns are considered, we should expect the demand to rise by so much in the next hour."
By automatically adjusting to new conditions, the system is said to be "self-healing," resulting in fewer complications and less frequent power failures.
Utility company workers currently perform the prediction role. They begin the day by trying to anticipate what the power demand will be over the next several hours in the entire grid. However, different parts of the service area sometimes have their own, distinctive microclimates that affect electricity use. Therefore, engineers are aiming to break the service area into "local area grids" that can be managed autonomously by computers.
The consortium will meet from 9:30 a.m. to 4 p.m. Monday (9/13) in Stewart Center at Purdue to discuss recent findings and the direction of future research. TELOS is scheduled to be ready for demonstration by the end of 2001. It will be operated on a trial basis for several years in the Commonwealth Edison and Tennessee Valley Authority service areas. If it works, it will be used on a national level.
