Spring 2004 – vol. 22, no. 1 logos home page Contact the logos editor Argonne home page Feature articles Superconductivity team wins top research prizes Creating stability in a world of unstable electricity distribution Argonne tests, creates fuel cells to power the future Fueling the hydrogen future with Argonne's ceramic membrane Nuclear plants may be clean hydrogen source Argonne Update Stars surrender secrets to supercomputers Electricity controls nanocrystal architecture

Creating stability in a world of unstable electricity distribution

by Evelyn Brown

2003 BLACKOUT – Top: A satellite image of the northeastern United States taken at 9 p.m. EDT August 13, 2003 – the evening befoe the blackout. Compare it to the lower satellite image, taken about the same time during the blackout. Images courtesy of the National Oceanic and Atmospheric Administration and the Defense Meteorological Satellite Program.

Extreme heat. Acts of nature. Human error. Deregulation. Equipment failure.

Several factors combined during the afternoon of August 14, 2003, to create a blackout that left 50 million people around the Great Lakes without power and cost the nation's economy an estimated $1 billion. This was only the latest in a string of electrical outages. At the other end of the country, bottlenecks in California's transmission grid caused notorious and costly outages throughout summer 2001.

The nation's electric transmission and distribution system is extremely complex. "There are many variables in this interconnected system," said Argonne engineer Yung Liu. "Its dynamic behavior is almost impossible to predict."

But researchers are challenged by the seemingly impossible. "There is evidence that small disturbances occurred several hours before the Aug. 14 blackout," Liu said. "With such warnings and a better understanding of the system's dynamics, we might be able to control events in the future."

Engineers in Argonne's Energy Technology Division are working to make the nation's electricity supply more reliable and flexible through the use of software and hardware innovations. Their three-pronged approach includes:

• training a computer to control a local-area power grid,
• developing techniques to monitor and control voltage instability, and
• patenting a hardier current-controlling device to handle power surges.

The U.S. power grid resembles the Interstate Highway System criss-crossing the country, and like the interstate system, the power grid has experienced some notorious traffic jams. Based on 1950s technology, the grid is composed of three main geographic sections – one in Texas and two others split roughly along the Continental Divide.

The original power grid

The grid was originally laid out to enable a utility to deliver power to local residential, commercial and industrial costumers. Deregulation has encouraged utilities and other generators to produce more electricity and to transfer it outside their local service area to meet electricity demand elsewhere. Researchers at the Electric Power Research Institute (EPRI) suggest that deregulation is stressing the national transmission system by causing it to operate beyond its design. This possibility is described in the foreword to the Nov. 20, 2003 white paper: "Factors Related to the Sources of Outages on August 14, 2003."

Growing power demand

POWER GRID – The electric power grid crosses the entire United States.

EPRI analysts also assert that power demand is outgrowing supply. Between 1988 and 1998, demand grew 30 percent, but according to the white paper, only enough new transmission capacity was added to handle about half that amount. A recent North American Electric Reliability Council assessment predicted that electricity demand would grow 20 percent from 2002-2011, but only 3.5 percent new transmission capacity is planned. This imbalance will further stress an already overtaxed electricity supply.

Before the electric power grid can be controlled effectively, researchers first need to understand the system. Argonne engineers decided to start at home with Argonne's own local-area grid. "It's one piece of the system," Liu said. "Studying Argonne's own local-area grid and using it to evaluate methods and technologies is both practical and representative of the larger system. Plus, the data are easily accessible.

"Argonne's local-area grid is ideal for testing software," Liu explained. "We are one of ComEd's 40 largest customers." More than 4,000 people work at Argonne's one-square-mile Illinois site, and in addition to maintaining office buildings and laboratories, Argonne operates three energy-intensive, national user facilities that operate around the clock.

TELOS tells a lot

Argonne's researchers have been training and testing an intelligent software program that is designed to monitor and control the laboratory's local electrical grid. Similar software may one day be able to detect and correct problems much faster than a human, and on a much larger scale than that of the laboratory system.

To accomplish this, Argonne engineers are using software created by the laboratory's research partners at Purdue University. The software is called Transmission Entities with Learning-Capabilities and On-Line Self-Healing (TELOS). Telos is also the Greek word for "purpose."

One of the most unique features of this software is that it was designed to have learning capability. Purdue originally developed the software to be "trainable" using relatively advanced programming techniques called "neural network" and "fuzzy logic." Compared to the "Boolean logic" that governs traditional software, these learning methods more closely resemble the way a human brain works. Neural network software is trained with information and rules about data relationships, much like humans are taught, and eventually the software becomes capable of making decisions.

The TELOS code is written in the Java language and is modular, so that key components can be modified or replaced without affecting other parts of the software. It has been "taught" to understand the sequence in which various electric loads can be "shed" and the best methods to bring new power generating units into service.

TELOS was adapted and set up to accommodate the Argonne local-area grid, which comprises two 138kV lines, eight transformers and a substation – enough equipment and capacity to power a small city. Argonne's researchers trained TELOS with a year's worth of data, including the laboratory's hourly electrical consumption, temperature and humidity information.

Using this historical data, the software "learned" to correlate the historical patterns of electricity consumption with temperature and humidity and, on that basis, to make intelligent forecasts of future energy loads. TELOS was then run for varying time periods – ranging from a few hours to slightly over three months – of Argonne electricity use and weather data to test how accurately it could predict the laboratory's power demand 30 minutes into the future. It proved to be accurate within 3 percent.

"We were surprised," Liu said. "It was encouraging that TELOS was so accurate with no fine-tuning."

The next step is to test TELOS in real time online when Argonne's supervisory power control and data acquisition system is installed to manage the laboratory's power systems. Researchers also want to test TELOS on larger parts of the grid and to study the dynamic behavior of local-area grid interactions.

The nation's electricity grid is so complex that it makes sense to have intelligent computers operating it to make the split-second decisions needed, Liu explained. "We would like to have a self-healing grid one day."