Copyright ANG Newspapers Aug 25, 2003
In less than 12 seconds, last week's electrical hiccup in Ohio escaped grid defenses and rippled across eight states. After that, the rapid-fire shutdown of power lines and plants was largely automatic, the darkening of tens of millions of homes inevitable. Investigators for the U.S. Department of Energy say it will take weeks to pinpoint a cause. But it already was clear that failed alarms and poor system awareness are likely to play as much a role as failures of humans and hardware. "Cascading" outages are complex, layered events. But they have warning signs for minutes, even days beforehand. Last week, several red flags went up more than an hour before the Midwest and Northeast went down. Soon, however, operators of California's electrical grid could be the first in the nation to use computer simulations to pre-dict, with 15 to 20 minutes' notice, some of the least understood grid- crashing problems. They're known as voltage collapses, and they can begin and end in little more than a blink of an eye. Such a collapse took down power for 7.5 million Westerners in 1996, and some experts already are fingering one as the culprit in last week's Northeast outage. Spotting those signals is no simple matter. In California alone, a $10 million computer sweeps in data from 200,000 sensors but only every four seconds, an eternity when disturbances travel at near light speed. But if successful, the simulations and other software for spotting problems on neighboring systems could serve as the foundation for a highly automated power grid, correcting itself with little human intervention. That's both alluring to energy scientists and yet more challenging as the grid becomes more complex. "It is moving in the direction of harder for the individual to get one's hands around and easier for the machine," said Kevin L. Stamber, a power systems analyst at Sandia National Laboratories. Turning more grid controls over to computers "is the dream of just about every systems operator." Lawrence Berkeley lab energy scientist Joseph Eto said the latest grid software lets operators "know where we are with respect to what they call the edge." The next generation, due in as little as five years, will add controls. "The goal is to enable the type of system that could detect and respond to problems very quickly, faster than humans can respond," said Eto, program manager for the Energy Department's Consortium for Electric Reliability Technology Solutions. Energy Department investigators warn that pinning last week's blackout on a voltage collapse is premature. They still are stitching together a timeline for the blackout, drawing on hundreds of thousands of readings on multiple electrical systems. The analysis is likely to take weeks. Trying to perceive such events beforehand is "mindbogglingly deep," said Ian Hobson, an engineering professor at the University of Wisconsin. "It's much worse than looking 10 deep in a chess game." Despite the lack of a diagnosis, federal lawmakers and power experts already are tallying their prescriptions: New market rules, more neighborhood-size power plants, faster electronic switches to replace mechanical ones and 30,000 miles of new high-voltage lines, perhaps triggering the first large-scale use of superconductors that carry 25 times the energy of ordinary wires. Energy Secretary Spencer Abraham pegs the cost at $50 billion. The Energy Power Research Institute, a Palo Alto-based nonprofit industry think tank, has ideas for a "smart grid" that will cost $100 billion and respond more automatically to changes in power flows. Building a more modern, resilient grid would start with a decade- long effort and continue indefinitely. Abraham says U.S. consumers will foot the bill. But as early as this fall, years before the lines go up, California will pioneer a faster, cheaper way to run the grid reliably, with software that significantly boosts the awareness of system operators. Today, controllers at the California Independent System Operator in Folsom can glance up at four, giant rear-projection screens and get a sense of power flows on their system, serving 80 percent of the state, and on the larger grid for the entire West. "When you walk on the floor, you get an instant feel for the status of the system," said David Hawkins, the ISO's manager of special projects engineering. But for the last 18 months, engineers and students have put thousands of hours into a pair of back-to-back computer simulations that begin online testing in September. The first, known as a state estimator model, creates a color snapshot of voltages grid-wide every five minutes. It corrects bad data and makes estimates in places where data is missing. The second, produced by Pasadena-based Electric Power Group for Eto's Energy Department consortium, draws on the voltage estimates and runs worst-case scenarios to predict trouble spots. Alarms and color 3-D maps show when and where. "It will look at, 'Gee, what happens if another transmission line or generator trips offline? Will you have a black hole open up in your system?' Then you'll have plenty of time to take action," Hawkins said. "We think we'll get 15 minutes' warning," he said. "In that time, you can usually run some other generation out or you can change the loads on lines." As a last resort, system operations can shut off customers to reduce demand on the system. The two pieces of software were built to predict the kind of voltage collapse that de-powered much of the West Coast on Aug. 10, 1996. In a series of wild system fluctuations, two major transmission lines shut themselves off and shunted their load onto a third line. As system managers struggled to figure out what was happening, the extra load heated the third line and caused it to sag down into a tree where it shorted out. That triggered a rapid-fire, cascading outage. In seconds, customers lost power from Baja to British Columbia. Hawkins expects the simulations will be ready for California's winter demand peak, running roughly from Thanksgiving to New Year's Day. Experts say more transmission lines are inevitable. Software may forecast a problem, but for system operators to avoid it, they must have excess line capacity to carry the load. Berkeley's Eto said new grid-managing software will have to be paired to new transmission lines, more generation close to consumers and energy conservation. Turning to automation also poses its own risks, notes University of Wisconsin's Dobson. "We need to be aware of system complexity," he said. "You may introduce many unusual failure modes just by sheer complexity, and I'm not sure how we make those systems reliable." Contact Ian Hoffman at ihoffman@angnewspapers.com . |