01 June 2004
Power quality tools shed light on 2003 blackout.
By David Pereles
Few facilities in the U.S. have escaped a power interruption during the course of operation. Construction crews sever underground power feeds, vehicles strike power poles, and storms bring down distribution networks. Despite the fact that minor power outages occur frequently, the blackout of 2003 was a sudden wake-up call.
Shortly after 4 p.m. Eastern Daylight Time on Thursday, 14 August 2003, the lights went out over much of the northeastern U.S. and southeast Canada, including major cities such as New York, Detroit, and Cleveland in the U.S. and Toronto and Ottawa in Canada. More than 85% of New York state lost power, as well as parts of New Jersey, Vermont, Connecticut, Ohio, and Michigan. The blackout ultimately affected an estimated 50 million people.
Although there has been substantial commentary about the event, few have written about the symptoms individual facilities experienced. Even facilities in unaffected areas experienced measurable changes in their voltage supply. On that day, engineers throughout the eastern U.S. were using instruments to perform power studies and obtaining documentation of the event as experienced at various locations.
The thought is the outage started in transmission lines and generating plants east of Cleveland, Ohio, and around the western banks of Lake Erie. While the overall outage lasted twenty-four hours, there were areas where the outages were more random. A facility in northern Ohio recorded power down for eleven and one-half hours. At monitoring points south of the New York and New Jersey transmission interconnections, the power was out for ten minutes during the initial event. Two more outages occurred within two hours. As far south as Georgia and as far west as Arkansas, the power recorders picked up frequency jitter and slight voltage fluctuations.
A team of engineers captured data from across the eastern U.S. on different three-phase power quality instruments: a portable power recorder model and permanently installed models.
Using sampling hardware and algorithms, each monitor captured thousands of events as the voltage, current, and frequency changed in response to the confusion that afflicted the system. The team captured fast events using fast sampling and summarized slow events. Adaptive threshold software adjusted capture levels according to the rate of incoming events as they took place. This full disclosure monitoring is a method of measuring all aspects of power quality, on every voltage cycle, in appropriate detail and duration. It created a rich forensic picture that otherwise we would have lost.
Full disclosure technology
Full disclosure monitors make many simultaneous measurements:
- Voltage trends and events:
- Voltage transients: between one-half microsecond and eight milliseconds (half a cycle)
- Voltage disturbances: between half a cycle and two seconds (typically wave shape changes and brief rms events)
- Root mean square (rms) voltage events greater than two seconds: sags, swells, or outages
- Voltage imbalance
- Flicker: periodic voltage fluctuations less than 25 hertz as defined by IEC 868
- Power consumption: watts, volt ampere (VA), volt ampere reactive (VAR), power factor (PF) (true and displacement), demand, and kilowatt hours
- Current in-phase and neutral conductors
- Ground current
- Harmonic distortion: harmonic spectra for voltage and current for all conductors, total harmonic distortion (THD), tracking of individual voltage and current harmonics
Managing all of this data requires a large memory capacity, high-speed digital signal processors, and adaptive threshold technology. In a full disclosure monitor, thresholds start at very low values. If the rate of incoming events is greater than the memory's capacity for the monitoring period, the monitor's software automatically raises the thresholds in small increments on successive cycles to regulate the rate of capture. Likewise, if events slow down, the threshold lowers. That way, the monitor can capture not only severe events, but also a continuous cycle-by-cycle record of rms voltage and current history, and power consumption for the entire monitoring period.
By continuously recording severe events and the underlying quiescent data that indicates incipient problems, full disclosure monitors make long-term power quality analysis and predictive maintenance possible.
You can perform predictive maintenance analysis by comparing data from instruments with identical data capture techniques, using the same analysis methods, and with full information about the conditions at the monitoring site. Full disclosure instruments faithfully record the true conditions at the inception of the maintenance program and on an ongoing basis. They also consistently capture and analyze data the same way every time you use them.
Maps of region tweny hours before and seven hours after blackout.
After finishing each survey, you can download data and save it into a computer for comparison, analysis, and reporting. Multiple databases collected over long periods of time provide engineers with a comprehensive power history of a plant's power system utility infrastructure. By continuously tracking the changes in the power situation and comparing events on a weekly or monthly basis, you can easily identify deteriorating conditions.
By using full disclosure monitors during the outage, engineers in various regions were able to capture real-time power events in unprecedented detail.
Here's a list of details of the power outages by location.
Cities bordering Lake Erie were the hardest hit; Cleveland and Detroit were among the first cities to lose power. Seconds later New York State, including New York City and Long Island overloaded and went offline as current tried to flow west. Eastern interconnection was severed at New Jersey and Pennsylvania. They maintained power south of this line.
Northeast New Jersey
There was a power recorder connected to the service entrance of an industrial facility in northeast New Jersey during the blackout. Here is detail of the three outages and the eight hours following the initial event.
One monitoring point was just south of the New York and New Jersey transmission interconnections. The power was out for about ten minutes during the initial event. Two more short-duration outages occurred two hours later.
The chart shows detail of the rms voltage during the first outage. Within two minutes of the initial instability, the voltage collapsed. Nineteen minutes later the power was restored, and the power recorder had resumed monitoring. The voltage had not yet stabilized. It took another one and one-half minutes to get back to 480 root mean squared voltage (VRMS).
Eight minutes later the system became unstable again. The voltage fluctuated between 500 and 250 VRMS for three minutes.
Two hours after the power was restored, a second, brief outage occurred, but the system recovered. It took roughly fifteen seconds to reach 10% of nominal voltage. Seventeen minutes later a third, brief outage occurred. Again, it took roughly fifteen seconds to reach 10% of nominal voltage.
The chart shows medium voltage primary at a large commercial campus measured on a 120:1 PT. Power was out for just over twenty-four seconds before being restored.
In northern Ohio, where many believe the outage originated, the situation was much worse. A power recorder in northern Ohio caught the outage on a commercial panel.
The outage might have started in transmission lines and generating plants around the western banks of Lake Erie. This plant in northern Ohio was down for eleven and one-half hours. Nominal voltage is 120 volts, on a wye connected panel.
Within seconds of the start of the event, the voltage had dropped to zero. Eleven and one-half hours later, we saw a relatively graceful recovery and resumption of the normal business cycle.
A power recorder in Georgia picked up a short-term voltage disturbance and recorded it as a frequency variation.
Data from two sites in Georgia experienced frequency jitter. One site experienced less than a minute of voltage fluctuations.
This recorder in central Arkansas picked up some frequency jitter at 3:10 p.m. Central Daylight Time.
What we learned
The data shows that outages like the ones on 14 August 2003 are not one-time events, and the transitions are not clean. One example shows three outages over the space of a few hours, any one of which could cause equipment malfunction. This raises certain questions: How can I tell when the coast is clear? Is the power stable or can I expect more problems? Should I restart my process now or wait until things settle down a bit?
By studying the power quality across different regions, it is possible to build improvements and make preparations for future outages. For example, in geographical areas that experienced sags of just a few cycles, or frequency jitter, some small uninterruptible power supplies (UPSs) on sensitive equipment would probably suffice. Areas with outages of ten minutes or longer, or multiple short outages, would require larger UPSs. For facilities that saw outages of hours, a backup generator would be the only viable option.
Full disclosure monitoring and predictive maintenance have arrived just in time to manage skyrocketing electricity demand, connection loads, and downtime costs. And with the 2003 power outage, electrical system safety practices are more critical than ever.
Behind the byline
David Pereles, power quality marketing manager at Fluke Corporation, has an MBA from Seattle University and a B.S. in electrical engineering from Trinity College in Hartford, Conn. Pereles is also a member of IEEE PES, NETA, and NFPA.
By Darrick Finan
Organizations that rely on electronic systems are not fully aware of the potential for power problems until they happen—and then it's too late. From manufacturing systems to information systems, communication networks to physical transport networks, whatever the infrastructure in question, proactive planning and the right uninterruptible power system (UPS) can prevent the potentially devastating consequences of power disturbances.
A UPS performs two primary and complementary functions. It conditions incoming power to smooth out the sags and spikes common on the grid and other primary sources of power. It also provides ride-through power to cover for sags and short-term outages (say, thirty minutes to one hour) by dynamically selecting and drawing power from the grid, batteries, backup generators, and other available sources.
Three key types of UPSs are in use today:
Advancements in computer chips and other components over the past decade have dramatically increased the capabilities of UPSs while reducing heat output and cost. Within any of these categories, you can configure systems for a broad range of output capacities. You can also deploy multiple units to accommodate loads up to megawatts. In this modular architecture, you can add or remove components as needed. In fact, the system's design should permit you to take individual modules offline for maintenance without removing the load from conditioned power. y
-Behind the byline
Darrick Finan is director of product management at PowerWare in Raleigh, N.C.
NERC investigates 2003 blackout
The North American Electric Reliability Council (NERC) conducted a comprehensive investigation of the 14 August 2003 blackout and concluded the following:
Source: NERC Actions to Prevent and Mitigate the Impacts of Future Cascading Blackouts, 10 February 2004 (www.nerc.com).