Variability is a well-known characteristic of windpower, and system operators know they must plan for changes in wind generation over the course of a day. But when those plans fall short, voltage levels can drop quickly, forcing grid operators to dispatch resources to make up the difference—either by shedding load or bringing reserve generation online.
That’s exactly what happened in ERCOT on an evening in February, when a combination of events left the system operator short on power and long on demand (See Figure 1).
“On a macro level we forecasted load pretty close to what it turned out to be,” says Kent Saathoff, vice president of operations at ERCOT. “But there was a substantial difference between what we were getting from wind generators and what was reported to us in our look-ahead studies.”
Ultimately what happened in ERCOT was a non-event, in the sense that only interruptible load was affected and no involuntary outages occurred. Nevertheless, the episode illustrates the value of modern forecasting technology to provide accurate data for utilities and system operators who must keep the lights on.
While a rare combination of events created the February 26 emergency, none of the individual factors was unusual by itself.
ERCOT suffered unexpected shortages in both wind and non-wind generation sources—totaling nearly 1,600 MW—starting at around 6:00 p.m., just as its load was rising, roughly in accord with forecasts.
The combination of factors caused system frequency to drop suddenly at about 6:30 p.m. ERCOT responded by first calling upon its balancing energy supplies, and then by declaring its step-two emergency protocols—most notably cutting power to about 200 MW of interruptible industrial load. This succeeded in restoring frequency, and when non-wind generation started coming back online between 6:50 and 7:00 p.m., the interruptible load was restored.
While the wind discrepancy was the single biggest factor in the episode—representing a shortfall of nearly 1,000 MW continuously for about 30 minutes—it was nothing ERCOT hadn’t seen before, even in recent months. “On December 22 of last year, there was an even quicker and more significant decline in wind generation,” Saathoff says. “But that time we had more generation available than we did on February 26, so we could fill in the blank.”
ERCOT also had access to additional generation on February 26, in the form of non-spinning reserves—800 MW of which started generating power at around 7 o’clock. But because these were non-spinning reserves, they weren’t immediately available to support voltage—mainly because ERCOT’s forecasts didn’t indicate they’d be needed.
“The primary factor is the need to accurately forecast wind,” Saathoff says. “Wind is certainly a good resource and it has a lot of benefits. We just have to take its variability into account and develop forecasting tools that we can reliably incorporate into our system.”
Today ERCOT’s daily resource plans include forecasts from wind generators who sell power into the grid. Those forecasts went awry on February 26, but as it happened, ERCOT possessed data that accurately predicted the decline it saw in wind generation (see Figure 2). Its system planners weren’t using that data, however, because the AWS TrueWind forecasting system was being tested for application only to the nodal market ERCOT is developing for Texas.
“When we go into our new nodal system at the end of the year, we won’t be relying on the generators for wind forecast data,” Saathoff says. “We’ll do the forecast, and our test system did pretty well tracking the wind.”
Similar systems are being tested and deployed in other regions that expect significant growth in wind generation. Xcel Energy, for example, is testing a WindLogics system to help it more accurately predict wind generation in its Northern States Power territory (see “Taming the Wind,” p. 60, Fortnightly, February 2008). And in early April the New York Independent System Operator (NYISO) contracted with AWS TrueWind to provide a centralized forecasting system for the NYISO grid.
NYISO proposes to require wind generators to finance the costs of forecasting, and also to penalize wind generators who repeatedly fail to provide required meteorological data. In return, it would increase the amount of wind generation eligible for exemption from under-generation penalties and full compensation for over-generation.
“As more windpower projects are connected to the grid, they will require enhanced attention,” said Robert Hiney, NYISO interim president, in a statement. “The advance forecasts will allow us to accommodate windpower more accurately and reliably.” NYISO now draws power from less than 500 MW of wind generating capacity, but nearly 7,000 MW of proposed wind projects are moving through the grid-interconnection process.
Similar quantities of new windpower capacity are being developed in Texas, adding to its existing 4,400 MW of windpower capacity. “We need to try to get that wind forecast into our current zonal operating system,” Saathoff says, “so we’ll have a better representation of what the wind generation is going to do and we can commit in advance any other units that might be needed.”–MTB