First Person With John Dumas
MJ: Texas weather stayed in the headlines in 2011, starting with the winter storm before the Super Bowl and extending through the terrible summer heat. How did this story play out with respect to the state’s large renewable energy base— particularly the wind resources?
Dumas: When we got into the week of August 1, there were about six days of what we call Emergency Energy Alert conditions. The loads were very, very high, and we used almost all of our reserves in supplying the load. We did not have to shed firm loads on any of those days. Any megawatt of load response that we got helped a lot, and any megawatt of wind generation that we got helped a lot. For three days in a row starting August 1, we set new peaks, and the peak on August 3 was 68,379 megawatts. On August 1, the wind accounted for about 4% of the generation on average and about 1.9% during the peak hour. On August 2, 5% for the day, 2.3% on peak. August 3, it was 4.7% for the day and 2.9% on the peak.
MJ: Watt for watt, was that 2% to 4% trickier to manage than gas turbines and so forth?
Dumas: The fundamental difference is that you have to think of wind more as “negative load.” Just like load, you have to try to forecast it, and just like load, you have to account for the error in your forecast. With a generator, I know what your lead time is, and I know when I need to tell you to be on-line and available. I know what your high sustainable limit is, so I know what your output capability is and where I can tell you to go. With a wind generation unit, you don’t really think of it so much as a generator but as negative load; it’s going to reduce my demand, and I’m going to have to cover with other generation by some amount. I have to be able to forecast how much that is and be able to account for the variability in the forecast.
MJ: How does that affect forecasting?
Dumas: We have models that forecast load, and we come up with our best load forecast. Historically, we know what the error is and what the volatility around that forecast usually is. The way we manage that is we submit ancillary services or reserves to put us in a position to manage any degree of error we had in our forecast. Wind is the same thing. You can net load and wind together and have a net load and develop some statistics around the net load forecast error. Your worst wind error doesn’t necessarily happen at the same time as your worst load forecast error. It’s really that combined load effect on the system you have to manage.
MJ: Does it introduce what feels like an element of randomness? Dumas: I would call it variability. It definitely introduces a new variable that you have to manage. Wind forecast error is another component of the risk you have to manage, and you have to be able to account for that in your ancillary services or the reserves that you carry.
MJ: What’s particularly challenging with regard to wind forecasting?
Dumas: What’s tricky about a wind forecast and why it has more volatility than load is you’re taking your wind speed and multiplying it by a power curve. The power curve is not linear, and at some point, small variations in wind speed can equate to significant variations in a wind turbine’s power output. Our forecasters John Dumas is Director of Wholesale Market Operations for the Electric Reliability Council of Texas (ERCOT), where he is responsible for real-time and day-ahead market operations and the monthly and annual congestion revenue rights auctions. He sat down with EPRI Journal to offer perspectives on the rambunctious Texas weather of 2011, the performance of the state’s large and growing wind generation, and the state of the forecasting art. “ The ability to forecast wind accurately increases your ability to manage the variation in wind; the more predictable wind is, the better you’re going to be able to plan your other generation around that. ” ~ John Dumas FIRST PERSON with John Dumas 2 6 EPRI JOURNAL are looking for a number of different weather events that can affect wind speed. Obviously, fronts moving through the area can affect wind speed, and in the summer when the daytime temperature heats up, it tends to reduce the wind speed. When the earth cools down at night, the wind speed tends to increase. Forecasters are also looking at wind feedback and adjusting their model as they move forward.
MJ: With the largest U.S. wind fleet right now connected to the ERCOT grid, how much wind power is ramping up or ramping down as wind condition changes?
Dumas: Well, you know you’re going to have a morning load peak and an evening load peak. And then in the summer, you’re going to have an evening load peak, and you know when it’s going to happen. The difference in the magnitude of that peak is very much temperature- and weather driven, and that’s what you’re trying to forecast for load. The peak hours for wind don’t always happen at the same time on the same day. What we’ve seen with wind generation is very large ramps—up to 3,500 megawatts.
MJ: Does that mean in a particular hour or in a relatively short time, there could be 3,500 megawatts ramping up?
Dumas: We observed a 3,000-megawatt drop-off in a 60-minute period in the morning on May 5 this year. Then, on the other side, we’ve seen the wind pick up almost 3,000 megawatts in less than 60 minutes. We saw that in the morning on September 22 this year.
MJ: That sounds like quite a challenge for system operators.
Dumas: It’s particularly challenging when it’s on the way down. When you see it start down, what you have to do is look at the forecast, make some estimate—is it going to go from 6,000 megawatts to zero, or is it going to go from 6,000 megawatts down to 4,000? And do I have enough other generation on my system available to offset that drop? Those are the challenges a system operator faces.
MJ: How does the presence of a large wind fleet with the potential for these large ramps either up or down affect the reserves that you have on hand and how you bring them onto the system?
Dumas: We use the forecast to determine how much generation we need to commit to serve load, given the wind generation forecast. We also buy what we call supplemental ancillary services—gas turbines that can be started in 30 minutes or less— to manage the change in wind generation. What we’ve done is incorporated that forecast uncertainty into our process for determining how many gas turbines we need or how much reserve we need that has to be able to ramp up in 30 minutes or less. We combine that and our simulation services and the units that are on-line with our dispatch and our new nodal system. This gives us the ability to re-dispatch the system every 5 minutes. Prior to that, we were having to make tight decisions every 30 minutes.
MJ: Are your operators slicing the day into 5-minute increments?
Dumas: They’re watching how things unfold in real time. We have a nodal market system that looks at the current demand and determines the most economical way to serve that demand. In between those 5-minute dispatches, you have regulation service, which does the second-to-second smoothing. If wind is ramping up, then it’s going to reduce the amount of gas generation that’s needed to serve the load. If wind is ramping down, then you’ve got to increase output so you maintain that power balance. That’s done by our nodal market system and our energy management system.
MJ: Last winter, national news coverage focused on the winter storm in Texas, with ice sliding off the roof of the new Cowboys arena and the damage to the power system. How did the weather and your wind resources play out during that pre–Super Bowl cold snap?
Dumas: That weather event was about instrumentation freezing up. We had over 100 generators that tripped off and a very high load because of the temperatures. Wind output was pretty good for the day; I don’t think wind was a factor. In that event, the story was all about the extreme weather conditions, the instrumentation that froze up, and the units that tripped.
MJ: Given what you’ve seen, as more wind resources have come online, have you formed an opinion about where the next waves of innovation need to come from? Where should the industry, meteorologists, and researchers focus our attention? “ We have a nodal market system that looks at the current demand and determines the most economical way to serve that demand. In between those 5-minute dispatches, you have regulation service, which does the second-to-second smoothing. ” ~ John Dumas WINTER 2011 2 7
Dumas: Wind integration, in general, is the first hurdle. We have got the CREZ (Competitive Renewable Energy Zones) project that’s going to build more transmission from west Texas into our load areas, which is going to increase our ability to transfer more wind output. And the ability to forecast wind accurately increases your ability to manage the variation in wind; the more predictable wind is, the better you’re going to be able to plan your other generation around that.
MJ: Will that be helped by more powerful computers? Or is it more dependent on spreading your wind resources across a wide area and averaging things out?
Dumas: I think that’s a good point. Your forecast is highly dependent on how accurate your weather models are. How volatile your wind output is––that’s dependent on the area your wind is in and the diversity. In Germany, forecasts are pretty good because their wind tends to be spread out over the country. It’s pretty diverse, and their wind volatility is much less than what we experience. In Alberta, Canada, in areas where there are mountains, wind volatility is pretty high, which makes forecasting very difficult. It’s all a matter of the weather patterns and the diversity. Another factor is the number of meteorological towers you have to measure and monitor the weather conditions.
MJ: Overall, you seem comfortable with the job you’ve got to do and your performance to date.
Dumas: You know, necessity is the mother of invention. Wind is here, and I think we’ve done a pretty good job of managing it and moving forward with the best ways to manage it. We developed a ramp rate forecaster that we put in place last year, which tries to predict the probability of a large ramp and the magnitude of that ramp over the next 6 hours. As far as I know, we’re the first ones to do that. So we look for better ways to predict what the wind’s going to do and improve our ability to manage and respond to that variability. I think we’ve shown that you can manage pretty large amounts of wind capacity and manage the variability of wind output through the use of your systems that dispatch, the use of ancillary services, and the use of your forecasting tools.