From left: Mohammad Hassanzadeh, Yaman Evrenosoglu, and Hanif Livani work together in the Power Systems Laboratory, generating models to forecast energy demands and power generation. The laboratory was made possible by a donation from Dominion Virginia Power.
In a world where almost everything we do requires some amount of electricity, we take for granted that when we flip a switch the lights will go on. With the integration of alternative power sources and increasingly uneven demand for power, the problem of generating the right amount of power at the right time is getting more complicated. Power engineers like C. Yaman Evrenosoglu, who work on power grid operations must continually find new ways to keep up with evolving power systems.
Evrenosoglu, who joined ECE last fall as an assistant professor, combines both technical and energy market expertise. The variety of his research activities gives testament to the complex issues behind keeping the lights on — especially as power companies add more large-scale energy sources, such as solar and wind power.
“We are trying to incorporate these kinds of energy sources while keeping the grid reliable and secure,” says Evrenosoglu. Forecasting the state of the grid (i.e., complex voltages at nodes and the power flows along the transmission lines), for example, becomes more important with unconventional generators joining the grid. Some power sources, such as wind or solar power, aren’t always available, he says, and we need to plan for this.
Collecting information from the network, power system engineers can estimate the current state of the system, run contingency analyses, forecast the future power consumption and plan for the worst case scenario. The price of the energy is determined in real-time energy markets, and it is bound by availability of energy as well as the demand. For the immediate reliability of the power system, power plant operators need to know how much power will be demanded in the immediate future: whether seconds or hours.
State estimation is an extremely important function, and according to Evrenosoglu it has to be complemented by forecasting. “Before, when we had extremely low levels of unconventional generation, the system state changed very slowly due to the slow nature of change in demand, so we didn’t need to forecast,” he says. However, integrating large percentages of large-scale alternative energy sources that depend upon the capricious weather, makes forecasting an imperative, he explains.
Conventional state estimation methods use information that is not time-stamped, along with the assumption that the state of the system won’t change significantly over a couple seconds, Evrenosoglu explains. That assumption is slowly changing with the introduction of advanced measurement devices such as phasor measurement units (PMUs) invented by researchers at the Center for Power and Energy. PMUs provide time-stamped and synchronized measurements at extremely fast rates (i.e., 30 samples/sec) contrary to the conventional data provided once every few seconds. Evrenosoglu proposes exploiting the data provided by PMUs and adding the use of historical measurements to forecast the system state for the future: a couple seconds, or even an hour ahead.
Yaman Evrenosoglu tackles forecasting, fault detection, integration of renewable resources, and other issues facing the power system.
The amount of historical data used for these calculations depends on whether the forecast is real time or offline. “For real-time forecasting, you can’t employ much data because it takes time. We use a couple minutes or hours of data. Offline, we can utilize from days to years of data.”
Although Evrenosoglu admits that some events affecting the power grid may be unpredictable, he says that there is still a need for these predictive capabilities. “We have historical data and there are seasonal changes we can capture, and these should be incorporated.”
A big challenge with state forecasting is the computational burden — especially if historical data is to be added. There are many possible states to analyze, compare, and calculate. He is currently pursuing two different approaches to reducing the size of the problem. The first takes a statistical perspective to determine the probable states using a dynamic programming algorithm. He is working with Chao Wang for abstracting the data and reducing the problem. The second approach exploits a regression-based model for the state transition model using historical data and a classical extended Kalman filter.
Forecasting is just one of the issues facing power systems researchers. It’s also important to make sure that power distribution is not interrupted, and that problems can be identified and fixed quickly.
Currently, when lightening strikes a transmission line, an operator can gather certain information from the line, and can send a crew to the general location. The crew then must pinpoint the exact location. Evrenosoglu has developed a technique that is faster and more accurate than conventional fault location methods.
Using very high frequency traveling waves, his method captures the first traveling wave within the first microseconds of the fault, and works with wavelet transformations and support vector machines. According to Evrenosoglu, “wavelet transformation tells you if and when a frequency is introduced to a signal and when it disappears. It’s much faster and more accurate than previous methods.”
Evrenosoglu comes to Virginia Tech from the University of Nevada, Reno (UNR), where he served as an assistant professor in the Department of Electrical and Biomedical Engineering since 2008. Prior to joining UNR, he worked for Software & Information Systems group at Nexant, Inc. as a senior consultant for wholesale competitive electric energy markets for two years.
While at UNR, he founded and directed the Electrical Power and Renewable Energy Systems research laboratory, and created and taught two graduate level courses in power system operations. He received the Outstanding Teacher Award in 2009 from the University of Nevada IEEE student branch.
Evrenosoglu earned his B.S. and M.S. in Electrical Engineering from Istanbul Technical University in Turkey in 1998 and 2001, respectively, and his Ph.D. from Texas A&M University in 2006.