Vassilis Kekatos

The power grid on which our society depends is a tangled network of interconnections that moves energy across continents. Aging infrastructure, a rise in domestic electricity consumption, and a trend toward incorporating more renewable energy has researchers seeking ways to modernize power systems.

ECE Assistant Professor Vassilis Kekatos studies how to smoothly integrate renewable sources and electric vehicles in a way that brings about a more reliable and efficient energy system. The integration of technologies and data processing to advance our power systems is often referred to as the smart grid, said Kekatos.

"We want our power system to be more green, more efficient, more robust, and to offer more participation for consumers to give back to the grid," said Kekatos. "The grid needs to be more controllable and we need better situational awareness of it."

By designing and employing optimization, real-time techniques, stochastic modeling, and tools from machine learning, Kekatos is seeking to model and predict the changing grid.

Kekatos likens the smart grid to a continent-wide electric circuit. Input from wind and solar introduces uncertainties into grid operation. In an effort to rapidly monitor grid conditions, Kekatos and his team have developed decentralized algorithms for processing measurements from synchrophasorsdevices that measure the electrical waves on the grid, which were invented by members of the Energy Systems Group at Virginia Techor information collected from smart meters to deduce the state of the grid.

Typical power output from the grid

Kekatos is experimenting with control systems to accomplish specific tasks. For instance, controlling different parameters allows your neighbor's rooftop solar panel to produce energy at a set voltage and pump it back into the grid. But when the sunny day clouds over, there's a large fluctuation in a matter of minutes or seconds, and this affects voltages for the rest of the neighborhood, and the power grid as a whole. Kekatos is looking for a way to correct the fluctuation and bring it back to a set level, thereby minimizing instability within the grid. He is currently researching a coordinated effort exploiting the power inverters that are already built into solar panels.

"Tools from machine learning and big data inform the shape of our future energy systems," said Kekatos. There are different variables that influence the power grid, including geographically specific prices in electricity markets, wind speeds, and weather patterns. Tools currently under development will exploit the patterns present in the data generated in transmission and distribution networks to respond in real-time to consumer behavior.

"Two decades ago, the Internet was realized by algorithms and TCP/IP protocols that control different layers of communications," said Kekatos. "We are trying to bring that knowledge to power grids."