Skip to content. Skip to main navigation.

UTA researchers optimize electricity distribution networks

Monday, October 22, 2018 • Media Contact: Herb Booth

computer scienceenergyresearchAll Topics

A team of researchers from The University of Texas at Arlington is combining several principles of machine learning to enable machines to control power networks and other complex dynamic systems more effectively during unexpected events.

Frank Lewis

Frank Lewis

Frank Lewis, the Moncrief-O’Donnell Professor of Electrical Engineering, Yan Wan and Ali Davoudi, associate professors of electrical engineering, are using a $220,000 Early-concept Grant for Exploratory Research, or EAGER, from the National Science Foundation to use real-time learning to create a unified theory on how to optimize microgrid capacity through DC distribution. Junfei Xie, an assistant professor of computer science at Texas A&M University – Corpus Christi, is assisting the team with data analysis.

Microgrid capacity in the United States is strained and DC, or direct current, distribution networks are emerging as alternatives to the current standard AC, or alternating current, networks. The networks are critical to the scalable integration of renewable energy resources and fleets of electric vehicles.

Yan Wan

Yan Wan

All of the systems operating in a complex environment act on each other. Uncertain spatiotemporal dynamics of the environmental patterns create new data types and enable the spatiotemporal-scenario data-driven decision approach.

Optimal control is the model-based process of determining how a system will react to cues over a period of time and how to make the system work at its most efficient based on that data. Reinforcement learning is a data-driven type of machine learning where computers learn to behave in an environment by performing actions, seeing the result and reacting accordingly to achieve a specific goal. The team’s goal is to combine both optimal control and reinforcement learning into a unified theory that will allow efficient, real-time feedback control of interacting complex systems.

“We think our methods of reinforcement learning show how to use deep learning for real-time feedback control,” Lewis said. “We’ll use the algorithms and tools we develop to create optimal power profiles for power electronic converters in DC distribution networks and help mitigate the adverse effects of intermittent source, uncertain load demand or faults.”

Ali Davoudi

Ali Davoudi

EAGER grants are awarded to researchers who propose potentially transformative research into high-risk, high-reward new subjects or methods that lead to innovation.

The team’s research is an example of innovative thinking in the area of data-driven discovery, one of the themes of UTA’s Strategic Plan 2020: Bold Solutions | Global Impact, said Jonathan Bredow, chair of the Department of Electrical Engineering.

“EAGER grants are awarded to faculty who are engaged in high-risk, high reward research, and Dr. Lewis, Dr. Wan and Dr. Davoudi are all accustomed working on research that pushes the boundaries of accepted engineering tenets,” Bredow said. “Their research is sure to set a solid foundation for future applications of machine learning that haven’t been considered.

--  Written by Jeremy Agor