A team led by David M. McKenney Family Associate Professor Andy Sun in the H. Milton Stewart School of Industrial and Systems Engineering (ISyE) placed sixth out of nine winners in the second round of the Department of Energy’s (DOE) ARPA-E Grid Optimization Competition. Previously, they placed third out of 27 teams in the first round of the competition. The team also includes A. Russell Chandler III Professor Santanu Dey and three Ph.D. students – Amin Gholami, Kaizhao Sun, and Shixuan Zhang – who are all advised by Associate Professor Sun.

The competition includes a series of challenges aimed at developing software management solutions to create a more resilient and secure American electricity grid. The first challenge tasked participating teams with finding solutions to a security-constrained optimal power flow (SCOPF) problem, a fundamental optimization problem in the operation and planning of power grids, while the second challenge expanded on the problem with additional complex constraints, such as unit commitment, line switching, and discretized transformer tap ratio and phase shift.

“Our team investigated the problem structure carefully and deployed various optimization techniques,” said Kaizhao. “Our code consists of two stages. In the first stage, we used parallel computing to explore different problem formulations, and report the best-found solution. In the second stage, we quickly recovered feasible solutions for contingencies, again, through parallelization of the computational tasks, and output them in a robust way.”

The team needed to figure out the optimal decisions for unit commitment – starting up or shutting down a generator, and for line switching – closing or opening a transmission line.

“We dug into a considerable amount of data and observed certain patterns, such as which generators are more likely to incur higher generation costs and which transmission lines are more likely to cause congestions in the network,” Kaizhao explained. “Based on these observations, we developed some efficient strategies that significantly improved our scores.”

Associate Professor Sun praised the team’s work.

“It is a tremendous achievement to develop such a robust and scalable optimization software, which for the first time enables U.S. power grid operators to solve the most complex grid-optimization problems in their daily operation,” he noted. “This is the result of years of hard work. I would like to applaud the three Ph.D. students – Amin, Kaizhao, and Shixuan – for their excellent teamwork, creativity, and perseverance, in addition to Santanu’s support.”

The winning teams will receive a combined total of $2.4 million in prize money, to be used for further development of their technologies. The details of the next step have not been released yet.

“We believe this series of competitions definitely benefits the development of algorithmic software for the modern power grid, and we look forward to learning about the new challenge,” Kaizhao concluded.