Over the last decade, the management of power through electrical grids has become increasingly efficient.
“The movement to mathematical optimization of grid scheduling has saved billions and billions of dollars,” said Dan Doulet Faculty Fellow Jim Ostrowski, the associate department head of the Department of Industrial and Systems Engineering.
Most electrical grids are managed by an independent system operator (ISO) that needs to balance a supply of electricity from multiple generators, at multiple price points, against the projected demand for electricity. Mathematical optimization of the problem allows ISOs to make minute adjustments almost instantly, adding up to considerable improvements in costs and adaptability.
Unfortunately, the grid software used in the United States was designed when electricity generation and demand were much simpler. Current programs struggle to account for distributed energy resources, like household solar panels, and intermittent power generated by renewables like wind and solar.
“Renewables are less expensive, but also add uncertainty,” Ostrowski said. “Better optimization will lead to more cost savings while transitioning to cleaner energy, but you need better techniques to run a grid in this kind of context.”
In 2018, the US Department of Energy’s Advanced Research Projects Agency-Energy (ARPA-E) created the nationwide Grid Optimization (GO) Competition to encourage the quick development of management software capable of handling the intricacies of the modern grid. The first two phases of the competition focused on maintaining instantaneous power flow and maintaining reliability during extreme events, while the final phase incorporated generator scheduling.
Ostrowski and two of his former students, Ben Knueven (PhD ’17) and Jonathan Schrock (PhD ’22), entered the final phase of the competition in 2022. They emerged as finalists, placing sixth out of 18 teams and securing $200,000 in future research contracts to continue developing their program.
Fast Development for a Fast Tool
Scheduling optimization can be difficult for many systems, but the immediacy of power grids adds an extra layer of challenge.
“For something like pit mining, where you’re moving trucks in and out of a site, you potentially have months to plan an optimal schedule,” said Ostrowski. “In the power systems world, it’s instantaneous. You need to always be monitoring and responding, so you can’t afford to spend hours trying to solve a problem. It’s all got to be done as fast as possible.”
The algorithm’s reaction time was not the only stopwatch to consider. The GO Challenge forced competing research teams to deliver their completed software in just one year, a lightning-fast deadline for such a complex project.
Fortunately, Ostrowski was in a team with people who were intimately familiar with the complex math behind the optimization problem—even if his status as their former advisor led to some minor bumps along the way.
“In the stressful moments, a former student may not feel comfortable saying what they really want to say,” Ostrowski reflected. “They’re trying to establish themselves and be independent, but the past power dynamics can make it awkward for them.”
By making it clear that he was receptive to their ideas and feedback, Ostrowski was able to let his former students take the reins. He is especially glad that Knueven’s talent with math got the chance to shine in the final product.
“Ben’s idea of a relaxing evening is to get a glass of scotch and read advanced math books,” Ostrowski said.
Beyond the Competition
When they receive their GO Competition contract award in January 2025, Ostrowski, Knueven, and Schrock are looking forward to rounding out their software with new features and other improvements.
“Now that we have time to actually sit down and think carefully about it, my expectation is that we’ll be able to cut the time to solution in half,” Ostrowski said.
They are also looking forward to commercializing their project, allowing the new capabilities they have added to streamline the addition of renewable and distributed power sources to real grids.
“Optimized grid scheduling saves money, which has an immediate impact on the general consumer,” said Ostrowski. “Now it can also have an environmental impact.”
Contact
Izzie Gall (865-974-7203, egall4@utk.edu)