The case for using more renewable energy is an easy sell: It’s cheaper than legacy power sources, reduces greenhouse gas emissions and curbs air and water pollution.
From a power company’s perspective, though, those benefits also come with a cost—uncertainty—particularly when it comes to planning and optimizing power flow. While renewable sources like wind and solar offer an enduring energy supply over the long haul, they also introduce fluctuations in the short term—both up and down—that can throw off the balance of power flowing through a network.
“We want to be able to use more and more renewable energy, but it is a challenge for the power companies who have to run this system,” says Jim Luedtke, a professor of industrial and systems engineering at the University of Wisconsin-Madison. “It’s a different type of energy resource.”
Luedtke specializes in stochastic and integer optimization, two techniques within the field of operations research that deal with uncertainty in a problem (stochastic) and questions that include a combination of discrete decisions (integer). Power systems optimization is a natural fit for those areas of inquiry, which is why he’s spent much of his career honing mathematical models to inform the field.
Luedtke is the principal investigator at UW-Madison for MACSER: Multifaceted Mathematics for Rare, High-Impact Events in Complex Energy and Environment Systems, a U.S. Department of Energy-funded project that spans three national labs, UW-Madison, the University of Chicago and Ohio State University.
As part of the project, Luedtke has worked with Line Roald, an assistant professor of electrical and computer engineering, and Rohit Kannan, a postdoctoral associate under Luedtke in the Wisconsin Institute for Discovery, to create a new mathematical model that could allow power operators to integrate more renewable energy—in part by sometimes backing off its production when needed to maintain system balance. They’ll present a paper on their work at the Power Systems Computation Conference in summer 2020 in Porto, Portugal.
Their proposed model also accounts for the ability to tap individual power generators within a network all the way up to their set limits before shifting to other generators, rather than striving to stay below those limits as current models do, Luedtke says.
“Because we’re being less conservative about making sure we’re always staying in these bounds when really you can do that after the fact, we’re able to provide a plan that’s more aggressive in using the renewable energy,” he says.
While this particular effort focused on optimization at short time scales, Luedtke is keen to examine longer intervals, which inherently include more decision variables and more uncertainty.
“I really like the power systems problem domain because it has these rich challenges,” he says.