Every day – multiple times a day – power grid operators make a complex calculation to solve a supply and demand problem.

They must quickly decide how much energy their systems need to produce to accommodate the demand from their users at that moment. On average, these calculations take about five to 15 minutes as dozens of factors are considered and resolved on how to get power from generators through lines to users. Thousands upon thousands of people rely on these calculations being correct for clean water, heat and many other aspects in their daily lives, making it a complex yet critical balancing act.

Assistant Professor Kyri Baker has been working to optimize those calculations in her research for years, developing algorithms that coordinate assets quickly to keep the lights on. But new demands, more intermittent renewable energy, and more general uncertainty in the system have made things increasingly complicated, she said. 

Charging a single electric vehicle can consume more energy than a house consumes in a whole day, for example, Baker said. That kind of demand has implications for our power generation and distribution infrastructure. At the same time, overall power generation is increasingly coming from renewable sources like solar and wind, which may fluctuate many times a day due to small changes in the weather. And the resulting power is being transmitted through systems that can fail at various points, as was seen during deadly winter storms in Texas in February 2021.

All of that means the calculations needed to keep the lights on will need to take more factors into account while also making decisions in seconds, not minutes, in the near future. That is going to be hard to do because these algorithms are complex, Baker said. 

“They take a lot of time to solve because there are so many variables and constraints at play. Solving this large optimization problem efficiently is a challenge we have been trying to address since the 1970s,” she said. “The key innovation here is that there are fundamental limits to how fast we can currently solve these large problems. 

“Ideally, we want to completely bypass solving an optimization and instead use all of the data that is already collected in the grid to train a predictive model to determine how power plants should operate. This takes seconds instead of minutes and can help us maintain grid stability and efficiency on faster timescales.”

Baker, who is based in the Department of Civil, Environmental and Architectural Engineering, was recently awarded an NSF CAREER Award to pursue that idea. The award provides up to $500,000 over five years to support the research and educational activities of early career faculty members who have the potential to become leaders in their field. Through the new award, Baker will work to speed those solution times through the use of historical data and machine learning – computer algorithms that improve automatically through experience over time.

Baker said operators have been collecting data for years, and appropriate machine learning models have been around for a while as well. But recent advancements in computing now allow for understanding of highly complex relationships that can be used to train the large-scale models needed to solve this problem. The most challenging piece, she said, will be convincing leaders at utilities this method is something they should adopt.

“We are trying to create these algorithms in a transparent way so operators can have confidence they will work,” she said. “This infrastructure is of one of the most complex and critical systems in our society, and it’s not just a matter of efficiency and cost, but also life and death. So there really is a need to show that this can be done in a safe and secure way where the reason for the decisions being made is clear to everyone involved.” 

Baker also has a courtesy appointment within the Department of Electrical, Computer and Energy Engineering and is a Fellow of the Renewable and Sustainable Energy Institute, a joint institute between the National Renewable Energy Laboratory (NREL) and CU Boulder. She completed her undergraduate degree and PhD in electrical and computer engineering from Carnegie Mellon University and served as a postdoctoral researcher at NREL before coming to CU Boulder.

As part of this CAREER award, Baker will also be creating a scholarship competition around machine learning for use with the power grid. She said she was particularly looking to engage with and reach LGBTQ students through that initiative as a member of that community herself.

“I think it is often a hidden trait about yourself and something I know I struggled with in my career – feeling like I couldn’t be myself going to conferences or applying for grants,” she said. “When I went to college there were only a few female professors, and I can’t think of one that was openly part of the LGBTQ community. So I want to let students that are part of that community know that someone like them is doing this work and is here for them.”

Baker said engaging younger researchers across disciplines – especially those with diverse perspectives and lived experiences – would be crucial when solving these issues and working to address larger ones like climate change.

“We are talking about completely changing how one of our biggest infrastructure systems operates with implications for transportation, water and many aspects of everyday life in the U.S.,” she said. “To do that we want and need everyone’s help.”