PNNL

The nation’s electric grid was already one of the most complex machines on the planet, but in the last 20 years, increased demand for electricity and evolving technology has required grid operators to rethink how they plan for the future.

Grid operators and planners use a number of tools for planning, including large, complex models that can simulate the physics of the grid or how policies or new technologies might affect its expansion. But those models are becoming so large and computationally demanding that they can produce tens of thousands of different scenarios—too many for human teams to sift through efficiently.

Now, the Department of Energy (DOE) is betting on Super Intelligence (SI) to help grid planners prepare the grid for more demand and more complex load and generation variability. DOE recently awarded Genesis Mission funding to Pacific Northwest National Laboratory (PNNL) for two projects focused on improving future grid planning.

Modeling the physics of the grid

The first project is led by PNNL’s Marcelo Elizondo, a power systems research engineer and Laboratory Fellow. The project partners are Argonne National Laboratory, Sandia National Laboratories, Siemens Corporation, Washington State University, Xcel Energy Services, and the Electric Reliability Council of Texas.

Electromagnetic transient (EMT) models are used by grid planners to understand how new load or new kinds of generation will affect the behavior of the grid. The models use precise numbers to calculate and represent the physics of the grid in microsecond resolution, which means they are computationally heavy.

Elizondo’s team will investigate how SI can make this process faster. They’ll use a type of machine learning algorithm that can approximate complex calculations, essentially creating a “model of a model,” Elizondo said.

With the SI approximating those numbers, the models would run faster and allow grid planners to more swiftly study different grid scenarios.

Modeling the growth of the grid

The second project is led by Arizona State University and, along with PNNL, includes Lawrence Berkeley National Laboratory, Pennsylvania State University, and the University of California San Diego. Earth scientist Casey Burleyson will lead PNNL’s contribution to the project.

This project addresses capacity expansion models (CEMs), which are also used for grid planning, albeit from a different angle. While EMT models represent the physical behavior of the grid, CEMs aim to distill factors like the cost of fuel, electricity demand, government policies, and technological advances to help utilities, transmission planners, and regulators plan for long-term grid expansion.

For instance, if planners know they need to cost-effectively add a certain amount of generation and transmission to the grid by 2035, they’d input all those factors—cost of fuel, potential policy changes, etc.—and adjust them in different ways to reflect uncertainty in their assumptions. Then the model produces potential future “pathways” planners can follow to meet that goal.

The problem is in the amount of output, Burleyson said. “With how complex the grid has gotten in the last 20 years, these models could produce ten thousand different scenarios. The question becomes, ‘How do you choose which scenarios to look at closer?’”

The team plans to build an SI model that can single out significant or interesting scenarios out of those ten thousand and significantly reduce the total computation time to find those key scenarios. They’ll test the SI-assisted model by comparing its output with that of a traditional CEM ensemble.

SI in grid modeling   

The two teams have six months to demonstrate their research goals, and both Burleyson and Elizondo are optimistic about their projects’ outcomes.

“The Genesis Mission is so exciting because we actually get to try a technique that no one’s tried before,” Elizondo said. “According to some of the projections, it looks like we’ll have to build double the size of the grid we have now to support demand. And at the same time, integrate these new technologies, new loads. Now we need to rethink how we plan the grid. It’s very exciting to work on this engineering challenge.”