NREL Researchers Collaborate To Optimize Transmission Modeling Efficiency
Advanced computing plays a pivotal role in accelerating clean energy research at the National Renewable Energy Laboratory (NREL). Whether researchers need simulations of traffic systems, models of biomass conversion, or optimization of renewable energy implementation, NREL researchers utilize advanced computing capabilities to speed up renewable energy research.
Within NREL's Energy Systems Integration Facility, the Modeling, Simulation, and Optimization Capability (MSOC) team—a part of the Computational Science Center—works with the Grid Planning and Analysis Center (GPAC) and with researchers across NREL on grid modernization, energy systems integration, and mobility electrification projects.
Researchers for the National Transmission Planning Study (NTP Study) came to MSOC with a need: They wanted to create a nodal representation of the U.S. transmission system with 95,000 buses and 130,000 transmission lines. But existing software for modeling the transmission system did not have the capability to scale and support the size of the analysis envisioned in the study, which sought to model the whole United States for every node and hour of the year and use that data to link with other modeling frameworks.
The NTP Study is led by the U.S. Department of Energy's Grid Deployment Office and conducted in partnership with NREL and the Pacific Northwest National Laboratory. With the goals of providing transmission that spreads benefits to electricity customers nationwide, informing transmission planning processes, and identifying strategies to accelerate decarbonization while maintaining system reliability, the NTP Study needed robust calculations and modeling of transmission constraints.
José Daniel Lara, a GPAC researcher who is the lead developer of NREL's Sienna framework, had a design in mind to enable the NTP Study to scale problems. He turned to MSOC to help bring his vision to life and found researchers Bernard Knueven and Devon Sigler, who specialize in applied mathematics.
"We looked at the software and knew we could help the NTP Study with more efficient processes, and we had a talented postdoctoral researcher, Alessandro Castelli, who could help," Knueven said. "We collaborated with scientists in GPAC and implemented new software to calculate the power flow sensitivity matrix—which describes how line flows change when load or generation changes—up to 20 times faster while using five times less memory on large-scale instances. We also made it so that you only need to compute the parts you want to focus on."
Those improvements were integrated into a new software package, PowerNetworkMatrices, which is part of the Sienna modeling ecosystem. Instead of calculating an entire matrix to pick out a small section, the new software directly calculates a specific section.
"The power flow sensitivity matrix is there to do a calculation that allows you to understand how power is going to flow through the whole network, but if you're only interested in certain lines in the network, you actually only need certain rows of the matrix," Sigler said. "That saves you a massive amount of time."
Think of a bucket of marbles of all different colors, Sigler offered as an analogy. You could try and select all blue marbles from a larger bucket.
"This method generates a smaller bucket with only blue marbles in the first place," he said. "That makes for a more efficient, easier task. When you can focus on the thing you're really interested in observing and evaluating, your work will be easier."
The approach is more efficient, which reduces computational bottlenecks, and means that researchers can even use their laptops for some work. The MSOC team plans to continue collaborating on the NTP Study in hopes of making more computational improvements to project workflows.
"We haven't changed anything about the simulation," Knueven said. "We haven't simplified it. You get identical results. It's just much more efficient."
MSOC wants to help researchers run simulations without simplifying them. "Through high-performance computing and advanced math, we can enable people to run the original simulation without simplifying, because there are things you sacrifice in your analysis when you have to simplify simulations," Sigler said.
MSOC's work helped make the optimization model solvable so that the NTP Study team could focus on the desired experiments they wanted to run.
"With the technical challenges solved through this NREL collaboration, the NTP Study can more effectively highlight transmission options that benefit customers and inform transmission planning throughout the country," Lara said. The NTP Study and its key findings will be released later this year.
Learn more about the Computational Science Center, NREL's grid research, and the National Transmission Planning Study.