The modern power grid is no longer a straightforward, one-way system.
Just 20 years ago, demand strictly dictated generation. Today, the grid is a massive, nonlinear web driven by distributed energy resources (DERs), such as solar arrays, wind farms, and battery storage. These resources rely on power electronics to interface with the grid, creating bidirectional flows where customers (including communities and even individual households) can both consume and supply electricity.
“So the question now is, how do we control these millions of devices?” says Javad Khazaei, an assistant professor of electrical and computer engineering in Lehigh University’s P.C. Rossin College of Engineering and Applied Science.
Khazaei recently received funding through the National Science Foundation’s Faculty Early Career Development (CAREER) Program for his proposal to develop a geometry-based control paradigm for DERs in smart grids.
The prestigious NSF CAREER award supports junior faculty members across the U.S. who exemplify the role of teacher-scholars through outstanding research, excellent education, and the integration of education and research. Each award provides approximately $500,000 over a five-year period.
The shift from centralized to data-driven control
Currently, grid control relies on a centralized optimization problem known as optimal power flow: data is collected from every generation/demand unit and sent to a single controller that optimizes the demand and generation balance on minute-based or hourly time frames.
That main controller, says Khazaei, decides exactly how much power should come from a specific power plant or resource without considering the grid conditions (i.e., grid models).
As the grid scales, the centralized approach is becoming impractical, as traditional control decisions cannot account for the complexity of a grid populated by millions of DERs. Khazaei and his team are pursuing a fundamentally different strategy to incorporate dynamic models into grid control, also known as predictive control, through a two-step process.
First, rather than relying on the precise physics of every individual device, which is computationally expensive and difficult to determine, they use systems behavioral theories to learn the behavior of complex nonlinear DERs directly from data. They then identify the geometric “shape” of that behavior—mapping its boundaries to ignore computational noise and focus only on the most critical variables. This allows them to simplify the system using reduced-order modeling.
“Rather than capturing every variable in a high-dimensional system, can we represent each DER with just a few simple differential equations?” he says. “In theory, this is possible.”
Simplicity as innovation
This emphasis on simplicity—mapping the geometric boundaries of a system’s behavior—is what makes Khazaei’s work novel.
By minimizing data requirements and reducing the computational burden of modeling the “full order” system, the team aims to slash the number of equations and data required to model large-scale power grids without sacrificing accuracy.
While the project centers on microgirds, these simplified models could eventually be applied to predictive control across the entire national grid, meaning systems could anticipate and prevent failures before they occur. Currently, he says, controlling the whole grid is nearly impossible because millions of nodes translate into millions of equations.
“With this new model, we will be able to predict what’s going to happen in the next few seconds or minutes given the current condition, and fix issues in real time,” he says. “Right now, we don’t have that ability. We have to take corrective action after the fact.”
The AI advantage
Preliminary results are promising, and the ongoing surge in data-driven methods and artificial intelligence is providing higher resolution data than ever before.
“We can use AI to our advantage,” he says. “It used to take months or years to develop control designs. I’m optimistic that our approach will eventually make control of large-scale systems more efficient, accurate, and robust.”
—Story by Christine Fennessy
About Javad Khazaei
Javad Khazaei, an assistant professor of electrical and computer engineering, leads the INTEGRITY Laboratory at Lehigh University, where his team develops the next generation of resilient and intelligent energy systems. His expertise spans data-driven and AI-based modeling, control, and optimization in microgrids and smart grids. Khazaei’s work is supported by the National Science Foundation, the Department of Energy, the Office of Naval Research, the Pennsylvania Department of Community and Economic Development, and other industry and government sponsors. He is a 2026 recipient of the NSF CAREER Award and holds a PhD from the University of South Florida.
