CAREER: Geometry-based Control of Nonlinear Distributed Energy Resources in Smart Grids

This NSF CAREER project aims to develop a new way to control the growing number of distributed energy resources (DERs) such as solar panels, wind turbines, batteries, and small generators, that are transforming the nation’s power grid. As these devices become more widespread, traditional model-based control methods struggle to keep pace with the grid’s increasing complexity and data volume. The project will bring transformative change by introducing a geometry-based control paradigm that learns how these systems behave directly from data, rather than relying on detailed physical models that are often difficult to obtain and maintain. This will be achieved by identifying low-dimensional patterns, or “manifolds,” hidden within system measurements and using them to design scalable, real-time control strategies. The intellectual merit of the project includes advancing fundamental knowledge at the intersection of nonlinear dynamics, machine learning, uncertainty quantification, and distributed control. This will enable resilient and adaptive operation of DER-rich power systems. The broader impacts of the project include improving grid reliability and energy resilience, integrating research into undergraduate and graduate education, offering industry internships and workforce training opportunities, and preparing U.S. students to lead in the modernization of electrified communities. The project integrates nonlinear manifold learning with predictive control to regulate voltage, frequency, and power sharing in microgrids that can operate independently or connected to the main grid. Instead of building full mathematical models of each device, the research extracts governing dynamics from time-series measurements and constructs reduced-order representations in a learned latent space. These representations are used to design predictive control algorithms that compute optimal control actions in real time. To address noise, missing data, and limited sensor availability, the project develops probabilistic latent-space models that estimate uncertainty and incorporate it into control decisions through stochastic optimization. A distributed coordination layer enables multiple DERs to cooperate using only local communication, avoiding centralized computation and improving scalability and cyber resilience. The framework will be validated on device-level and microgrid testbeds and is designed to generalize to larger power networks. By combining geometry-based learning with real-time control, the project will establish a new data-centric foundation for managing complex electrified infrastructure systems. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2538831 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Javad Khazaei | Institution: Lehigh University, BETHLEHEM, PA | Award Amount: $533,119 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2538831 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2538831.html