Monday, September 29, 2025 - 4:00pm
Event Calendar Category
LIDS Seminar Series
Speaker Name
Jan Drgona
Affiliation
Johns Hopkins University
Building and Room Number
32-155
The transition to sustainable energy systems demands new computational paradigms that integrate physics-based reasoning with the adaptability of modern machine learning. This talk presents a scientific machine learning (SciML) perspective on modeling, optimization, and control of dynamical systems in energy applications. We outline how physics-informed models, learning-to-optimize strategies, and learning-based control can be unified within a single SciML framework to achieve both interpretability and scalability.
Two recent advances will be highlighted. First, an operator-splitting approach for neural differential-algebraic equations (NDAEs), which integrates mechanistic models with neural networks and demonstrates robust extrapolation in engineering modeling tasks involving implicit constraints, such as conservation laws. Second, a self-supervised learning-to-optimize method for mixed-integer nonlinear programs (MINLPs) with feasibility guarantees, delivering high-quality approximate solutions in milliseconds, even on large-scale problems where traditional solvers fail.
Together, these advances illustrate how SciML can unlock new capabilities for sustainable energy systems, from power grid optimization to physics-informed digital twins for building operations.
Jan Drgona is an Associate Professor in the Department of Civil and Systems Engineering at Johns Hopkins University (JHU), where he is a core faculty member of the Ralph S. O’Connor Sustainable Energy Institute (ROSEI) and an associate member of the Data Science and AI Institute (DSAI). Before joining JHU, Jan was a Senior Data Scientist in the Physics and Computational Sciences Division at the Pacific Northwest National Laboratory (PNNL), where he continues to hold a joint appointment.
Jan previously worked as a postdoctoral researcher in the Mechanical Engineering Department at KU Leuven, Belgium, and received his PhD in Control Engineering from the Slovak University of Technology in Slovakia.
His research focuses on scientific machine learning for modeling, optimization, and control of cyber-physical systems with applications to sustainable energy.