DGRI Lab advances electrical energy transition with a focus on machine learning (ML) applications, especially in distribution systems and microgrids. We are interested in modeling and operational challenges arising from the increasing penetration of heterogeneous stochastic resources and demand, often inverter-interfaced. These include distributed energy resources such as batteries and solar PV, novel power-electronics-based controls, flexible demand, and electrified transportation. We study definition, modeling, control, and optimization of power systems, covering a broad range of phenomena. Our research is grounded in a pragmatic engineering philosophy and addresses domain-specific challenges that can drive fundamental ML innovation with practical and transformative impact.

We develop theoretically grounded, computationally efficient, and safety-aware ML approaches for grids characterized by multi-scale dynamics, high stochasticity, limited observability, and stringent operational constraints. The applications span data-driven modeling, physics-informed estimation and simulation, and learning-based control and optimization. We apply supervised and unsupervised learning to classify, categorize, estimate, or forecast parameters of a partially observable model (plant), leverage supervised learning for its simulation, and use imitation and reinforcement learning for the optimal control of its high-dimensional, stochastic, and non-stationary state-action space.

DGRI Lab is currently involved in several major initiatives and pursue the following specific research areas:

• - ML-Driven power systems modeling, simulation, control, and optimization;
• - Grid Resiliency to Extreme Weather Events.