Computer Science Thesis Oral
— 4:00pm
Location:
In Person and Virtual - ET
-
Traffic21 Classroom, Gates Hillman 6501 and Zoom
Speaker:
PRIYA LEKHA DONTI
,
Ph.D. CandidateComputer Science DepartmentCarnegie Mellon University
https://priyadonti.com/
Bridging Deep Learning and Electric Power Systems
Addressing climate change will require deep cuts in greenhouse gas emissions over the next several decades, in which electric power systems will play a key role. In this thesis, we provide several directions for the principled design and use of machine learning methods (with a particular focus on deep learning) to address climate-relevant problems in the electric power sector. In the first part of this thesis, we present statistical and optimization-based approaches to estimate critical quantities on power grids. Specifically, we employ regression-based tools to assess the climate- and health-related emissions factors that are used to evaluate power system interventions. We also propose a matrix completion-based method for estimating voltages on power distribution systems, to enable the integration of solar power. Motivated by insights from these projects, in the second part of this thesis, we focus on the design of deep learning methods that capture the physics and domain knowledge relevant to the settings in which they are employed. In particular, we leverage the toolkit of implicit layers to design (a) decision-cognizant forecasts in the context of stochastic optimization, (b) fast, feasibility-preserving neural approximators for optimization problems, and (c) provably robust deep reinforcement learning methods. These methods are directly applicable to power systems, as well as being broadly relevant for other physical and safety-critical domains. While part two demonstrates how power systems can yield interesting directions for deep learning, in the last part of this thesis, we demonstrate vice versa how insights from deep learning can yield fruitful directions for power systems research. Specifically, we show how methods inspired by the implicit layers literature can be used to assess policy-relevant inverse problems on the power grid. We further show how combining insights from implicit layers and adversarially robust deep learning can allow us to provide scalable heuristic solutions to two central problems in power systems: N-k security-constrained optimal power flow and stochastic optimal power flow. Overall, this thesis demonstrates how bridging insights from deep learning and electric power systems can help significantly advance methods in both fields, in addition to addressing high-impact problems of relevance to climate action. Thesis Committee: J. Zico Kolter (Co-chair) Inês Azevedo (Co-chair) Jeff Schneider M. Granger Morgan Yoshua Bengio (Université de Montréal) In Person and Zoom Participation. See announcement.
For More Information:
deb@cs.cmu.edu