Title: End-to-End Learning and Optimization with the Applications in Power Systems
Date: August 9th, 2023
Time: 3:00-4:00 pm EDT
Room Location: Coda C1215 Midtown
Meeting ID: 253 032 974 435
Passcode: 5R52yk
Wenbo Chen
Machine Learning PhD Student
School of Industrial & Systems Engineering
Georgia Institute of Technology
Committee
1. Dr. Pascal Van Hentenryck (Advisor)
2. Dr. Yao Xie
3. Dr. Siva Theja Maguluri
Abstract
The increasing penetration of renewable generation and distributed energy resources requires new operating practices for power systems, wherein risk is explicitly quantified and managed. However, traditional risk-assessment frameworks are not fast enough for real-time operations, because they require numerous simulations, each of which requires solving multiple Security Constrained Economic Dispatch (SCED) problems sequentially. To tackle this challenge, a comprehensive set of studies is proposed to systematically develop end-to-end learning and optimization methodologies.
The first study proposes a novel just-in-time machine learning (ML) pipeline by analyzing the market-clearing optimizations of MISO in a principled manner. This pipeline addresses the main challenges faced in learning SCED solutions, i.e., the variability in load, renewable output, and production costs, as well as the combinatorial structure of commitment decisions. A novel combined classification-plus-regression architecture is also proposed, to further capture the behavior of SCED solutions.
The second study presents a novel End-to-End Learning and Repair (E2ELR) architecture for training optimization proxies for SCED problems. E2ELR combines deep neural networks with closed-form, differentiable optimization layers, thereby integrating learning and feasibility in an end-to-end fashion. E2ELR is also trained with self-supervised learning, removing the need for labeled data and solving of numerous optimization problems offline. E2ELR is evaluated on industry-sized power grids with tens of thousands of buses using an economic dispatch that co-optimizes energy and reserves. The results demonstrate that the self-supervised E2ELR achieves state-of-the-art performance, with optimality gaps that outperform other baselines by at least an order of magnitude.
Wenbo Chen
PhD Student | Machine Learning
Georgia Institute of Technology