EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
4256. A Data-Driven Methodology for Contextual Unit Commitment Using Regression Residuals
Invited abstract in session TB-35: Risk Averse and Contextual Stochastic Optimization, stream Stochastic, Robust and Distributionally Robust Optimization.
Tuesday, 10:30-12:00Room: 44 (building: 303A)
Authors (first author is the speaker)
| 1. | Guzin Bayraksan
|
| Integrated Systems Engineering, Ohio State University | |
| 2. | Ogun Yurdakul
|
| Techniche Universität Berlin |
Abstract
We present a data-driven approach that integrates machine learning prediction models with optimization under uncertainty in the presence of contextual information by utilizing residuals from the learning models. We discuss an application of this approach for making unit commitment decisions under uncertain net load conditions to improve short-term power system operations. We propose several enhancements to the basic methodology and assess their effectiveness on several case studies conducted using real-world data collected from the California and New York Independent system operators. Results show that the proposed approach can significantly improve the out-of-sample performance.
Keywords
- Stochastic Optimization
- OR in Energy
- Machine Learning
Status: accepted
Back to the list of papers