Use of Machine Learning Models to Warmstart Column Generation for Unit Commitment

arXiv2110.06872MaRDI QIDQ6380244FDOQ6380244

Authors: Nagisa Sugishita, Andreas Grothey, K. I. M. McKinnon 

Publication date: 13 October 2021

Abstract: This work focuses on a solution method for unit commitment problems which are to be solved repeatedly with different data but with the same problem structures. Dantzig-Wolfe decomposition with a column generation procedure has been proved successful for the unit commitment problem. We demonstrate use of a neural network to generate initial dual values for the column generation procedure, and show how the neural network can be trained efficiently using dual decomposition. Our numerical experiments compare our methods with baselines and reveal that our approach solves test instances to high precision in shorter computational time and scales to handle larger-scale instances.

This page was built for publication: Use of Machine Learning Models to Warmstart Column Generation for Unit Commitment

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6380244)