https://doi.org/10.26089/NumMet.v27r438

Application of machine learning methods to accelerate the solution of the unit commitment problem

Authors

  • Sergey A. Romanov

Keywords:

unit commitment
machine learning
integer linear programming
classification
solution acceleration

Abstract

The unit commitment problem is central to short-term power system planning and is usually formulated as a large-scale mixed-integer linear programming problem. The paper investigates an approach to accelerating its solution using machine learning methods. Trained classifiers predict the state of each generator in every planning period, after which confidently classified generators are fixed, and the remaining lower-dimensional problem is solved by an exact method. The existing approach based on logistic regression is extended: six classification algorithms (logistic regression, random forest, gradient boosting, multilayer perceptron, support vector machine, and k-nearest neighbors) are compared and applied to the full model including minimum up-time and minimum down-time constraints. Experiments are carried out on three datasets from the pglib-uc library using leave-one-out cross-validation. Tree-based methods accelerate the exact solution by a factor of 2.2–3.4 with a median objective deviation below 0.05% on the largest FERC dataset (up to 1028 generators), whereas logistic regression, support vector machine, and multilayer perceptron lose reliability at this scale.


Downloads

Published

2026-10-21

Issue

Vol. 27 (2026): Issue 4.

Section

Methods and algorithms of computational mathematics and their applications

Author

Sergey A. Romanov

Omsk Department of the Sobolev Institute of Mathematics SB RAS

• PhD Student

References

  1. eliteme:knueven2020 B. Knueven, J. Ostrowski, and J.-P. Watson, “On Mixed-Integer Programming Formulations for the Unit Commitment Problem,” INFORMS J. Comput. 32 (4), 857–876 (2020).
    doi 10.1287/ijoc.2019.0944
  2. A. V. Ramesh and X. Li, “Machine Learning Assisted Model Reduction for Security-Constrained Unit Commitment,” in 2022 North American Power Symposium (NAPS), Salt Lake City, UT, USA, October 9–11, 2022 (IEEE Press, New York, 2022), pp. 1–6.
    doi 10.1109/NAPS56150.2022.10012236
  3. Y. Bengio, A. Lodi, and A. Prouvost, “Machine Learning for Combinatorial Optimization: A Methodological Tour d’Horizon,” Eur. J. Oper. Res. 290 (2), 405–421 (2021).
    doi 10.1016/j.ejor.2020.07.063
  4. A. S. Xavier, F. Qiu, and S. Ahmed, “Learning to Solve Large-Scale Security-Constrained Unit Commitment Problems,” INFORMS J. Comput. 33 (2), 739–756 (2020).
    doi 10.1287/ijoc.2020.0976
  5. S. Park, W. Chen, D. Han, et al., “Confidence-Aware Graph Neural Networks for Learning Reliability Assessment Commitments,” IEEE Trans. Power Syst. 39 (2), 3839–3850 (2024).
    doi 10.1109/TPWRS.2023.3298735
  6. F. Pourahmadi and J. Kazempour, “Unit Commitment Predictor with a Performance Guarantee: A Support Vector Machine Classifier,” IEEE Trans. Power Syst. 40 (1), 715–727 (2025).
    doi 10.1109/TPWRS.2024.3400405
  7. N. Sugishita, A. Grothey, and K. McKinnon, “Use of Machine Learning Models to Warmstart Column Generation for Unit Commitment,” INFORMS J. Comput. 36 (4), 1129–1146 (2024).
    doi 10.1287/ijoc.2022.0140
  8. S. Pineda and J. M. Morales, “Is learning for the unit commitment problem a low-hanging fruit?,” Electric Power Syst. Res. 207, Article Number 107851 (2022).
    doi 10.1016/j.epsr.2022.107851
  9. Y. Yang and L. Wu, “Machine Learning Approaches to the Unit Commitment Problem: Current Trends, Emerging Challenges, and New Strategies,” Electr. J. 34 (1), Article Number 106889 (2021).
    doi 10.1016/j.tej.2020.106889
  10. J. Milla and J. I. Pérez-D’iaz, “The Use of Artificial Intelligence for the Unit Commitment Problem: State of the Art,” in ELECTRIMACS 2024, Lecture Notes in Electrical Engineering (Springer, Cham, 2025), Vol. 1275, pp. 713–725.
    doi 10.1007/978-3-031-73921-7_54
  11. Power Grid Lib – Unit Commitment.
    https://github.com/power-grid-lib/pglib-uc Cited October 8, 2026.
  12. D. W. Hosmer, S. Lemeshow, and R. X. Sturdivant, Applied Logistic Regression , 3rd ed. (Wiley, Hoboken, 2013).
    doi 10.1002/9781118548387
  13. L. Breiman, “Random Forests,” Machine Learning 45 (1), 5–32 (2001).
    doi 10.1023/A:1010933404324
  14. J. H. Friedman, “Greedy Function Approximation: A Gradient Boosting Machine,” Ann. Statist. 29 (5), 1189–1232 (2001).
    doi 10.1214/aos/1013203451
  15. L. Prokhorenkova, G. Gusev, A. Vorobev, et al., “CatBoost: Unbiased Boosting with Categorical Features,” Advances in Neural Information Processing Systems 31, 6638–6648 (2018).
    https://arxiv.org/abs/1706.09516 Cited October 8, 2026.
  16. D. E. Rumelhart, G. E. Hinton, and R. J. Williams, “Learning representations by back-propagating errors,” Nature 323, 533–536 (1986).
    doi 10.1038/323533a0
  17. D. P. Kingma and J. Ba, “Adam: A Method for Stochastic Optimization,” arXiv:1412.6980 (2014).
    https://arxiv.org/abs/1412.6980 Cited October 8, 2026.
  18. C. Cortes and V. Vapnik, “Support-Vector Networks,” Machine Learning 20 (3), 273–297 (1995).
    doi 10.1007/BF00994018
  19. J. C. Platt, “Probabilistic Outputs for Support Vector Machines and Comparisons to Regularized Likelihood Methods,” Advances in Large Margin Classifiers 61–74 (1999).
  20. T. Cover and P. Hart, “Nearest Neighbor Pattern Classification,” IEEE Trans. Inf. Theory 13 (1), 21–27 (1967).
    doi 10.1109/TIT.1967.1053964
  21. W. E. Hart, J.-P. Watson, and D. L. Woodruff, “Pyomo: modeling and solving mathematical programs in Python,” Math. Program. Comput. 3 (3), 219–260 (2011).
    doi 10.1007/s12532-011-0026-8
  22. F. Pedregosa, G. Varoquaux, A. Gramfort, et al., “Scikit-learn: Machine Learning in Python,” J. Mach. Learn. Res. 12, 2825–2830 (2011).
    https://jmlr.org/papers/v12/pedregosa11a.html Cited October 9, 2026.
  23. Q. Huangfu and J. A. J. Hall, “Parallelizing the Dual Revised Simplex Method,” Math. Program. Comput. 10 (1), 119–142 (2018).
    doi 10.1007/s12532-017-0130-5

License

Copyright (c) 2026 С. А. Романов

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.