Coordinated Distribution Network Reconfiguration and Power Flow Optimization Based on Optimal Policy Tree and Constraint Learning
IEEE Transactions on Power Systems · 2026-07 · 41 (4) · 3172-3185
DOI: 10.1109/tpwrs.2026.3661279 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Power Systems, 2026-07. Verfasst von Jianfeng Zeng und 2 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 31 Literaturangaben. Gefördert von: Science and Technology Development Fund, Macau SAR, Guangdong Provincial Department of Science and Technology.
Abstract (Originalsprache)
Distribution network reconfiguration can help enhance grid flexibility and promote integration of distributed renewable generation. However, achieving rapid and accurate solutions for distribution network reconfiguration remains challenging due to the combinatorial complexity introduced by numerous integer variables. This paper proposes a learning-based two-stage method for coordinated distribution network reconfiguration and power flow optimization. The method decomposes the problem into two sequential stages: distribution network reconfiguration and power flow optimization. In the distribution network reconfiguration stage, an optimal policy tree model incorporating hyperplane splits is proposed, transforming the optimization problem into a classification task based on historical operational data. The proposed model guarantees a globally optimal structure by formulating the tree construction process as a mixed-integer program. Based on this structure, topology strategies can be efficiently and reliably identified. In the subsequent power flow optimization stage, three neural network-based constraint learning models are employed to train surrogate models for power flow constraints. These surrogate models are then linearized and embedded into the power flow optimization model, removing the dependency on exact topology information. Simulation results on the IEEE 33-bus, 119-bus, and 301-bus systems demonstrate that the proposed method guarantees topology optimality in 88.3%–96.6% of scenarios, achieves an optimality gap below 5.7%, and delivers solutions within 36 seconds, without requiring precise topology and line parameter information.
Förderung
- Science and Technology Development Fund
- Macau SAR
- Guangdong Provincial Department of Science and Technology
Zitieren
APA
Zeng, J., Liu, D., & Zhang, H. (2026). Coordinated Distribution Network Reconfiguration and Power Flow Optimization Based on Optimal Policy Tree and Constraint Learning. IEEE Transactions on Power Systems, 41(4), 3172-3185. https://doi.org/10.1109/tpwrs.2026.3661279
Vancouver
Zeng J, Liu D, Zhang H. Coordinated Distribution Network Reconfiguration and Power Flow Optimization Based on Optimal Policy Tree and Constraint Learning. IEEE Transactions on Power Systems. 2026;41(4):3172-3185. doi:10.1109/tpwrs.2026.3661279
BibTeX
@article{Zeng2026,
author = {Zeng, Jianfeng and Liu, Dundun and Zhang, Hongcai},
title = {Coordinated Distribution Network Reconfiguration and Power Flow Optimization Based on Optimal Policy Tree and Constraint Learning},
journal = {IEEE Transactions on Power Systems},
year = {2026},
volume = {41},
number = {4},
pages = {3172--3185},
doi = {10.1109/tpwrs.2026.3661279}
}
EUVIMED