Doan, Hien Thanh, Lim, Yeji, Kim, Hongseok · Nov 8, 2026
Feasibility-Aware Learning for Network-Constrained Unit Commitment
Markdown · JSON · Original source
Authors Hien Thanh Doan, Yeji Lim and Hongseok Kim propose FALCON-UC, a learning framework for network-constrained unit commitment (NCUC), which plans generator commitments and dispatch while respecting power-system constraints. They say many learning-based approaches still need an optimization solver to repair or complete predictions, limiting speed; FALCON-UC instead uses a differentiable correction during training and deterministic repairs at inference, without solving a mixed-integer program then. On the evaluated 2383-bus system, the authors report 99.09% commitment accuracy, a 0.4350% mean absolute cost gap, and all evaluated constraints within tolerance at 0.0562 seconds per instance. They report a 2.27×10³ speedup over the solver. Their comparison method, MTLR R-SCUC-FL, is slightly more accurate but still requires a solver, taking 262.2 seconds per instance. A label-only baseline without repair violates balance and branch limits on every test instance in that system; FALCON-UC also meets all evaluated constraint tolerances on a 1354-bus system. The results suggest a way to reduce solver use in these evaluations, but the abstract does not establish performance beyond the tested systems and conditions.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
eess.SY · Article
Source details
Original source date: Nov 8, 2026 (2026-11-08).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-12T11:02:29.747Z. First observed: 2026-11-10T08:36:48.097Z. Retrieved: 2026-11-12T10:47:21.640Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
The evidence is the paper's abstract; it does not provide the evaluation setup or further details needed to assess the reported results. The reported performance is for the stated 2383-bus and 1354-bus evaluations; results on other systems or operating conditions are not established by this evidence.
Open question
How does FALCON-UC perform on systems and operating conditions beyond the 2383-bus and 1354-bus evaluations?
Source-stated
The authors propose FALCON-UC, which predicts commitment and dispatch jointly and produces schedules without solving a mixed-integer program at inference time.
This paper proposes FALCON-UC, a feasibility-aware graph learning framework that predicts commitment and dispatch jointly and produces final schedules without solving a mixed-integer program at inference time.
Source-stated
On the evaluated 2383-bus system, the authors report 99.09% commitment accuracy, a 0.4350% mean absolute cost gap and all evaluated constraints within tolerance at 0.0562 seconds per instance.
On the 2383-bus system, FALCON-UC achieves $99.09\%$ commitment accuracy and a $0.4350\%$ mean absolute cost gap with all evaluated constraints within tolerance at $0.0562$ s per instance
Source-stated
The authors report a 2.27×10³ speedup over the solver; their comparison method is slightly more accurate but still uses a solver to complete each schedule.
a $2.27\times10^{3}$ speedup over the solver. MTLR R-SCUC-FL is slightly more accurate but still needs an optimization solver to complete each schedule
Source-stated
The authors report that a label-only baseline without repair violates balance and branch limits on every 2383-bus test instance, while FALCON-UC meets evaluated constraint tolerances on a 1354-bus system.
A label-only baseline without repair violates the balance and branch limits on every 2383-bus test instance, and FALCON-UC also meets all evaluated constraint tolerances on a 1354-bus system.
Source and contributor context
Doan, Hien Thanh, Lim, Yeji, Kim, Hongseok · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Editorial provenance
Version 1. Publication ID: 74b59040-a300-4fb2-8a6b-a88684a3e743.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-12T11:02:29.747Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: cdc809ac423d19f92779fadd0484179e6a5937e24816c57c0b7f3f2a0940e5aa. Source text hash: f82a9b89da6112c0d91466a4df674476c28fff9b1345510100e94fdcd0d56b99.
Revision and source-check history
Version 1 · 2026-11-12T11:02:29.747Z
Feasibility-Aware Learning for Network-Constrained Unit Commitment
Authors Hien Thanh Doan, Yeji Lim and Hongseok Kim propose FALCON-UC, a learning framework for network-constrained unit commitment (NCUC), which plans generator commitments and dispatch while respecting power-system constraints. They say many learning-based approaches still need an optimization solver to repair or complete predictions, limiting speed; FALCON-UC instead uses a differentiable correction during training and deterministic repairs at inference, without solving a mixed-integer program then. On the evaluated 2383-bus system, the authors report 99.09% commitment accuracy, a 0.4350% mean absolute cost gap, and all evaluated constraints within tolerance at 0.0562 seconds per instance. They report a 2.27×10³ speedup over the solver. Their comparison method, MTLR R-SCUC-FL, is slightly more accurate but still requires a solver, taking 262.2 seconds per instance. A label-only baseline without repair violates balance and branch limits on every test instance in that system; FALCON-UC also meets all evaluated constraint tolerances on a 1354-bus system. The results suggest a way to reduce solver use in these evaluations, but the abstract does not establish performance beyond the tested systems and conditions.
Original source
eess.SY · Article
Source details
Original source date: Nov 8, 2026 (2026-11-08).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-12T11:02:29.747Z. First observed: 2026-11-10T08:36:48.097Z. Retrieved: 2026-11-12T10:47:21.640Z.
Limits
The evidence is the paper's abstract; it does not provide the evaluation setup or further details needed to assess the reported results. The reported performance is for the stated 2383-bus and 1354-bus evaluations; results on other systems or operating conditions are not established by this evidence.
Open question
How does FALCON-UC perform on systems and operating conditions beyond the 2383-bus and 1354-bus evaluations?
Source-stated
The authors propose FALCON-UC, which predicts commitment and dispatch jointly and produces schedules without solving a mixed-integer program at inference time.
This paper proposes FALCON-UC, a feasibility-aware graph learning framework that predicts commitment and dispatch jointly and produces final schedules without solving a mixed-integer program at inference time.
Source-stated
On the evaluated 2383-bus system, the authors report 99.09% commitment accuracy, a 0.4350% mean absolute cost gap and all evaluated constraints within tolerance at 0.0562 seconds per instance.
On the 2383-bus system, FALCON-UC achieves $99.09\%$ commitment accuracy and a $0.4350\%$ mean absolute cost gap with all evaluated constraints within tolerance at $0.0562$ s per instance
Source-stated
The authors report a 2.27×10³ speedup over the solver; their comparison method is slightly more accurate but still uses a solver to complete each schedule.
a $2.27\times10^{3}$ speedup over the solver. MTLR R-SCUC-FL is slightly more accurate but still needs an optimization solver to complete each schedule
Source-stated
The authors report that a label-only baseline without repair violates balance and branch limits on every 2383-bus test instance, while FALCON-UC meets evaluated constraint tolerances on a 1354-bus system.
A label-only baseline without repair violates the balance and branch limits on every 2383-bus test instance, and FALCON-UC also meets all evaluated constraint tolerances on a 1354-bus system.
Source and contributor context
Doan, Hien Thanh, Lim, Yeji, Kim, Hongseok · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Editorial provenance
Version 1. Publication ID: 74b59040-a300-4fb2-8a6b-a88684a3e743.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-12T11:02:29.747Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: cdc809ac423d19f92779fadd0484179e6a5937e24816c57c0b7f3f2a0940e5aa. Source text hash: f82a9b89da6112c0d91466a4df674476c28fff9b1345510100e94fdcd0d56b99.
passed · 2026-11-12T11:02:29.747Z · anthropic/claude-opus-5.5
Publication: 74b59040-a300-4fb2-8a6b-a88684a3e743
attribution — pass: The reading correctly credits Hien Thanh Doan, Yeji Lim and Hongseok Kim as the authors. Performance figures are kept as author reports ('the authors report'), not presented as established fact.
Authors:Hien Thanh Doan, Yeji Lim, Hongseok Kim
summary — pass: The summary leads with the actor and the proposal. It gives a plain-language gloss of NCUC and states the core argument: earlier learning methods still need a solver to repair predictions, which limits speed, while FALCON-UC avoids this at inference. The numbers, the comparison and the scope limits are faithful. 'Mixed-integer program', 'bus' and 'inference' are not explained in the text and are candidates for term-library tooltips, but the central idea is still understandable. The question follows directly from the stated scope.
usually still rely on an optimization solver to repair or complete their predictions, which limits the speedup they can deliver
date — pass: The originalPublished value, 2026-11-08, matches the arXiv v1 submission date. The feed date of 2026-11-10 is correctly not used.
[Submitted on 8 Nov 2026]
support — pass: All four claims match the quoted abstract text. The closing inference, that the results 'suggest a way to reduce solver use in these evaluations', is scoped to the tested systems. The reading explicitly does not extend the results to real-world performance. No benefits or costs are invented.
MTLR R-SCUC-FL is slightly more accurate but still needs an optimization solver to complete each schedule, which takes $262.2$ s per instance
media — pass: The selected media is the verified arXiv abstract page for this paper. It is the direct original source.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXiv