Vertex-Guided Redundant Constraints Identification for Unit Commitment

Danny H.K. Tsang; Xuan He; Yize Chen; Yuxin Pan

REVIEW 3 major objections 5 minor 33 references

Vertex-Guided Redundant Constraints Identification for Unit Commitment

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Checking a constraint at the vertices of a bounding box around the relaxed UC region identifies it as redundant, reducing screening to one matrix operation and accelerating it by up to 8.8x.

desk verdict Solid, correct core algorithm for UC constraint screening; the ML-based cut enhancements are empirically useful but lack a formal safety guarantee. read the letter →

arxiv 2507.09280 v1 pith:WURT7DO5 submitted 2025-07-12 eess.SY cs.SY

Xuan He , Yuxin Pan , Yize Chen , Danny H.K. Tsang This is my paper · ORCID

classification eess.SYcs.SY

keywords unitcommitmentconstraintscreeningredundantconstraintsvertex-guidedouterapproximationlinearprogrammingloadoperatingrange

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper's central claim is that redundant transmission constraints in the unit commitment (UC) problem can be identified by checking constraints against the vertices of a hyperrectangle that contains the UC feasible region, rather than solving one linear program per line limit. The hyperrectangle is built from per-variable bounds obtained by solving a small number of LPs over a relaxation of the UC model, and a matrix operation collapses all vertex checks into one pass. Because power systems typically have far fewer generator variables than transmission lines, this shifts the screening cost from the number of constraints to the number of variables. An ensemble of vertex-guided screening followed by classic line-flow-guided screening provably removes the same constraints as the classic method while taking up to 8.8x less time on systems up to 2,383 buses. The practical payoff is a smaller, faster-to-solve UC model whose optimal solution is unchanged.

What carries the argument

The load-bearing machinery is the outer-approximation hyperrectangle $R_{\text{bound}} = \{\underline{y} \le y \le \bar{y}\}$, whose facets are the per-variable bounds obtained by solving $2|G|$ LPs over the relaxed UC region, together with the matrix identity $\omega = \Phi(\tilde A) \circ \tilde A(\bar{y} - \underline{y}) + \tilde A\underline{y} - \tilde b$ ($\Phi$ = elementwise unit step). The hyperrectangle replaces the unknown relaxed polytope; its vertices are just all combinations of the variable bounds, so no vertex enumeration is needed. Theorem 1 links strict satisfaction at every vertex to redundancy, and Theorem 2's $\omega_j < 0$ test checks every vertex at once through a matrix operation. The ensemble EOVL stacks VGS with LFGS on the leftover constraints to guarantee the same removal set as LFGS at a lower LP count.

What would settle it

For any load in the test set, compute the true optimal UC cost $C^*$ by solving the full UC model and compare it with the predicted cost bound $C = f_{NN}(\ell)(1+\epsilon)$. If $C < C^*$ on even one instance, run the cost-cut screening and solve the reduced UC model: a different optimal objective or a violation of a removed line limit would show that the tightened screening region excluded the true optimum, falsifying the claim that the cost-cut variant preserves the UC solution.

Watch

Extended reading notes

Core claim

The core discovery is a redundancy criterion with a formal guarantee: after relaxing the binary commitment variables to $[0,1]$, the paper solves one LP per decision variable to obtain upper and lower bounds, forming a hyperrectangle that contains the relaxed LP region and therefore the true UC feasible region. Theorem 1 proves that a constraint strictly satisfied at every vertex of this hyperrectangle is redundant for the true UC region. Theorem 2 turns the geometric condition into the algebraic test $\omega_j < 0$, with $\omega = \Phi(\tilde A) \circ \tilde A(\bar{y} - \underline{y}) + \tilde A\underline{y} - \tilde b$, where $\Phi$ is the elementwise unit step; this lets all constraints be screened simultaneously by one matrix operation. The ensemble EOVL runs this vertex-guided screening first and then applies classic line-flow-guided screening to the remaining constraints, which the authors prove yields exactly the same removed-constraint set as classic screening while solving far fewer LPs.

Load-bearing premise

The main load-bearing assumption is that the predicted cost bound $f_{NN}(\ell)(1+\epsilon)$ lies above the true optimal UC cost and that the predicted unit statuses are correct for the unseen load instance; if either fails, the tightened screening region can cut off the true optimum and the removed constraints may not be redundant for the original UC feasible region.

Editorial extensions

If this is right

Screening time for UC no longer has to scale with the number of transmission line limits; it can scale with the number of generator dispatch variables, which is typically much smaller.
The EOVL ensemble provably removes the same constraints as classic LFGS while requiring roughly |G| + (|J| - N_V) LPs instead of |J|, yielding up to 8.8x acceleration on the 2,383-bus test system.
A single screening run over a load operating range produces constraints redundant for every load in that range, so the reduced model can be reused across load scenarios without re-screening.
Adding a cost cutting plane or fixing predicted unit statuses removes additional constraints and shortens screening time on all tested systems, and the reduced UC models solve 12.44% to 80.21% faster with zero solution gap.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The same box-and-vertices argument should transfer to other mixed-integer programs where a tractable relaxation can be boxed by per-variable bounds, so the screening speed-up is not inherently limited to power-system UC problems.
The cost-cut and commitment-cut variants trade away the unconditional redundancy guarantee: they are safe only when the predictor gives a valid upper bound on the true optimal cost and correct unit statuses. One cheap safeguard would be to run a single feasibility LP to verify the predicted bound before accepting the reduced model.
For multi-period UC with ramping constraints, the number of decision variables grows with the time horizon, so the current |G| versus |J| advantage may shrink; whether the vertex-guided approach still accelerates screening there is an open question the authors leave for future work.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The thing to know: the base method—VGS plus the EOVL ensemble—is correct, and it does what it says. Building a hyperrectangle from variable bounds and then checking all constraints with one matrix operation is a genuinely neat trick. The equivalence result (EOVL removes exactly the same set as classic LFGS) is real, and the experiments on systems up to 2383 buses show consistent acceleration. This is the paper's actual contribution, and it deserves serious referee time.

Where the paper is also good: Theorems 1 and 2 are sound. I checked the scaling argument; the formula computes the maximum violation over the outer box, and strict negativity gives redundancy for the original UC region via Corollary 3. The load-range extension (Proposition 3) is clean and useful. The complexity analysis is honest. The writing is mostly clear, and the authors open-source the code, which makes the empirical claims reproducible.

The soft spot is the machine-learning part, and it is the paper's most overstated element. The cost cut and commitment cut tighten the screening region using predictions, but the formal redundancy guarantee applies only to the un-tightened region. If the learned cost bound falls below the true optimum, or if the KNN-predicted unit statuses are wrong, constraints removed for the tightened region are not guaranteed redundant, and the reduced UC model can silently lose optimality or feasibility. The authors acknowledge this for the commitment cut, but they still report 0% solution gap without error bars, and epsilon is chosen from validation minima. That is finite-sample evidence, not a certificate. The abstract's claim about "finding the same redundant constraints" is only literally true for the base EOVL; the schemes with cuts intentionally remove more.

These issues are real but partial. The base VGS/EOVL result is solid and the ML enhancements are clearly labeled as improvements, not the core theoretical claim. I would like to see the ML section reframed as model-dependent screening with explicit safety warnings, or replaced by a certified tightening. The experiments would also benefit from variance across training seeds and test samples.

Recommendation: deserve peer review, conditional accept. The core contribution is exact, reproducible, and practically relevant; the ML add-ons need revision before publication.

Referee Report

3 major / 5 minor

Summary. The paper proposes a vertex-guided constraint screening method (VGS) for single-period unit commitment. It relaxes the binary commitment variables, solves 2|G| LPs to obtain bounds on the decision variables, builds a hyperrectangle outer approximation of the feasible region, and uses the matrix operation ω = Φ(Ã)∘Ã(ȳ - y̲) + Ãy̲ - b̃ to identify constraints with ω_j < 0 as redundant. The paper then introduces EOVL, an ensemble of VGS and classic line-flow-guided screening (LFGS), and adds three enhancements: load operating ranges, cost cuts predicted by neural networks, and commitment-status cuts predicted by KNN. Case studies on systems up to 2,383 buses report up to 8.8x screening acceleration over LFGS and substantial solution-time reductions.

Significance. The core VGS/EOVL result is a valuable and mostly sound contribution: it replaces one LP per line limit with 2|G| LPs plus a matrix operation, and Theorem 2 gives a sufficient condition for redundancy that is correctly derived from the hyperrectangle vertices. The open-source implementation and experiments on multiple test systems are strengths. However, the ML-based enhancements in Sections IV-B and IV-C are not covered by the same formal redundancy guarantee, and the abstract's claim that the schemes 'find the same redundant constraints' is only valid for the base methods, not for the prediction-based schemes. The paper would be significantly strengthened by clearly separating certified results from heuristic enhancements and by either proving or explicitly disclaiming optimality preservation for the cost-cut and commitment-cut variants.

major comments (3)

[Section IV-B, Proposition 4] Proposition 4 proves only that constraints removed from Rbound,C are redundant for Ruc,C, not for the original feasible region Ruc. If the cost cut (17c) is not enforced in the final reduced UC model, removing such constraints can enlarge the feasible region with points whose cost is below C* and that violated only the removed constraints, thereby changing the optimal solution. The paper does not state that the final model retains (17c); Section V-D reports solution gaps only empirically. Furthermore, the relaxation parameter ε is selected from the minimum validation gap (Table VI), which gives no certificate that C ≥ C* on unseen load instances. Please either add the cost cut explicitly to the final model and state this, or provide a proof of optimality preservation for the original model, and qualify the abstract's 'same redundant constraints' claim accordingly.
[Section IV-C, Proposition 5] The commitment cut has the same structural issue, and the paper itself acknowledges it at the end of Section IV-C: 'these screened may involve more constraints that are non-redundant for Ruc, while ˆu(k) can be non-optimal.' Because the KNN predictors are selected for 100% validation accuracy (Table VII), a wrong prediction on an unseen instance means that constraints removed because they are redundant for Rbound,ˆu need not be redundant for Ruc, so the reduced model can be suboptimal or infeasible. This does not invalidate VGS/EOVL, but it means S6 and S7 are heuristics whose safety is only demonstrated on a finite test set. The paper should either provide a formal safeguard for these variants or present them explicitly as heuristic extensions, separating their claims from the certified S3/S4 results.
[Section V-B and Abstract] The claim that EOVL identifies the same removed set as LFGS is asserted rather than proved. It should be stated as a proposition with a clear proof, since it depends on the subset relationship between VGS-removed constraints and LFGS-removed constraints. More importantly, the abstract's 'up to 8.8x acceleration while finding the same redundant constraints' is supported only for EOVL (S3); the cost-cut and commitment-cut schemes S5-S7 remove additional constraints that are not certified redundant for the original UC. Please qualify the abstract and conclusion so that the formal guarantee is attributed only to the base VGS/EOVL methods.

minor comments (5)

[Algorithm 1, Step 5] Step 5 writes 'ω = ε(Ã) ∘ Ã ∗ (y − y) + Ã ∗ y − b̃', but Theorem 2 defines the same operation with Φ(Ã). Use a consistent symbol for the unit-step function.
[Theorem 1 proof] There is a typo: 'Colloary 1' should be 'Corollary 1'.
[Theorem 2, Eq. (13)-(14)] The displayed formula in (13) is ambiguous because the overlines and underlines on y are not clearly distinguished in the typeset version. Please ensure the final version explicitly shows the upper and lower bounds, e.g., φ(a_{j,p}) a_{j,p} (ȳ_p - y̲_p) + a_{j,p} y̲_p, so that readers can verify the derivation.
[Section II-B, Corollary 1] The proof of Corollary 1 would be easier to follow if the two cases (screening model with and without the target constraint) were separated into two paragraphs; currently the mixture can confuse readers about which region is being screened.
[Section III-C] Theorem 2 gives a sufficient condition for redundancy, not a necessary one. It would be helpful to state this explicitly, since the experiments show VGS removes fewer constraints than LFGS and readers should not infer that ω_j ≥ 0 implies non-redundancy.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the VGS/EOVL redundancy guarantee follows from polyhedral containment and LP bounds, and the ML enhancement caveats are explicitly stated rather than definitionally forced.

full rationale

The central derivation is self-contained. Theorem 1 states a standard sufficient condition: if a linear constraint is strictly satisfied at every vertex of a polytope, then its maximum over the polytope is below the bound, so the constraint is redundant. The variable bounds obtained from Eq. (6) define a hyperrectangle R_bound containing the relaxed region R_u (Section III-B, Proposition 2), and Corollary 3 transfers redundancy from R_bound to R_uc by monotonicity; no fitted parameter or target outcome is used to define these regions. Theorem 2's matrix condition (14) is derived by bounding each coefficient's vertex contribution by the appropriate variable bound (Eqs. 10-13), so the condition omega_j < 0 is a sufficient condition for Theorem 1; it is not equivalent to the screening outcome by construction. The EOVL 'same redundant set as LFGS' claim follows logically from the subset property (VGS-removed constraints are redundant for R_u and hence would also be removed by exact LFGS) plus running LFGS on the remaining constraints; it is an algorithmic consequence, not a fit. The ML enhancements (Sections IV-B and IV-C) do involve fitted inputs - epsilon selected from validation gaps and KNN status predictors chosen for 100% validation accuracy - and the reported 0% solution gap for cost cuts is conditional on C >= C*, which the paper explicitly states as necessary: 'To ensure UC feasibility and prevent the optimal UC solution from being excluded, it is necessary to ensure C is above the optimal UC cost C*' (Section IV-B). Section IV-C also explicitly acknowledges that with fixed statuses, 'these screened may involve more constraints that are non-redundant for Ruc, while u_hat can be non-optimal.' These are stated robustness and correctness caveats about the enhanced schemes, not circular derivations; the core S3/EOVL acceleration claim does not reduce to the fitted epsilon or KNN thresholds. Self-citations [12] and [13] are cited only as related improvements and are not load-bearing for Theorem 1, Theorem 2, or the EOVL equivalence. No uniqueness theorem or ansatz is smuggled in via self-citation. Therefore, no step in the claimed derivation chain is equivalent to its input by definition.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The core VGS derivation relies only on the convex polytope structure of the relaxed UC and standard LP vertex properties; it has no free parameters. The enhancements introduce fitted parameters (ϵ, KNN/NN settings, β) and empirical assumptions about predictor generalization. No new physical entities are postulated.

free parameters (4)

epsilon (cost-cut relaxation) = 0.005 for 39/118/300-bus, 0.01 for 500/2383-bus
Chosen from validation gap between predicted and optimal UC costs to make the cost cut C ≥ C*; it is fitted to training/validation data, not derived.
beta (load operating range) = 0.2 to 1.0 in experiments
User-specified width of the polyhedral load range in (20); controls how many constraints are screened out, not derived from first principles.
KNN configuration (number of predicted/fixed units) = e.g., 5/4 for 39-bus, 3/175 for 2383-bus
Selected to achieve 100% validation accuracy for fixed unit statuses; data-dependent model selection.
NN architecture = hidden sizes 100-50-30, ReLU
Architecture chosen by the authors for the cost prediction model; not derived from the problem.

assumptions (5)

domain assumption The relaxed UC region with u ∈ [0,1] is a convex polytope that contains the original mixed-integer feasible region; line flows follow the DC PTDF model.
Section II-A formulates the UC model with linear DC line limits and relaxes binary u in (4); the theoretical screening guarantees rely on this convex relaxation containing Ruc.
domain assumption The single-period UC formulation is representative; multi-period ramping and logic constraints are excluded.
Section II-A explicitly restricts analysis to single-period UC following [10],[22], so the method's guarantees do not automatically extend to multi-period UC.
standard math The LP optimum of the screening models lies at a vertex, and the strict inequality condition in Theorem 1 is sufficient for redundancy.
Used in Corollary 1 and Theorem 1; relies on fundamental properties of linear programming and polyhedra.
ad hoc to paper The relaxation parameter ϵ guarantees C ≥ C* for all unseen load instances in the cost-cut scheme.
Section IV-B sets C = f_NN(ℓ)(1+ϵ) with ϵ tuned on validation data; this only holds empirically, not as a theorem.
ad hoc to paper KNN predictors with 100% validation accuracy for unit statuses will remain correct on unseen instances.
Section IV-C and V-E use this to fix unit statuses; the paper explicitly notes a tradeoff if predictions are wrong.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Vertex-Guided Redundant Constraints Identification for Unit Commitment." pith.science (2026). https://pith.science/paper/WURT7DO5

@misc{pith2026250709280,
  author       = {Pith},
  title        = {Pith review of: Vertex-Guided Redundant Constraints Identification for Unit Commitment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WURT7DO5}},
  note         = {Machine review of arXiv:2507.09280}
}

read the original abstract

Power systems Unit Commitment (UC) problem determines the generator commitment schedule and dispatch decisions to realize the reliable and economic operation of power networks. The growing penetration of stochastic renewables and demand behaviors makes it necessary to solve the UC problem timely. It is possible to derive lightweight, faster-to-solve UC models via constraint screening to eliminate redundant constraints. However, the screening process remains computationally cumbersome due to the need of solving numerous linear programming (LP) problems. To reduce the number of LPs to solve, we introduce a novel perspective on such classic LP-based screening. Our key insights lie in the principle that redundant constraints will be satisfied by all vertices of the screened feasible region. Using the UC decision variables' bounds tightened by solving much fewer LPs, we build an outer approximation for the UC feasible region as the screened region. A matrix operation is then designed and applied to the outer approximation's vertices to identify all redundant constraints on-the-fly. Adjustments for the outer approximation are further explored to improve screening efficiency by considering the load operating range and cutting planes derived from UC cost and discrete unit status prediction. Extensive simulations are performed on a set of testbeds up to 2,383 buses to substantiate the effectiveness of the proposed schemes. Compared to classic LP-based screening, our schemes can achieve up to 8.8x acceleration while finding the same redundant constraints.

Figures

Figures reproduced from arXiv: 2507.09280 by the authors.

**Figure 1.** (a) Line flow-guided Screening. By relaxing binary variables u to [0, 1] in the original UC model, we can get a relaxed region containing the UC feasible region. The feasible region of the screening model (1), i.e., the screened region will be formulated to contain the relaxed region. An actual bound corresponding to the checked limit can be determined by the objective value of (1). The actual bound is proved to be … view at source ↗

**Figure 2.** Toy example for VGS. (a) is a 5-bus system with 2 units and 6 lines. To visualize the outer approximation, we let u = 1, and then this UC instance only has two variables x1 and x2 whose solution space is given in (b). The red lines are the bounds of x1 and x2 obtained by (6). They formulate the hyperrectangle H containing the UC feasible region FR. It shows the limits of f1 − f2 are under the vertex (3,3), which mea… view at source ↗

**Figure 3.** Illustration of the LP-based screening schemes. ( [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (7 more)

**Figure 4.** Figure 4: The screening ability for redundant constraints o [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: The screening speed of schemes S1-S3. The proposed [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

**Figure 6.** Figure 6: Comparison on computationally cost-effective pe [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

**Figure 7.** Figure 7: The screening results of S4 on different load opera [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

**Figure 8.** Figure 8: The screening performance of cost-cut integratio [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

**Figure 9.** Figure 9: The screening ability of S5-S7 [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]

**Figure 10.** Figure 10: The screening time of S5-S7. Table VII COMMITMENT CUT SETTINGS ON PREDICTED AND FIXED UNITS. systems 39 118 300 500 2383 K(No. Predicted units) 5(6) 5(5) 5(7) 3(13) 3(15) No.fixed units 4 45 58 74 175 Despite the increased complexity from additional constraints, scree…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 33 canonical work pages

[1]

A comp rehensive review on evolutionary optimization techniques applied fo r unit com- mitment problem,

N. Muralikrishnan, L. Jebaraj, and C. C. A. Rajan, “A comp rehensive review on evolutionary optimization techniques applied fo r unit com- mitment problem,” IEEE Access , vol. 8, pp. 132 980–133 014, 2020

work page 2020
[2]

A genetic algorithm solution to the unit commitment problem,

S. A. Kazarlis, A. Bakirtzis, and V . Petridis, “A genetic algorithm solution to the unit commitment problem,” IEEE transactions on power systems, vol. 11, no. 1, pp. 83–92, 2002

work page 2002
[3]

Applying reinforcement le arning and tree search to the unit commitment problem,

P . de Mars and A. O’Sullivan, “Applying reinforcement le arning and tree search to the unit commitment problem,” Applied Energy, vol. 302, p. 117519, 2021

work page 2021
[4]

Security-constrained unit commit- ment for electricity market: Modeling, solution methods, a nd future challenges,

Y . Chen, F. Pan, F. Qiu, A. S. Xavier, T. Zheng, M. Marwali, B. Knueven, Y . Guan, P . B. Luh, L. Wu et al. , “Security-constrained unit commit- ment for electricity market: Modeling, solution methods, a nd future challenges,” IEEE Transactions on Power Systems , vol. 38, no. 5, pp. 4668–4681, 2022

work page 2022
[5]

On the comple xity of the unit commitment problem,

P . Bendotti, P . Fouilhoux, and C. Rottner, “On the comple xity of the unit commitment problem,” Annals of Operations Research , vol. 274, no. 1-2, pp. 119–130, 2019

work page 2019
[6]

Tight a nd compact milp formulation for the thermal unit commitment problem,

G. Morales-Espa˜ oa, J. M. Latorre, and A. Ramos, “Tight a nd compact milp formulation for the thermal unit commitment problem,” IEEE Transactions on Power Systems , vol. 28, no. 4, pp. 4897–4908, 2013

work page 2013
[7]

Tran smission constraint ﬁltering in large-scale security-constrained unit commitment,

A. S. Xavier, F. Qiu, F. Wang, and P . R. Thimmapuram, “Tran smission constraint ﬁltering in large-scale security-constrained unit commitment,” IEEE Transactions on Power Systems , vol. 34, no. 3, pp. 2457–2460, 2019

work page 2019
[8]

Fast screen of redunda nt transmission constraints in line contingency-constraine d dispatch,

Z. Guo, W. Wei, L. Chen, and S. Mei, “Fast screen of redunda nt transmission constraints in line contingency-constraine d dispatch,” IEEE Transactions on Power Systems , vol. 35, no. 4, pp. 3305–3307, 2020

work page 2020

Show all 33 references

[9]

Fast identiﬁcation of inactive security constraints in scuc problems,

Q. Zhai, X. Guan, J. Cheng, and H. Wu, “Fast identiﬁcation of inactive security constraints in scuc problems,” IEEE Transactions on Power Systems, vol. 25, no. 4, pp. 1946–1954, 2010

1946
[10]

Implied constraint satis faction in power system optimization: The impacts of load variations,

L. A. Roald and D. K. Molzahn, “Implied constraint satis faction in power system optimization: The impacts of load variations, ” in 2019 57th Annual Allerton Conference on Communication, Control , and Computing (Allerton) . IEEE, 2019, pp. 308–315

2019
[11]

Cost- driven screening of network constraints for the unit commit ment prob- lem,

A. Porras, S. Pineda, J. M. Morales, and A. Jimenez-Cord ero, “Cost- driven screening of network constraints for the unit commit ment prob- lem,” arXiv preprint arXiv:2104.05746 , 2021

2021 arXiv
[12]

Enabling fast unit c om- mitment constraint screening via learning cost model,

X. He, H. Wen, Y . Zhang, and Y . Chen, “Enabling fast unit c om- mitment constraint screening via learning cost model,” arXiv preprint arXiv:2212.00483, 2022

2022 arXiv
[13]

Fast constr aint screening for multi-interval unit commitment,

X. He, J. Tian, Y . Zhang, H. Wen, and Y . Chen, “Fast constr aint screening for multi-interval unit commitment,” in 2023 62nd IEEE Conference on Decision and Control (CDC) . IEEE, 2023, pp. 577–583

2023
[14]

Identifying redundant constraints and imp licit equalities in systems of linear constraints,

J. Telgen, “Identifying redundant constraints and imp licit equalities in systems of linear constraints,” Management Science, vol. 29, no. 10, pp. 1209–1222, 1983

1983
[15]

Fast identiﬁ cation of redundant transmission capacity constraints with changea ble power and load injections,

Q. Zhai, W. Y uan, X. Lei, X. Li, and X. Guan, “Fast identiﬁ cation of redundant transmission capacity constraints with changea ble power and load injections,” in Proceeding of the 11th W orld Congress on Intelligent Control and Automation . IEEE, 2014, pp. 3139–3143

2014
[16]

Fa st identifying redundant security constraints in scuc in the p resence of uncertainties,

T. Ding, M. Qu, J. Bai, W. Jia, J. Wu, Y . He, and T. Chen, “Fa st identifying redundant security constraints in scuc in the p resence of uncertainties,” IET Generation, Transmission & Distribution , vol. 14, no. 13, pp. 2441–2449, 2020

2020
[17]

Tight and compact data-dr iven linear relaxations for constraint screening in unit commitment,

M. Awadalla and F. Bouffard, “Tight and compact data-dr iven linear relaxations for constraint screening in unit commitment,” IEEE Trans- actions on Energy Markets, Policy and Regulation , 2023

2023
[18]

Fast security-constrained o ptimal power ﬂow through low-impact and redundancy screening,

R. Weinhold and R. Mieth, “Fast security-constrained o ptimal power ﬂow through low-impact and redundancy screening,” IEEE Transactions on Power Systems , vol. 35, no. 6, pp. 4574–4584, 2020

2020
[19]

Redundant a nd nonbinding transmission constraints identiﬁcation metho d combining physical and economic insights of unit commitment,

Z. Ma, H. Zhong, T. Cheng, J. Pi, and F. Meng, “Redundant a nd nonbinding transmission constraints identiﬁcation metho d combining physical and economic insights of unit commitment,” IEEE Transactions on Power Systems , vol. 36, no. 4, pp. 3487–3495, 2021

2021
[20]

Eliminating re dundant line ﬂow constraints in composite system reliability evalu ation,

B. Hua, Z. Bie, C. Liu, G. Li, and X. Wang, “Eliminating re dundant line ﬂow constraints in composite system reliability evalu ation,” IEEE Transactions on Power Systems , vol. 28, no. 3, pp. 3490–3498, 2013

2013
[21]

A dat a-aided security constraint prescreening technique and applicati on to real-world system,

S. Zhang, H. Y e, F. Wang, Y . Chen, S. Rose, and Y . Ma, “A dat a-aided security constraint prescreening technique and applicati on to real-world system,” in 2019 North American Power Symposium (NAPS) . IEEE, 2019, pp. 1–6

2019
[22]

Dat a-driven screen- ing of network constraints for unit commitment,

S. Pineda, J. M. Morales, and A. Jim´ enez-Cordero, “Dat a-driven screen- ing of network constraints for unit commitment,” IEEE Transactions on Power Systems, vol. 35, no. 5, pp. 3695–3705, 2020

2020
[23]

Prediction of umbrella co nstraints,

A. Ardakani and F. Bouffard, “Prediction of umbrella co nstraints,” in 2018 Power Systems Computation Conference (PSCC) . IEEE, 2018, pp. 1–7

2018
[24]

Learning to solve large-s cale security-constrained unit commitment problems,

´A. S. Xavier, F. Qiu, and S. Ahmed, “Learning to solve large-s cale security-constrained unit commitment problems,” INFORMS Journal on Computing, vol. 33, no. 2, pp. 739–756, 2021

2021
[25]

War m-starting con- straint generation for mixed-integer optimization: A mach ine learning approach,

A. Jim´ enez-Cordero, J. M. Morales, and S. Pineda, “War m-starting con- straint generation for mixed-integer optimization: A mach ine learning approach,” Knowledge-Based Systems , vol. 253, p. 109570, 2022

2022
[26]

Learning to predict security constraints for large-scale unit commitm ent problems,

R. Sterzinger, J. Poland, M. B. Paulus, and D. Ch´ etelat , “Learning to predict security constraints for large-scale unit commitm ent problems,” IN SUBMISSION 11 in 2023 IEEE PES Innovative Smart Grid Technologies Europe (IS GT EUROPE). IEEE, 2023, pp. 1–5

2023
[27]

An algorithm for ﬁnding all vertices of convex polyhe- dral sets,

M. L. Balinski, “An algorithm for ﬁnding all vertices of convex polyhe- dral sets,” Journal of the Society for Industrial and Applied Mathemati cs, vol. 9, no. 1, pp. 72–88, 1961

1961
[28]

An algorithm for determining irrelev ant constraints and all vertices in systems of linear inequalities,

T. H. Mattheiss, “An algorithm for determining irrelev ant constraints and all vertices in systems of linear inequalities,” Operations Research, vol. 21, no. 1, pp. 247–260, 1973

1973
[29]

Mult i-parametric toolbox 3.0,

M. Herceg, M. Kvasnica, C. N. Jones, and M. Morari, “Mult i-parametric toolbox 3.0,” in 2013 European control conference (ECC) . IEEE, 2013, pp. 502–510

2013
[30]

Generating all vertices of a polyhedron is hard,

L. Khachiyan, E. Boros, K. Borys, V . Gurvich, and K. Elba ssioni, “Generating all vertices of a polyhedron is hard,” Twentieth Anniversary V olume: Discrete & Computational Geometry , pp. 1–17, 2009

2009
[31]

Identiﬁcation of umbre lla constraints in dc-based security-constrained optimal power ﬂow,

A. J. Ardakani and F. Bouffard, “Identiﬁcation of umbre lla constraints in dc-based security-constrained optimal power ﬂow,” IEEE Transactions on Power Systems , vol. 28, no. 4, pp. 3924–3934, 2013

2013
[32]

Inﬂuence of stochastic de pendence on network constraints screening for unit commitment,

M. Awadalla and F. Bouffard, “Inﬂuence of stochastic de pendence on network constraints screening for unit commitment,” arXiv preprint arXiv:2208.12139, 2022

2022 arXiv
[33]

Mat- power: Steady-state operations, planning, and analysis to ols for power systems research and education,

R. D. Zimmerman, C. E. Murillo-S´ anchez, and R. J. Thoma s, “Mat- power: Steady-state operations, planning, and analysis to ols for power systems research and education,” IEEE Transactions on power systems , vol. 26, no. 1, pp. 12–19, 2010

2010

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

A comp rehensive review on evolutionary optimization techniques applied fo r unit com- mitment problem,

N. Muralikrishnan, L. Jebaraj, and C. C. A. Rajan, “A comp rehensive review on evolutionary optimization techniques applied fo r unit com- mitment problem,” IEEE Access , vol. 8, pp. 132 980–133 014, 2020

work page 2020

[2] [2]

A genetic algorithm solution to the unit commitment problem,

S. A. Kazarlis, A. Bakirtzis, and V . Petridis, “A genetic algorithm solution to the unit commitment problem,” IEEE transactions on power systems, vol. 11, no. 1, pp. 83–92, 2002

work page 2002

[3] [3]

Applying reinforcement le arning and tree search to the unit commitment problem,

P . de Mars and A. O’Sullivan, “Applying reinforcement le arning and tree search to the unit commitment problem,” Applied Energy, vol. 302, p. 117519, 2021

work page 2021

[4] [4]

Security-constrained unit commit- ment for electricity market: Modeling, solution methods, a nd future challenges,

Y . Chen, F. Pan, F. Qiu, A. S. Xavier, T. Zheng, M. Marwali, B. Knueven, Y . Guan, P . B. Luh, L. Wu et al. , “Security-constrained unit commit- ment for electricity market: Modeling, solution methods, a nd future challenges,” IEEE Transactions on Power Systems , vol. 38, no. 5, pp. 4668–4681, 2022

work page 2022

[5] [5]

On the comple xity of the unit commitment problem,

P . Bendotti, P . Fouilhoux, and C. Rottner, “On the comple xity of the unit commitment problem,” Annals of Operations Research , vol. 274, no. 1-2, pp. 119–130, 2019

work page 2019

[6] [6]

Tight a nd compact milp formulation for the thermal unit commitment problem,

G. Morales-Espa˜ oa, J. M. Latorre, and A. Ramos, “Tight a nd compact milp formulation for the thermal unit commitment problem,” IEEE Transactions on Power Systems , vol. 28, no. 4, pp. 4897–4908, 2013

work page 2013

[7] [7]

Tran smission constraint ﬁltering in large-scale security-constrained unit commitment,

A. S. Xavier, F. Qiu, F. Wang, and P . R. Thimmapuram, “Tran smission constraint ﬁltering in large-scale security-constrained unit commitment,” IEEE Transactions on Power Systems , vol. 34, no. 3, pp. 2457–2460, 2019

work page 2019

[8] [8]

Fast screen of redunda nt transmission constraints in line contingency-constraine d dispatch,

Z. Guo, W. Wei, L. Chen, and S. Mei, “Fast screen of redunda nt transmission constraints in line contingency-constraine d dispatch,” IEEE Transactions on Power Systems , vol. 35, no. 4, pp. 3305–3307, 2020

work page 2020