Quantum Inspired QUBO Assisted ALNS for Reliability Driven Hurricane Restoration of Distribution Networks

Anurag Srivastava; Hooman Torkaman; Jignesh Solanki

REVIEW 3 major objections 4 minor 16 references

A QUBO-assisted repair-batch selection layer inside an ALNS scheduler reduces mean restoration makespan by 50.71% and improves SAIDI/ENS by 2.24% over classical energized ALNS on a 100 m/s hurricane stress test of the IEEE 123-node feeder.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 04:55 UTC pith:LIL2L4Y4

load-bearing objection A sensible QUBO-in-ALNS integration with an honest frame, but the makespan headline is likely an artifact of compute and rule changes, not the QUBO itself. the 3 major comments →

arxiv 2607.29544 v1 pith:LIL2L4Y4 submitted 2026-07-31 eess.SY cs.SY

Quantum Inspired QUBO Assisted ALNS for Reliability Driven Hurricane Restoration of Distribution Networks

Hooman Torkaman , Jignesh Solanki , Anurag Srivastava This is my paper

classification eess.SY cs.SY

keywords QUBOALNSdistribution system restorationhurricane resilienceSAIDIenergy not suppliedcrew schedulingsimulated annealing

verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper tries to establish that a quantum-inspired quadratic unconstrained binary optimization (QUBO) layer can improve the local batch-selection step of an adaptive large neighborhood search (ALNS) scheduler for post-hurricane distribution repair. On the IEEE 123-node feeder under a 100 m/s wind scenario, the QUBO-assisted method reduces mean system average interruption duration index (SAIDI) and energy not supplied (ENS) by 2.24% and restoration makespan by 50.71% compared with classical energized ALNS. The authors argue the benefit grows with damage severity because severe damage creates a larger combinatorial space of individual and multi-job repair alternatives. A deterministic decoder and OpenDSS replay keep the schedules operationally and electrically feasible, so the suggestion is that a modest local optimization layer can be added without disturbing trusted logistics and validation layers.

Core claim

The central claim is that replacing the local stage-batch selector in energized ALNS with a QUBO-assisted sampler—while keeping the deterministic crew-truck decoder and OpenDSS validation intact—produces better restoration schedules when the repair space is large. At 100 m/s, where 22 repair series and 40 expanded mapped lines must be coordinated, the QUBO-assisted method cuts average makespan from about 7.10 h to 3.50 h and improves SAIDI and ENS by 2.24%. The paper also shows that the benefit is damage-dependent: at 80 m/s, with only four active repair series, the classical selector performs better on SAIDI and ENS, and the QUBO layer helps mainly by shortening makespan. The authors theref

What carries the argument

The central object is the stage-level QUBO model over repair 'atoms', where each atom is either an individual repair job or a multi-job combination near the energized frontier. The objective minimizes a binary Hamiltonian whose linear terms combine normalized restoration utility per atom (newly reachable load and customers per unit time, minus completion time), a useful-crew capacity reward, and a mild atom penalty; the quadratic terms penalize overlapping atoms and reward graph synergy between compatible jobs. This QUBO is solved locally on a CPU by simulated annealing (300 reads, 1200 sweeps per call), and its sampled solutions are passed through the original deterministic logistics decode

Load-bearing premise

The reported gains assume the QUBO formulation itself—not its far larger inner-loop compute budget—is responsible for the better schedules, since classical ALNS was given only about 16 seconds per run while QUBO-assisted ALNS used about 201–361 seconds, and no wall-clock-matched or iteration-matched classical baseline was included.

What would settle it

Run the classical ALNS selector with the same local search budget (300 simulated-annealing reads and 1200 sweeps per stage) and check whether the 50.71% makespan reduction over the QUBO-assisted method persists; if it shrinks to near zero, the claimed QUBO benefit would be a compute artifact rather than a scheduling-formulation advantage.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

Under severe hurricane damage, the QUBO-assisted scheduler can reduce restoration makespan by roughly half, largely by eliminating the long restoration tail seen in the classical schedule.
The improvement is damage-dependent: utilities should activate QUBO-assisted batch selection only when the candidate repair pool is large enough to make coordination valuable.
Because the QUBO layer plugs into the existing ALNS and deterministic logistics decoder, it can be tested on other distribution feeders without changing crew-truck or OpenDSS validation logic.
The QUBO formulation is BQM/Ising compatible, so the same stage-selection layer can be ported to hybrid quantum-classical solvers in future work without redesigning the surrounding scheduler.
The reported crew utilization shows that QUBO assistance helps assign all 100 useful crews after a source-side bottleneck is cleared, pointing to better parallel use of field resources.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the QUBO benefit survives a wall-clock-matched comparison, the same stage-level selection pattern could be applied to other resource-constrained repair scheduling problems, such as post-earthquake gas or water network restoration where combinatorial batch coordination dominates.
The makespan reduction of 50.71% appears driven by tail compression; an editorial extension is to test whether adding a tail-penalty term to the objective (rather than just makespan weight) yields even larger gains in severe scenarios.
The reported zero standard deviation on makespan across three stochastic runs at 90 and 100 m/s suggests the runs may share initialization or random seed; independent trials with varied seeds would provide a stronger statistical basis for the claimed improvement.
A concrete testable prediction is that the QUBO advantage will grow monotonically with the number of compatible multi-job atoms in the pool; this could be checked by scaling damage severity while holding the logistics decoder fixed.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

A sensible QUBO-in-ALNS integration with an honest frame, but the makespan headline is likely an artifact of compute and rule changes, not the QUBO itself.

read the letter

The QUBO-assisted ALNS architecture is a real integration: a stage-level QUBO batch selector inside a logistics-preserving ALNS, with OpenDSS replay validation. The damage-dependence story (mild gain at 90 m/s, stronger effect at 100 m/s) is plausible, and the paper deserves credit for openly reporting the 80 m/s regression, the ~20x runtime, and the CPU-only sampler. The 2.24% SAIDI/ENS improvement at 100 m/s is statistically defensible in the narrow sense — the difference is several standard errors even with n=3 — though the absolute gain is small enough that a utility might not care.

The soft spots are the load-bearing ones. The stress-test note is right: the proposed method adds two non-QUBO mechanisms (adaptive candidate pool, hard full-crew rule) and a much larger inner-loop search budget. The comparison against classical ALNS cannot separate the QUBO formulation from those other changes. The 50.71% makespan reduction at 100 m/s is exactly the kind of number that could shrink or vanish when the classical selector is given the same computation time or the same candidate-pool logic. The zero makespan standard deviation across three stochastic runs in Table III is another red flag; it makes the runs look deterministic or the metric insensitive, and it weakens the headline.

The paper's own limitations section does not fix these confounds. It states the runtime tradeoff but does not attempt a matched-budget baseline. The synthetic damage model parameters are unpublished, but that is minor because they are held fixed across both schedulers. The bigger issue is attribution: the paper frames the QUBO as the reason for the gains, but the evidence only supports "the proposed ensemble of mechanisms" as the reason.

Who is this for? Researchers working on hybrid quantum/classical optimization for power systems, especially those who need an example of how to evaluate such hybrids honestly. It would be a good reading-group paper because it demonstrates both a useful architectural pattern and a textbook attribution error. The right fix is straightforward: add an ablation that removes the QUBO but keeps the adaptive pool and full-crew rule, and run a classical batch selector with equivalent compute (same number of SA reads/sweeps or wall-clock time).

I would send this to peer review, not desk reject it. The core idea is legitimate and the validation pipeline is reproducible, but it needs a major revision to support the central claim. A careful referee could turn this into a solid paper.

Referee Report

3 major / 4 minor

Summary. The paper proposes a QUBO-assisted ALNS scheduler for post-hurricane distribution-system restoration. At each energized stage, a local CPU simulated-annealing QUBO sampler ranks individual repair jobs and multi-job combinations; a deterministic decoder then enforces crew/truck logistics, material trips, and a full-useful-crew rule, and final schedules are replayed in OpenDSS. The method is evaluated on the IEEE 123-node feeder under 80, 90, and 100 m/s hurricane scenarios, with the reported headline result being a 2.24% mean SAIDI/ENS improvement and a 50.71% makespan reduction at 100 m/s relative to classical energized ALNS. The paper argues that QUBO assistance provides increasing value as the repair space becomes more combinatorial.

Significance. The problem is operationally relevant, and the proposed decomposition (QUBO as a local stage-batch selector, deterministic logistics decoder, OpenDSS replay) is a pragmatic architecture that preserves physical feasibility. The independent OpenDSS replay (Eq. 10) is a genuine strength, as is the explicit reporting of simulation settings and parameters. If the central claim were cleanly isolated, this would be a useful contribution showing that a QUBO-based local sampler can improve restoration schedules in severe damage conditions. However, the reported gains are not yet attributable to the QUBO layer because the method differs from the baseline in several non-QUBO ways and consumes far more compute; the statistical basis is also weakened by suspicious zero standard deviations.

major comments (3)

[§IV-B, Table III, with §II-D/Eq. (7)] The headline 50.71% makespan reduction and the 2.24% SAIDI/ENS improvement are confounded with the two non-QUBO changes added in Contributions 2 and 3: the adaptive candidate pool and the hard full-useful-crew rule (Eq. 7). The classical energized ALNS baseline does not include these mechanisms, nor is it given a comparable compute budget (~361 s vs ~16 s per run). A matched-budget ablation with the QUBO layer disabled but the pool/crew rule retained is essential to support the claim that the QUBO formulation, rather than the wider pool, forced crew utilization, or larger inner-loop search, is load-bearing.
[Table III, 90 and 100 m/s rows] The reported zero standard deviation for makespan in both classical and QUBO columns (e.g., 3.4996 ± 0.0000 h for QUBO at 100 m/s) across three stochastic evaluations is implausible for ALNS with stochastic destroy/repair and simulated-annealing sampling. If the runs are not actually independent, the 50.71% makespan and 2.24% SAIDI/ENS figures lack a reliable statistical basis. The authors should report more repetitions, verify run independence, and provide confidence intervals or significance tests.
[Table II and §II-D] The QUBO coefficients (λ_o=12.0, λ_c=0.35, λ_a=0.03, λ_s=0.35), utility weights (wE^q=1.0, wD^q=0.15, wT^q=0.03), the nominal atom bound (28), and the SA reads/sweeps (300/1200) are all hand-set, and no sensitivity analysis is presented. Given that the 80 m/s case shows classical ALNS performing better, the reader cannot judge whether the 100 m/s advantage is robust to reasonable parameter variation. A small sensitivity study or ablation would substantially strengthen the damage-dependent-value claim.

minor comments (4)

[§I (Introduction)] There are minor formatting and typographical issues (e.g., 'sta rt' on the first page). Please perform a careful proofread.
[§IV-A and Fig. 5] The restoration trajectory plot would be more informative with labeled axes and an annotation indicating the reported makespan difference; currently the reader must infer the 7.10 h vs 3.50 h gap from text.
[Eq. (5)] The utility Ub mixes newly reachable load ΔP_b, newly restored customers ΔNb, and completion time τb, but the normalization of Ub is not explicitly defined. Clarify how the normalized utility entering Eq. (6) is computed.
[References] Several references are dated 2026 (e.g., [8], [9], [18]); if these are early-access or conference papers, please ensure the publication status is accurate and complete.

Circularity Check

0 steps flagged

No circularity: the QUBO selector and OpenDSS-evaluated metrics are distinct, and no fitted parameter is renamed as a prediction.

full rationale

The derivation chain is self-contained. The QUBO layer (Eqs. 5-6) is explicitly a local stage-batch selector: U_b is a surrogate utility and H(z) is minimized to propose candidate atoms. These candidates are then passed through the deterministic logistics decoder, which enforces crew/truck constraints, and the final SAIDI, ENS, and makespan are computed from OpenDSS replay acceptance (Eq. 10). The final metrics are therefore not defined by the QUBO objective; they are external evaluations of the committed schedules. No parameter is fitted to the reported outcomes: the local weights (λ_o=12.0, λ_c=0.35, λ_a=0.03, λ_s=0.35; utility weights 1.0/0.15/0.03) are stated as fixed reproducibility settings, not inferred from the test cases. Self-citations [15]-[17] are positioning references and are not load-bearing for the central claim; no uniqueness theorem or prior ansatz is imported from the authors' own work to force the QUBO design. The 50.71% makespan reduction at 100 m/s is an empirical comparison between two schedulers, not an equivalence-by-construction. A wall-clock- or iteration-matched ablation would strengthen the attribution of the gain to the QUBO layer (the classical baseline uses ~16 s while the QUBO method uses ~201-361 s), but that is a confound in experimental design, not a circular derivation, and does not meet the standard for a circularity flag.

Axiom & Free-Parameter Ledger

10 free parameters · 6 axioms · 0 invented entities

No new physical entities (particles, forces, mediators, conserved quantities) are postulated. The 'repair atom', 'energized frontier', and 'synergy' terms are computational abstractions within the scheduler, not entities with independent falsifiable handles.

free parameters (10)

λ_o (QUBO overlap penalty) = 12.0
Hand-set in Table II; penalizes selecting overlapping repair atoms; no sensitivity analysis reported.
λ_c (QUBO capacity reward) = 0.35
Hand-set in Table II; rewards atoms that consume useful crew capacity; drives the full-crew rule.
λ_a (QUBO atom penalty) = 0.03
Hand-set in Table II; mild cost per selected atom.
λ_s (QUBO synergy weight) = 0.35
Hand-set in Table II; rewards pair combinations with positive graph synergy.
QUBO utility weights (wE^q, wD^q, wT^q) = 1.0, 0.15, 0.03
Section III; weights for ENS, SAIDI, and completion time inside each atom's local utility U_b (Eq. 5); chosen by hand, no tuning disclosure.
Global objective weights (wD, wE, wT, εF) = 1, 0.1, 0.001, 0.001
Eq. (4); schedule quality measure; weights chosen by hand and determine what 'better' means.
Fragility failure/materials thresholds = 0.70 / 0.80
Section III; a line is declared failed when p_fail ≥ 0.70 and material trips counted at ≥ 0.80; arbitrary thresholds that set the damage scenario.
Wind field radius R_w = not reported
Eq. (2); defines the spatial extent of the radial exponential wind envelope; value not given in the paper, affecting all damage scenarios.
Per-line fragility parameters (V50,i, β_i) = not published ('persistent line asset table')
Eq. (3); Section III says the values are held fixed but the table is not provided, making the damage realizations unreproducible.
Nominal atom bound / SA reads / sweeps = 28 / 300 / 1200
Section II-F; computational limits of the inner QUBO sampler; paper states the 28-atom bound is non-binding by design.

axioms (6)

domain assumption The internal graph evaluator's estimates of newly reachable load ΔP_b, newly restored customers ΔN_b, and completion time τ_b are faithful surrogates for the OpenDSS-validated outcome.
QUBO utilities (Eq. 5) and all stage decisions are computed from the fast internal evaluator; OpenDSS (Eq. 10) is only a final validator, so search quality depends on the surrogate's accuracy.
domain assumption The synthetic hazard/fragility model (Eqs. 1–3) with thresholds at 0.70/0.80 generates a realistic enough damage field on the IEEE 123-node feeder for the scheduling conclusions to transfer.
Section II-A; the entire comparison is run on one synthetic storm per wind speed, with unpublished V50/β values and unspecified R_w.
domain assumption Keeping every useful crew busy (hard full-useful-crew rule, Eq. 7) is never worse than strategically idling crews to preserve resources for later stages.
Section II-D; acceptance requires assigned crews meet C_t^target, which rules out schedules that deliberately leave crews unassigned while waiting for a bottleneck repair.
domain assumption CPU simulated annealing with 300 reads × 1200 sweeps finds sufficiently low-energy QUBO samples that the stage-batch ranking is meaningful.
Section II-F; the inner sampler is a stochastic heuristic; its sample quality is not measured against exact QUBO solutions.
domain assumption The IEEE 123-node feeder without distributed generation represents real post-hurricane restoration conditions well enough for the damage-dependence trend to generalize.
Section III/IV; one test feeder, no DG, no spatial/seasonal variety; the 'value grows with damage' claim is inferred from three operating points.
standard math The QUBO-to-Ising transformation x_b = (1 + s_b)/2 is exact and the objective is unchanged under it.
Section II-G; elementary change of variables, no hidden assumption.

pith-pipeline@v1.3.0-daily-deepseek · 8229 in / 28391 out tokens · 263795 ms · 2026-08-03T04:55:34.977246+00:00 · methodology

0 comments

read the original abstract

Post hurricane distribution system restoration requires rapid repair scheduling subject to feeder topology, field logistics, and electrical feasibility. This paper presents a quantum inspired quadratic unconstrained binary optimization (QUBO) assisted adaptive large neighborhood search (ALNS) framework. At each restoration stage, a local CPU simulated annealing sampler ranks individual repairs and multi job combinations near the energized frontier. A deterministic decoder preserves crew truck logistics, enforces full useful crew utilization, and rejects infeasible batches. Final schedules are validated through OpenDSS replay. The framework is evaluated on the IEEE 123 node test feeder without distributed generation under 80, 90, and 100 m/s wind scenarios. In the 100 m/s stress test, the proposed method reduces mean system average interruption duration index and energy not supplied by 2.24% and restoration makespan by 50.71% relative to classical energized ALNS. Results show that QUBO assistance is most valuable when severe damage creates a larger combinatorial repair space.

Figures

Figures reproduced from arXiv: 2607.29544 by Anurag Srivastava, Hooman Torkaman, Jignesh Solanki.

**Figure 2.** Figure 2: Stage level QUBO assisted energized restoration scheduling procedure. Low energy sampled batches are passed to the deterministic crew truck decoder, and only physically feasible batches are committed. The stage level procedure is summarized as follows: (1) update the energized graph; (2) generate individual and multi job atoms; (3) expand the candidate boundary when productive crew capacity remains unused;… view at source ↗

**Figure 4.** Figure 4: shows the feeder level line risk map for the 100 m/s case. The spatial distribution of fragility values produces both upstream and downstream repair requirements. In this stress test, damage close to the source creates an initial bottleneck, while distributed downstream failures create a larger parallel scheduling problem after the bottleneck is removed [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗

**Figure 6.** Figure 6: Productive crew utilization in the 100 m/s stress [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗

**Figure 5.** Figure 5: illustrates a representative OpenDSS validated restoration trajectory under the 100 m/s stress test scenario. The classical scheduler restores some customers slightly earlier after the source side bottleneck is cleared, but it leaves a long final restoration tail. The QUBO assisted schedule coordinates the remaining repair batches more effectively and completes restoration in approximately 3.50 h instead o… view at source ↗

discussion (0)

Reference graph

Works this paper leans on

16 extracted references · 2 canonical work pages

[1]

Large scale visual data driven probabilistic risk assessment of utility poles regarding the vulnerability of power distribution infrastructure systems,

J. Kim, M. Kamari, S. Lee, and Y. Ham, “Large scale visual data driven probabilistic risk assessment of utility poles regarding the vulnerability of power distribution infrastructure systems,” J. Constr. Eng. Manage., vol. 147, no. 10, Art. no. 04021121, 2021, doi: 10.1061/(ASCE)CO.1943 7862.0002153

arXiv 2021
[2]

Repair and resource scheduling in unbalanced distribution systems using neighborhood search,

A. Arif, Z. Wang, C. Chen, and J. Wang, “Repair and resource scheduling in unbalanced distribution systems using neighborhood search,” IEEE Trans. Smart Grid, vol. 11, no. 1, pp. 673–685, 2020, doi: 10.1109/TSG.2019.2927739

arXiv 2020
[3]

An adaptive large neighborhood search heuristic for the pickup and delivery problem with time windows,

S. Ropke and D. Pisinger, “An adaptive large neighborhood search heuristic for the pickup and delivery problem with time windows,” Transp. Sci., vol. 40, no. 4, pp. 455–472, 2006, doi: 10.1287/trsc.1050.0135

arXiv 2006
[4]

The unconstrained binary quadratic programming problem: A survey,

G. Kochenberger, J. K. Hao, F. Glover, M. Lewis, Z. Lu, H. Wang, and Y. Wang, “The unconstrained binary quadratic programming problem: A survey,” J. Comb. Optim., vol. 28, no. 1, pp. 58–81, 2014, doi: 10.1007/s10878 014 9734 0

work page doi:10.1007/s10878 2014
[5]

Ising formulations of many NP problems,

A. Lucas, “Ising formulations of many NP problems,” Front. Phys., vol. 2, Art. no. 5, 2014, doi: 10.3389/fphy.2014.00005

arXiv 2014
[6]

Optimization by simulated annealing,

S. Kirkpatrick, C. D. Gelatt, and M. P. Vecchi, “Optimization by simulated annealing,” Science, vol. 220, no. 4598, pp. 671–680, 1983, doi: 10.1126/science.220.4598.671

doi:10.1126/science.220.4598.671 1983
[7]

Quantum annealing infused microgrids formation: Distribution system restoration and resilience enhancement,

N. Nikmehr, P. Zhang, H. Zheng, T. C. Wei, G. He, and Y. A. Shamash, “Quantum annealing infused microgrids formation: Distribution system restoration and resilience enhancement,” IEEE Trans. Power Syst., vol. 40, no. 1, pp. 463–474, 2025, doi: 10.1109/TPWRS.2024.3399122

arXiv 2025
[9]

Learning optimal crew dispatch for grid restoration following an earthquake,

F. Amani, F. Ardali, and A. Kargarian, “Learning optimal crew dispatch for grid restoration following an earthquake,” IEEE Trans. Smart Grid, early access, 2026, doi: 10.1109/TSG.2026.3700512

arXiv 2026
[10]

Fault classification and location in power distribution networks using 1D CNN with residual learning,

K. D. Khattak, M. A. Choudhry, and A. Feliachi, “Fault classification and location in power distribution networks using 1D CNN with residual learning,” in Proc. 2024 IEEE Power & Energy Society Innovative Smart Grid Technologies Conf. (ISGT), 2024, pp. 1–5, doi: 10.1109/ISGT59692.2024.10454243

arXiv 2024
[11]

Radial distribution test feeders,

W. H. Kersting, “Radial distribution test feeders,” IEEE Trans. Power Syst., vol. 6, no. 3, pp. 975–985, 1991, doi: 10.1109/59.119237

work page doi:10.1109/59.119237 1991
[12]

An analytic model of the wind and pressure profiles in hurricanes,

G. J. Holland, “An analytic model of the wind and pressure profiles in hurricanes,” Mon. Weather Rev., vol. 108, no. 8, pp. 1212–1218, 1980, doi: 10.1175/1520 0493(1980)108<1212:AAMOTW>2.0.CO;2

doi:10.1175/1520 1980
[13]

IEEE Guide for Electric Power Distribution Reliability Indices, IEEE Std 1366 2022, 2022

2022
[14]

An open source platform for collaborating on smart grid research,

R. C. Dugan and T. E. McDermott, “An open source platform for collaborating on smart grid research,” in Proc. IEEE PES Gen. Meeting, 2011, pp. 1–7, doi: 10.1109/PES.2011.6039829

arXiv 2011
[16]

Comparative multi objective investigation of radial and ring distribution system in the presence of DGs,

N. Deyhimi, H. Torkaman, M. Shadaei, M. Shabanirad, and M. Kermani, “Comparative multi objective investigation of radial and ring distribution system in the presence of DGs,” in Proc. IEEE EEEIC/I&CPS Europe, Bari, Italy, 2021, pp. 1–6, doi: 10.1109/EEEIC/ICPSEurope51590.2021.9584527

arXiv 2021
[17]

Protection coordination enhancement in tied uncompleted parallel lines via two step Grey Wolf Optimization and multi strategy machine learning for jumper placement,

H. Torkaman and J. Solanki, “Protection coordination enhancement in tied uncompleted parallel lines via two step Grey Wolf Optimization and multi strategy machine learning for jumper placement,” TechRxiv preprint, Sep. 2025, doi: 10.36227/techrxiv.175761364.49143203/v1

arXiv 2025
[18]

A distributed variational quantum eigensolver algorithm for unit commitment,

M. Hasanzadeh, A. Rajabi, and A. Kargarian, “A distributed variational quantum eigensolver algorithm for unit commitment,” in Proc. 2026 IEEE Texas Power and Energy Conf. (TPEC), 2026, pp. 1–6, doi: 10.1109/TPEC67884.2026.11513141

arXiv 2026

[1] [1]

Large scale visual data driven probabilistic risk assessment of utility poles regarding the vulnerability of power distribution infrastructure systems,

J. Kim, M. Kamari, S. Lee, and Y. Ham, “Large scale visual data driven probabilistic risk assessment of utility poles regarding the vulnerability of power distribution infrastructure systems,” J. Constr. Eng. Manage., vol. 147, no. 10, Art. no. 04021121, 2021, doi: 10.1061/(ASCE)CO.1943 7862.0002153

arXiv 2021

[2] [2]

Repair and resource scheduling in unbalanced distribution systems using neighborhood search,

A. Arif, Z. Wang, C. Chen, and J. Wang, “Repair and resource scheduling in unbalanced distribution systems using neighborhood search,” IEEE Trans. Smart Grid, vol. 11, no. 1, pp. 673–685, 2020, doi: 10.1109/TSG.2019.2927739

arXiv 2020

[3] [3]

An adaptive large neighborhood search heuristic for the pickup and delivery problem with time windows,

S. Ropke and D. Pisinger, “An adaptive large neighborhood search heuristic for the pickup and delivery problem with time windows,” Transp. Sci., vol. 40, no. 4, pp. 455–472, 2006, doi: 10.1287/trsc.1050.0135

arXiv 2006

[4] [4]

The unconstrained binary quadratic programming problem: A survey,

G. Kochenberger, J. K. Hao, F. Glover, M. Lewis, Z. Lu, H. Wang, and Y. Wang, “The unconstrained binary quadratic programming problem: A survey,” J. Comb. Optim., vol. 28, no. 1, pp. 58–81, 2014, doi: 10.1007/s10878 014 9734 0

work page doi:10.1007/s10878 2014

[5] [5]

Ising formulations of many NP problems,

A. Lucas, “Ising formulations of many NP problems,” Front. Phys., vol. 2, Art. no. 5, 2014, doi: 10.3389/fphy.2014.00005

arXiv 2014

[6] [6]

Optimization by simulated annealing,

S. Kirkpatrick, C. D. Gelatt, and M. P. Vecchi, “Optimization by simulated annealing,” Science, vol. 220, no. 4598, pp. 671–680, 1983, doi: 10.1126/science.220.4598.671

doi:10.1126/science.220.4598.671 1983

[7] [7]

Quantum annealing infused microgrids formation: Distribution system restoration and resilience enhancement,

N. Nikmehr, P. Zhang, H. Zheng, T. C. Wei, G. He, and Y. A. Shamash, “Quantum annealing infused microgrids formation: Distribution system restoration and resilience enhancement,” IEEE Trans. Power Syst., vol. 40, no. 1, pp. 463–474, 2025, doi: 10.1109/TPWRS.2024.3399122

arXiv 2025

[8] [9]

Learning optimal crew dispatch for grid restoration following an earthquake,

F. Amani, F. Ardali, and A. Kargarian, “Learning optimal crew dispatch for grid restoration following an earthquake,” IEEE Trans. Smart Grid, early access, 2026, doi: 10.1109/TSG.2026.3700512

arXiv 2026

[9] [10]

Fault classification and location in power distribution networks using 1D CNN with residual learning,

K. D. Khattak, M. A. Choudhry, and A. Feliachi, “Fault classification and location in power distribution networks using 1D CNN with residual learning,” in Proc. 2024 IEEE Power & Energy Society Innovative Smart Grid Technologies Conf. (ISGT), 2024, pp. 1–5, doi: 10.1109/ISGT59692.2024.10454243

arXiv 2024

[10] [11]

Radial distribution test feeders,

W. H. Kersting, “Radial distribution test feeders,” IEEE Trans. Power Syst., vol. 6, no. 3, pp. 975–985, 1991, doi: 10.1109/59.119237

work page doi:10.1109/59.119237 1991

[11] [12]

An analytic model of the wind and pressure profiles in hurricanes,

G. J. Holland, “An analytic model of the wind and pressure profiles in hurricanes,” Mon. Weather Rev., vol. 108, no. 8, pp. 1212–1218, 1980, doi: 10.1175/1520 0493(1980)108<1212:AAMOTW>2.0.CO;2

doi:10.1175/1520 1980

[12] [13]

IEEE Guide for Electric Power Distribution Reliability Indices, IEEE Std 1366 2022, 2022

2022

[13] [14]

An open source platform for collaborating on smart grid research,

R. C. Dugan and T. E. McDermott, “An open source platform for collaborating on smart grid research,” in Proc. IEEE PES Gen. Meeting, 2011, pp. 1–7, doi: 10.1109/PES.2011.6039829

arXiv 2011

[14] [16]

Comparative multi objective investigation of radial and ring distribution system in the presence of DGs,

N. Deyhimi, H. Torkaman, M. Shadaei, M. Shabanirad, and M. Kermani, “Comparative multi objective investigation of radial and ring distribution system in the presence of DGs,” in Proc. IEEE EEEIC/I&CPS Europe, Bari, Italy, 2021, pp. 1–6, doi: 10.1109/EEEIC/ICPSEurope51590.2021.9584527

arXiv 2021

[15] [17]

Protection coordination enhancement in tied uncompleted parallel lines via two step Grey Wolf Optimization and multi strategy machine learning for jumper placement,

H. Torkaman and J. Solanki, “Protection coordination enhancement in tied uncompleted parallel lines via two step Grey Wolf Optimization and multi strategy machine learning for jumper placement,” TechRxiv preprint, Sep. 2025, doi: 10.36227/techrxiv.175761364.49143203/v1

arXiv 2025

[16] [18]

A distributed variational quantum eigensolver algorithm for unit commitment,

M. Hasanzadeh, A. Rajabi, and A. Kargarian, “A distributed variational quantum eigensolver algorithm for unit commitment,” in Proc. 2026 IEEE Texas Power and Energy Conf. (TPEC), 2026, pp. 1–6, doi: 10.1109/TPEC67884.2026.11513141

arXiv 2026