Pith. sign in

REVIEW 3 major objections 7 minor 1 cited by

PGLearn -- An Open-Source Learning Toolkit for Optimal Power Flow

T0 review · 3 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read PGLearn claims to be the first public collection of over ten million optimal power flow samples with complete primal and dual solutions for AC, SOC, and DC formulations, including realistic time-series data for large-scale grids and…

desk verdict PGLearn is a strong, standard-setting OPF dataset toolkit whose clean accept is held back only by an overstated AC-dual claim that is easy to fix. read the letter →

arxiv 2505.22825 v1 pith:DFOOIE2Z submitted 2025-05-28 cs.LG cs.AIcs.SYeess.SYmath.OC

Michael Klamkin , Mathieu Tanneau , Pascal Van Hentenryck This is my paper
classification cs.LGcs.AIcs.SYeess.SYmath.OC
keywords optimalpowerflowmachinelearningdatasetsAC-OPFSOC-OPFDC-OPFLagrangiandualitydataaugmentationgridbenchmarking
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

PGLearn sets out to end the fragmentation of machine-learning-for-OPF research by releasing one public, reproducible collection of standardized datasets. The paper's central claim is that this collection is the first to combine realistic time-series data for large-scale power systems with complete primal and dual solutions for AC, SOC, and DC OPF formulations, totalling more than ten million solved samples. A sympathetic reader would care because it gives any research group the same training and test splits, the same ground-truth solutions, and the same evaluation metrics, removing data generation as a variable when comparing methods. The paper also contributes open-source code for data generation and model benchmarking, so the corpus can be reproduced and extended.

What carries the argument

The load-bearing object is Algorithm 1, the demand-sampling scheme: a global per-sample scaling factor $b \sim \mathrm{Uniform}(b_l, b_u)$ multiplies the reference load, and each load then receives independent active and reactive noise factors drawn uniformly from $[1-\epsilon, 1+\epsilon]$. That two-level structure is what lets the datasets cover both global operating regimes and local variability. The complementary machinery is the solver pipeline that records dual information: for SOC-OPF and DC-OPF the stored duals are dual-feasible conic or linear solutions that give valid certificates, while for AC-OPF the paper stores the multipliers returned by a local nonlinear solver. The pipeline is wrapped in the paper's own open-source generation code and a training and evaluation toolkit, with HDF5 data files and a case JSON defining all grid parameters.

What would settle it

Take any stored AC-OPF sample, form the Lagrangian with its saved multipliers, and minimize that Lagrangian over the bounded variable box with a global method; if the result is not a lower bound on the saved primal objective for even one sample, the claim that PGLearn provides complete AC-OPF dual solutions fails. A simpler version of the same test is to check whether the stored dual objective values are below the primal objective values across the test set.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is a dataset-generation pipeline together with the released corpus it produced. Each instance is created by taking a reference grid snapshot and sampling every load's active and reactive demand with a per-sample global scaling factor plus independent per-load noise; the global factor is drawn uniformly over a case-specific range and the local noise is ±20%, so datasets cover wide total-demand ranges while retaining diversity. N-1 instances are generated by disabling one generator or one non-bridge branch, and for the two largest cases one year of hourly synthetic time series is interpolated with cubic splines to five- or ten-minute granularity. Every feasible sample is solved for AC, SOC, and DC formulations, and the paper stores primal solutions, dual solution objects, and metadata such as solve times in HDF5 format, with an 80/20 seeded split into training and testing sets. PGLearn claims this is the first OPF dataset collection to include both time-series data for large-scale systems and complete primal and dual solutions for multiple formulations.

Load-bearing premise

The load-bearing premise is that storing the multiplier values a local solver returns for AC-OPF counts as a complete dual solution, even though the paper's own appendix says valid dual bounds for that nonconvex problem would require solving an NP-hard problem.

Editorial extensions

If this is right

  • Any two groups can train on the same PGLearn training split and report against the same held-out test split, making published accuracy and timing numbers directly comparable.
  • Researchers can train dual proxies or learn Lagrange multipliers because dual solutions are stored for every formulation, with SOC and DC duals giving valid certificates.
  • The time-series cases (Texas7k and Midwest24k) allow models to be tested on chronologically coherent demand profiles rather than only independent random perturbations.
  • Because the generation code is open source, the dataset can be reproduced, extended to new formulations or sampling schemes, and re-generated for new grids without re-inventing the pipeline.
  • The evaluation-metrics guidance (optimality gap, constraint violations, distance to feasible set, timing) gives the community a common reporting standard.

Reading between the lines

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

  • The stored AC-OPF dual objects are local multipliers, not valid Lagrangian certificates: the paper's own appendix notes that valid dual bounds for the nonconvex problem would require solving an NP-hard problem, so those duals should be treated as local sensitivity information unless separately validated.
  • A quick validation pass could check whether every stored AC dual objective lies below its primal objective; any violation would confirm that the AC duals are not a certificate, while SOC and DC duals should pass by weak duality.
  • The global-plus-local sampling scheme could be extended to spatially correlated renewable availability (for example, wind and solar zones) to generate scenarios that stress the grid in a more weather-realistic way.
  • If PGLearn becomes the common benchmark, reported performance gains in the ML-for-OPF literature may shift from data-generation choices to model architectures and training procedures.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 7 minor

Summary. PGLearn is a dataset and tooling contribution to the ML-for-OPF literature. The paper introduces a collection of OPF instances built from 14 reference snapshots (up to the Midwest24k case with 23,643 buses) by sampling demand via a global multiplier plus independent per-load noise, generating N-1-style line/generator outages, and, for Texas7k and Midwest24k, interpolating synthetic time series to five- and ten-minute resolution. Each sample is solved in AC, SOC, and DC formulations using Ipopt, Clarabel, and HiGHS, respectively, with primal and dual information stored in HDF5. The paper also describes PGLearn.jl (data-generation code) and ML4OPF (PyTorch parsers, differentiable layers, and baseline models), and proposes evaluation metrics for optimality gap, constraint violations, feasibility distance, and computational performance. The headline claims are that this is the first dataset collection to include realistic time-series data for large-scale systems and to provide complete primal and dual solutions for AC, SOC, and DC OPF.

Significance. The resource is potentially very valuable. The open-source release of dataset generation code, seeded train/test splits, standardized HDF5 formats, and a PyTorch toolkit lowers entry barriers and enables fair comparisons. The scale (over 10M samples, cases up to 23,643 buses) and multiple formulations are genuine strengths. The SOC and DC dual solutions provide valid certificates, which supports the stated motivation for dual-proxy learning. However, the two headline claims need qualification: the AC-OPF 'dual solutions' are local KKT multipliers rather than certifying duals, and the 'realistic sampling' claim is stronger than the actual uniform-perturbation scheme. These are fixable with rewording and documentation changes, and do not undermine the dataset's usefulness as a benchmark.

major comments (3)
  1. [Section 2.2, Appendix A.1.1, Table 4] Section 2.2 describes the stored AC-OPF duals as part of 'complete primal and dual solutions,' but Appendix A.1.1 states that a valid Lagrangian lower bound for the nonconvex problem requires solving the NP-hard inner minimization in Eq. (3). The AC-OPF duals in Table 4 are Ipopt KKT multipliers for a local optimum; they are not dual-feasible certificates and the dual_objective_value field in Table 3 is not a valid lower bound. This contradicts the unqualified claim in the abstract and Section 1.3. Please reword to 'complete primal solutions and local KKT multipliers for AC, plus valid dual certificates for SOC and DC,' and add a prominent caveat in Section 2.2 and in the HDF5 metadata documentation.
  2. [Abstract, Algorithm 1, Section 6.1] The paper repeatedly describes the sampling scheme as 'realistic' and 'representative of real-life operating conditions' (Abstract, Section 1.3). What Algorithm 1 implements is a single global multiplier b, uniform on [bl,bu], times independent per-load uniform multipliers on [1-epsilon, 1+epsilon], with epsilon=20%. This is a tractable parametric model, but it does not model spatial/temporal load correlations beyond a single scalar, and the time-series are synthetic as acknowledged in Section 6.1. To make the claim supportable, either temper the wording to 'a parametric approximation inspired by day-ahead forecasting ranges' or provide validation of the generated demand distributions against real system data (e.g., distribution of total demand, coincidence factors). This is a central contribution claim and needs to be supported or qualified.
  3. [Section 1.3 vs Appendix A.2.1] The statement in Section 1.3 that 'The code used to generate PGLearn is fully open-source and relies only on open-source solvers' is contradicted by the default AC-OPF configuration in Appendix A.2.1, which uses Ipopt with the MA27 linear solver from LibHSL. LibHSL is not an open-source library. Since reproducible generation is a central claim, the paper must either change the default to a genuinely open-source linear solver (e.g., MUMPS) or clearly state that LibHSL/MA27 is a non-free dependency and specify how the published datasets were generated.
minor comments (7)
  1. [Section 1.1 heading] 'Data Scarsity' should be 'Data Scarcity'.
  2. [Section 1.1, second paragraph] 'the resulting data distribution' should be 'of the resulting data distribution'.
  3. [Section 4.1] 'GNUparallel' should be 'GNU Parallel'.
  4. [Section 3, Train-Test Split] The 'infeasible' split is defined as samples for which a locally optimal solution could not be found, not as proven infeasible; rename it 'unsolved' or add a note that the termination_status field in Table 3 must be used to distinguish solver failure from infeasibility.
  5. [Appendix A.2.2] 'slighly' should be 'slightly'.
  6. [References] Jabr [2006a] and [2006b] are the same paper; combine or cross-reference them.
  7. [Table 1] Please define 'Total PG' precisely (it appears to be the sum of generator upper bounds) and state whether the 'Global Range' interval applies to both active and reactive demand.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: PGLearn's dataset construction is a forward pipeline from independent reference cases and open-source solvers, not a derivation whose outputs are assumed in its inputs.

full rationale

PGLearn is a dataset and tooling paper rather than a parametric model of its own outputs. The data pipeline starts from externally maintained reference cases (PGLib-OPF, RTE/PEGASE, Texas A&M, ARPA-E PERFORM) and public solvers (Ipopt, Clarabel, HiGHS), then applies an explicitly stated sampling scheme (Algorithm 1) with free design parameters (global range, epsilon=20%). These choices are not fitted to any target claim, so there is no fitted-input-called-prediction or self-definitional reduction: the resulting AC/SOC/DC primal-dual data are forward artifacts. The paper's self-citations (e.g., Tanneau and Van Hentenryck [2024] for dual feasibility, Qiu et al. [2024] for dual proxies) are motivational or background support, not load-bearing uniqueness theorems that force the paper's conclusions, and the claims about time-series novelty and complete datasets are benchmarked against external prior work (OPFLearn, OPFData). The legitimate caveat in Appendix A.1.1 -- that valid Lagrangian lower bounds for nonconvex AC-OPF require solving an NP-hard inner problem, so Ipopt multipliers are not certified dual bounds -- is a correctness/validity limitation that the paper itself states; it does not make the dataset construction circular. Thus no specific circular step can be quoted, and the appropriate finding is no significant circularity.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central artifact is a dataset, so there are no invented physical entities. The main assumptions are the realism of the synthetic sampling procedure and the meaningfulness of AC-OPF duals, both of which are only weakly supported.

free parameters (2)
  • Local noise level epsilon = 20% for all cases
    In Algorithm 1, each load is multiplied by Uniform(1-epsilon, 1+epsilon). The value is hand-chosen, not derived or validated against real load data.
  • Global demand scaling range [bl, bu] = Varies per case, e.g., 0.6-1.0, 0.7-1.1, 0.8-1.2, 0.9-1.3
    The upper bound bu is found by incrementally scaling the reference load in 10% steps until the OPF becomes infeasible, then by fixing the range width at 40%. This is an adaptive data-generation parameter that depends on each grid's feasibility boundary.
assumptions (4)
  • domain assumption The reference snapshots (PGLib, PEGASE, RTE, Texas A&M cases) are realistic starting points for generating useful OPF datasets.
    Section 3 relies on these public test cases as ground truth for the grid topology and parameters, assuming they are representative of real systems.
  • ad hoc to paper Global scaling plus independent per-load noise yields a distribution that is representative of real operating conditions.
    Algorithm 1 and Section 3 claim realism, but this is an unvalidated modeling choice. The paper's own Section 6.1 notes the absence of public granular data and the limitation of synthetic time series.
  • domain assumption For AC-OPF, local solver (Ipopt) solutions are good enough proxies for global optima, and the associated KKT multipliers can be called 'dual solutions'.
    Section 2.2 and Appendix A.2.1 assume that Ipopt outputs are usable as ground truth. Appendix A.1.1 acknowledges that valid Lagrangian bounds for nonconvex problems are NP-hard, but the paper does not reconcile this with the release of AC-OPF duals.
  • standard math Conic duality holds for SOC-OPF and DC-OPF, so dual-feasible solutions provide valid certificates.
    The formulations in Models 2-5 follow standard conic duality, which is correct for convex problems.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PGLearn -- An Open-Source Learning Toolkit for Optimal Power Flow." pith.science (2026). https://pith.science/paper/DFOOIE2Z

@misc{pith2026250522825,
  author       = {Pith},
  title        = {Pith review of: PGLearn -- An Open-Source Learning Toolkit for Optimal Power Flow},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DFOOIE2Z}},
  note         = {Machine review of arXiv:2505.22825}
}
read the original abstract

Machine Learning (ML) techniques for Optimal Power Flow (OPF) problems have recently garnered significant attention, reflecting a broader trend of leveraging ML to approximate and/or accelerate the resolution of complex optimization problems. These developments are necessitated by the increased volatility and scale in energy production for modern and future grids. However, progress in ML for OPF is hindered by the lack of standardized datasets and evaluation metrics, from generating and solving OPF instances, to training and benchmarking machine learning models. To address this challenge, this paper introduces PGLearn, a comprehensive suite of standardized datasets and evaluation tools for ML and OPF. PGLearn provides datasets that are representative of real-life operating conditions, by explicitly capturing both global and local variability in the data generation, and by, for the first time, including time series data for several large-scale systems. In addition, it supports multiple OPF formulations, including AC, DC, and second-order cone formulations. Standardized datasets are made publicly available to democratize access to this field, reduce the burden of data generation, and enable the fair comparison of various methodologies. PGLearn also includes a robust toolkit for training, evaluating, and benchmarking machine learning models for OPF, with the goal of standardizing performance evaluation across the field. By promoting open, standardized datasets and evaluation metrics, PGLearn aims at democratizing and accelerating research and innovation in machine learning applications for optimal power flow problems. Datasets are available for download at https://www.huggingface.co/PGLearn.

Figures

Figures reproduced from arXiv: 2505.22825 by the authors.

Figure 1
Figure 1. The PGLearn Toolkit: publicly available AC, DC, and SOC optimal power flow datasets, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Limitations of sampling strategies that do not consider correlations across individual loads. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Constrained Diffusion Models for Synthesizing Representative Power Flow Datasets

    cs.LG 2025-06 conditional novelty 6.0 of 10

    A physics-guided diffusion model produces synthetic power flow samples that are more AC-feasible and slightly closer to ground truth than unconstrained diffusion on IEEE 5, 24, and 118 bus systems.

Reference graph

Works this paper leans on

56 extracted references · 38 canonical work pages · cited by 1 Pith paper

  1. [1]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should not add it explicitly Type <Return> for now, but then later remove the command n...

  2. [2]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@first@sw \@firstoftwo \@ifundefined NAT@b*@#2 \@firstoftwo @num @NAT@ctr \@secondoft...

  3. [3]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibsetup #1 @NAT@ctr @ @openbib .11em \@plus.33em \@minus.07em 4000 4000 `\.\@m @bibit...

    work page

  4. [4]

    Future of grid operations and markets: Uncertainty management., 2021

    Junshan Zhang, Yonghong Chen, Mohammad Faqiry, Bernard Knueven, Manuel Joseph Garcia, and Roger Treinen. Future of grid operations and markets: Uncertainty management., 2021. URL https://www.osti.gov/biblio/1861473

    work page arXiv 2021

  5. [5]

    Real-time risk analysis with optimization proxies

    Wenbo Chen, Mathieu Tanneau, and Pascal Van Hentenryck. Real-time risk analysis with optimization proxies. Electric Power Systems Research, 235: 0 110822, 2024

    work page 2024

  6. [6]

    Review of machine learning techniques for optimal power flow

    Hooman Khaloie, Mihaly Dolanyi, Jean-Francois Toubeau, and Fran c ois Vall \'e e. Review of machine learning techniques for optimal power flow. Available at SSRN 4681955, May 2024. doi:10.2139/ssrn.4681955

    work page doi:10.2139/ssrn.4681955 2024

  7. [7]

    OPFData : Large-scale datasets for AC optimal power flow with topological perturbations

    Sean Lovett, Miha Zgubic, Sofia Liguori, Sephora Madjiheurem, Hamish Tomlinson, Sophie Elster, Chris Apps, Sims Witherspoon, and Luis Piloto. OPFData : Large-scale datasets for AC optimal power flow with topological perturbations. arXiv preprint arXiv:2406.07234, 2024

    arXiv 2024

  8. [8]

    Luh, Lei Wu, Bing Yan, Mikhail A

    Yonghong Chen, Feng Pan, Feng Qiu, Alinson S Xavier, Tongxin Zheng, Muhammad Marwali, Bernard Knueven, Yongpei Guan, Peter B. Luh, Lei Wu, Bing Yan, Mikhail A. Bragin, Haiwang Zhong, Anthony Giacomoni, Ross Baldick, Boris Gisin, Qun Gu, Russ Philbrick, and Fangxing Li. Security-constrained unit commitment for electricity market: Modeling, solution methods...

    work page arXiv 2023

Show all 56 references
  1. [9]

    A survey on applications of machine learning for optimal power flow

    Fouad Hasan, Amin Kargarian, and Ali Mohammadi. A survey on applications of machine learning for optimal power flow. In 2020 IEEE Texas Power and Energy Conference (TPEC), pages 1--6. IEEE, 2020

    2020

  2. [10]

    Molzahn and Ian A

    Daniel K. Molzahn and Ian A. Hiskens. A survey of relaxations and approximations of the power flow equations. Foundations and Trends® in Electric Energy Systems, 4 0 (1-2): 0 1--221, 2019. ISSN 2332-6557. doi:10.1561/3100000012. URL http://dx.doi.org/10.1561/3100000012

    2019 doi

  3. [11]

    The security-constrained commitment and dispatch for midwest iso day-ahead co-optimized energy and ancillary service market

    Xingwang Ma, Haili Song, Mingguo Hong, Jie Wan, Yonghong Chen, and Eugene Zak. The security-constrained commitment and dispatch for midwest iso day-ahead co-optimized energy and ancillary service market. In 2009 IEEE Power & Energy Society General Meeting, pages 1--8. IEEE, 2009

    2009

  4. [12]

    End-to-end feasible optimization proxies for large-scale economic dispatch

    Wenbo Chen. End-to-end feasible optimization proxies for large-scale economic dispatch. INFORMS Annual meeting , 2023

    2023

  5. [13]

    The Power Grid Library for benchmarking AC optimal power flow algorithms

    Sogol Babaeinejadsarookolaee, Adam Birchfield, Richard D Christie, Carleton Coffrin, Christopher DeMarco, Ruisheng Diao, Michael Ferris, Stephane Fliscounakis, Scott Greene, Renke Huang, et al. The Power Grid Library for benchmarking AC optimal power flow algorithms. arXiv pre...

    1908 arXiv

  6. [14]

    Neural networks for power flow: Graph neural solver

    Balthazar Donon, R \'e my Cl \'e ment, Benjamin Donnot, Antoine Marot, Isabelle Guyon, and Marc Schoenauer. Neural networks for power flow: Graph neural solver. Electric Power Systems Research, 189: 0 106547, 2020

    2020

  7. [15]

    Data-driven time series reconstruction for modern power systems research

    Minas Chatzos, Mathieu Tanneau, and Pascal Van Hentenryck. Data-driven time series reconstruction for modern power systems research. Electric Power Systems Research, 212: 0 108589, 2022

    2022

  8. [16]

    Bucketized active sampling for learning ACOPF

    Michael Klamkin, Mathieu Tanneau, Terrence WK Mak, and Pascal Van Hentenryck. Bucketized active sampling for learning ACOPF . Electric Power Systems Research, 235: 0 110697, 2024

    2024

  9. [17]

    Learning an optimally reduced formulation of OPF through meta-optimization

    Alex Robson, Mahdi Jamei, Cozmin Ududec, and Letif Mones. Learning an optimally reduced formulation of OPF through meta-optimization. arXiv preprint arXiv:1911.06784, 2019

    1911 arXiv

  10. [18]

    OPF-Learn : An open-source framework for creating representative AC optimal power flow datasets

    Trager Joswig-Jones, Kyri Baker, and Ahmed S Zamzam. OPF-Learn : An open-source framework for creating representative AC optimal power flow datasets. In 2022 IEEE Power & Energy Society Innovative Smart Grid Technologies Conference (ISGT), pages 1--5. IEEE, 2022

    2022

  11. [19]

    Powermodels.jl: An open-source framework for exploring power flow formulations

    Carleton Coffrin, Russell Bent, Kaarthik Sundar, Yeesian Ng, and Miles Lubin. Powermodels.jl: An open-source framework for exploring power flow formulations. In 2018 Power Systems Computation Conference (PSCC), pages 1--8, June 2018. doi:10.23919/PSCC.2018.8442948

    2018

  12. [20]

    Julia: A fresh approach to numerical computing

    Jeff Bezanson, Alan Edelman, Stefan Karpinski, and Viral B Shah. Julia: A fresh approach to numerical computing. SIAM R eview , 59 0 (1): 0 65--98, 2017. doi:10.1137/141000671. URL https://epubs.siam.org/doi/10.1137/141000671

    2017 doi

  13. [21]

    PyTorch 2: Faster machine learning through dynamic Python bytecode transformation and graph compilation

    Jason Ansel, Edward Yang, Horace He, Natalia Gimelshein, Animesh Jain, Michael Voznesensky, Bin Bao, Peter Bell, David Berard, Evgeni Burovski, Geeta Chauhan, Anjali Chourdia, Will Constable, Alban Desmaison, Zachary DeVito, Elias Ellison, Will Feng, Jiong Gong, Michael Gschwi...

    2024

  14. [22]

    Radial distribution load flow using conic programming

    Rabih A Jabr. Radial distribution load flow using conic programming. IEEE transactions on power systems, 21 0 (3): 0 1458--1459, 2006 a

    2006

  15. [23]

    Transmission management in the deregulated environment

    Richard D Christie, Bruce F Wollenberg, and Ivar Wangensteen. Transmission management in the deregulated environment. Proceedings of the IEEE, 88 0 (2): 0 170--195, 2000

    2000

  16. [24]

    Dual conic proxies for ac optimal power flow

    Guancheng Qiu, Mathieu Tanneau, and Pascal Van Hentenryck. Dual conic proxies for ac optimal power flow. Electric Power Systems Research, 236: 0 110661, 2024

    2024

  17. [25]

    Dual lagrangian learning for conic optimization

    Mathieu Tanneau and Pascal Van Hentenryck. Dual lagrangian learning for conic optimization. arXiv preprint arXiv:2402.03086, 2024

    2024 arXiv

  18. [26]

    Learning constrained optimization with deep augmented lagrangian methods

    James Kotary and Ferdinando Fioretto. Learning constrained optimization with deep augmented lagrangian methods. arXiv preprint arXiv:2403.03454, 2024

    2024 arXiv

  19. [27]

    Learning lagrangian multipliers for the travelling salesman problem, 2023

    Augustin Parjadis, Quentin Cappart, Bistra Dilkina, Aaron Ferber, and Louis-Martin Rousseau. Learning lagrangian multipliers for the travelling salesman problem, 2023. URL https://arxiv.org/abs/2312.14836

    2023 arXiv

  20. [28]

    Predicting lagrangian multipliers for mixed integer linear programs

    Francesco Demelas, Joseph Le Roux, Mathieu Lacroix, and Axel Parmentier. Predicting lagrangian multipliers for mixed integer linear programs. In Ruslan Salakhutdinov, Zico Kolter, Katherine Heller, Adrian Weller, Nuria Oliver, Jonathan Scarlett, and Felix Berkenkamp, editors, ...

    2024

  21. [29]

    Contingency ranking with respect to overloads in very large power systems taking into account uncertainty, preventive, and corrective actions

    St \'e phane Fliscounakis, Patrick Panciatici, Florin Capitanescu, and Louis Wehenkel. Contingency ranking with respect to overloads in very large power systems taking into account uncertainty, preventive, and corrective actions. IEEE Transactions on Power Systems, 28 0 (4): 0...

    2013

  22. [30]

    AC power flow data in MATPOWER and QCQP format: iTesla , RTE snapshots, and PEGASE

    C \'e dric Josz, St \'e phane Fliscounakis, Jean Maeght, and Patrick Panciatici. AC power flow data in MATPOWER and QCQP format: iTesla , RTE snapshots, and PEGASE . arXiv preprint arXiv:1603.01533, 2016

    2016 arXiv

  23. [31]

    University of Wisconsin-Madison ARPA-E PERFORM Electric Grid Test Cases , 2024

    University of Wisconsin-Madison . University of Wisconsin-Madison ARPA-E PERFORM Electric Grid Test Cases , 2024. URL https://electricgrids.engr.tamu.edu/university-of-wisconsin-madison-perform-cases/

    2024

  24. [32]

    A description of the Texas A&M University Electric Grid Test Case Repository for power system studies

    Sanjana Kunkolienkar, Farnaz Safdarian, Jonathan Snodgrass, Adam Birchfield, and Thomas Overbye. A description of the Texas A&M University Electric Grid Test Case Repository for power system studies. In 2024 IEEE Texas Power and Energy Conference (TPEC), pages 1--6. IEEE, 2024

    2024

  25. [33]

    of Electrical Engineering

    University of Washington, Dept. of Electrical Engineering . Power systems test case archive, 1999. URL http://www.ee.washington.edu/research/pstca/

    1999

  26. [34]

    Learning optimization proxies for large-scale security-constrained economic dispatch

    Wenbo Chen, Seonho Park, Mathieu Tanneau, and Pascal Van Hentenryck. Learning optimization proxies for large-scale security-constrained economic dispatch. Electric Power Systems Research, 213: 0 108566, 2022

    2022

  27. [35]

    The entso-e transparency platform – a review of europe’s most ambitious electricity data platform

    Lion Hirth, Jonathan Mühlenpfordt, and Marisa Bulkeley. The entso-e transparency platform – a review of europe’s most ambitious electricity data platform. Applied Energy, 225: 0 1054--1067, 2018. ISSN 0306-2619. doi:https://doi.org/10.1016/j.apenergy.2018.04.048. URL https://w...

    2018 doi

  28. [36]

    JuMP 1.0: R ecent improvements to a modeling language for mathematical optimization

    Miles Lubin, Oscar Dowson, Joaquim Dias Garcia , Joey Huchette, Beno \^i t Legat, and Juan Pablo Vielma. JuMP 1.0: R ecent improvements to a modeling language for mathematical optimization. Mathematical Programming Computation, 2023. doi:10.1007/s12532-023-00239-3

    2023 doi

  29. [37]

    MATPOWER : Steady-state operations, planning, and analysis tools for power systems research and education

    Ray Daniel Zimmerman, Carlos Edmundo Murillo-S \'a nchez, and Robert John Thomas. MATPOWER : Steady-state operations, planning, and analysis tools for power systems research and education. IEEE Transactions on power systems, 26 0 (1): 0 12--19, 2010

    2010

  30. [38]

    MathOptSymbolicAD.jl , 2022

    LANL-ANSI . MathOptSymbolicAD.jl , 2022. URL https://github.com/lanl-ansi/MathOptSymbolicAD.jl

    2022

  31. [39]

    GNU Parallel 20220522 ('NATO') , May 2022

    Ole Tange. GNU Parallel 20220522 ('NATO') , May 2022. URL https://doi.org/10.5281/zenodo.6570228

    2022 doi

  32. [40]

    Architecture of the Slurm workload manager

    Morris A Jette and Tim Wickberg. Architecture of the Slurm workload manager. In Workshop on Job Scheduling Strategies for Parallel Processing, pages 3--23. Springer, 2023

    2023

  33. [41]

    Lagrangian duality for constrained deep learning

    Ferdinando Fioretto, Pascal Van Hentenryck, Terrence WK Mak, Cuong Tran, Federico Baldo, and Michele Lombardi. Lagrangian duality for constrained deep learning. In Machine Learning and Knowledge Discovery in Databases. Applied Data Science and Demo Track: European Conference, ...

    2020

  34. [42]

    DC3 : A learning method for optimization with hard constraints

    Priya Donti, David Rolnick, and J Zico Kolter. DC3 : A learning method for optimization with hard constraints. In International Conference on Learning Representations, 2021

    2021

  35. [43]

    Differentiable convex optimization layers

    Akshay Agrawal, Brandon Amos, Shane Barratt, Stephen Boyd, Steven Diamond, and J Zico Kolter. Differentiable convex optimization layers. Advances in neural information processing systems, 32, 2019

    2019

  36. [44]

    Building highly detailed synthetic electric grid data sets for combined transmission and distribution systems

    Hanyue Li, Jessica L Wert, Adam Barlow Birchfield, Thomas J Overbye, Tomas Gomez San Roman, Carlos Mateo Domingo, Fernando Emilio Postigo Marcos, Pablo Duenas Martinez, Tarek Elgindy, and Bryan Palmintier. Building highly detailed synthetic electric grid data sets for combined...

    2020

  37. [45]

    Lectures on modern convex optimization: analysis, algorithms, and engineering applications

    Aharon Ben-Tal and Arkadi Nemirovski. Lectures on modern convex optimization: analysis, algorithms, and engineering applications. SIAM, 2001

    2001

  38. [46]

    Zimmerman and Carlos E

    Ray D. Zimmerman and Carlos E. Murillo-Sánchez. Matpower user's manual, May 2024. URL https://doi.org/10.5281/zenodo.11212313

    2024 doi

  39. [47]

    Large-scale nonlinear programming using Ipopt : An integrating framework for enterprise-wide dynamic optimization

    Lorenz T Biegler and Victor M Zavala. Large-scale nonlinear programming using Ipopt : An integrating framework for enterprise-wide dynamic optimization. Computers & Chemical Engineering, 33 0 (3): 0 575--582, 2009

    2009

  40. [48]

    MA27 - a set of Fortran subroutines for solving sparse symmetric sets of linear equations

    Iain S Duff and John K Reid. MA27 - a set of Fortran subroutines for solving sparse symmetric sets of linear equations . UKAEA Atomic Energy Research Establishment, 1982

    1982

  41. [49]

    LibHSL : the ultimate collection for large-scale scien-tific computation

    J Fowkes, A Lister, A Montoison, and D Orban. LibHSL : the ultimate collection for large-scale scien-tific computation. Les Cahiers du GERAD ISSN, 711: 0 2440, 2024

    2024

  42. [50]

    Gopinath, H.L

    S. Gopinath, H.L. Hijazi, T. Weisser, H. Nagarajan, M. Yetkin, K. Sundar, and R.W. Bent. Proving global optimality of acopf solutions. Electric Power Systems Research, 189: 0 106688, 2020. ISSN 0378-7796. doi:https://doi.org/10.1016/j.epsr.2020.106688. URL https://www.scienced...

    2020

  43. [51]

    R. A. Jabr. Radial distribution load flow using conic programming. IEEE Transactions on Power Systems, 21 0 (3): 0 1458--1459, Aug 2006 b . ISSN 0885-8950. doi:10.1109/TPWRS.2006.879234

    2006

  44. [52]

    A conic quadratic format for the load flow equations of meshed networks

    Rabih A Jabr. A conic quadratic format for the load flow equations of meshed networks. IEEE Transactions on Power Systems, 22 0 (4): 0 2285--2286, 2007

    2007

  45. [53]

    Goulart and Yuwen Chen

    Paul J. Goulart and Yuwen Chen. Clarabel: An interior-point solver for conic programs with quadratic objectives. arXiv prepring arXiv:2405.12762, 2024

    2024 arXiv

  46. [54]

    Huangfu and J

    Q. Huangfu and J. A. J. Hall. Parallelizing the dual revised simplex method. Mathematical Programming Computation, 10 0 (1): 0 119--142, 2018. doi:10.1007/s12532-017-0130-5

    2018 doi

  47. [55]

    Hierarchical Data Format, version 5 , 2024

    The HDF Group . Hierarchical Data Format, version 5 , 2024. URL https://github.com/HDFGroup/hdf5

    2024

  48. [56]

    P artnership for an A dvanced C omputing E nvironment ( PACE ) , 2017

    PACE. P artnership for an A dvanced C omputing E nvironment ( PACE ) , 2017. URL http://www.pace.gatech.edu

    2017

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.