Pith. sign in

REVIEW 5 major objections 5 minor 1 cited by

Voltage Regulation in Distribution Systems with Data Center Loads

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

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

Pith's one-line read The paper claims that data centers can regulate distribution voltage by adjusting GPU clock frequency (DVFS) for active power and rectifiers for reactive power through a local, communication-free droop controller, keeping more than 97% of…

desk verdict Useful idea with real data, but the printed control law has a sign error that would make it positive feedback rather than regulation; fix the equations and it deserves a serious look. read the letter →

arxiv 2507.06416 v1 pith:4KAQIUEB submitted 2025-07-08 eess.SY cs.SY

Yize Chen , Baosen Zhang This is my paper · ORCID
classification eess.SYcs.SY
keywords voltageregulationdatacentersDVFSdistributedcontroldroopdistributionsystemsLLMinferenceloadreactivepower
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 claims that the volatile power draw of AI data centers—GPU ramps that can swing within seconds—can be turned from a voltage problem into a voltage solution. By running a simple linear droop controller that reads local voltage and adjusts active power via dynamic voltage and frequency scaling (DVFS) and reactive power via rectifiers, each data center can counteract its own disturbances without real-time communication. In simulations on the IEEE 123-bus system with real LLM inference loads, more than 97% of buses stay within 0.05 p.u. of nominal voltage, and the scheme beats inverter-only droop control by at least 12.8% in average voltage deviation. If the hardware can execute the commanded frequency changes within the one-second control interval, data centers become a practical fast voltage-regulation resource at the distribution level.

What carries the argument

The controller is a linear droop law on active power, $u^P_{i,t+1} = [P^{\mathrm{ref}}_{i,t} - k_{p,i}(V_{i,t}-1) + \alpha_{i,t} S_{i,t}]_{\underline{P}_i}^{\overline{P}_i}$, where $S_{i,t}$ accumulates the unserved compute load so that total energy over the horizon stays invariant, and a standard reactive droop $u^Q_{i,t+1} = [u^Q_{i,t} - k_{q,i}(V_{i,t}-1)]_{\underline{Q}_i}^{\overline{Q}_i}$. The active-power knob is GPU clock frequency, which the paper treats as linearly related to power draw ($\tilde{P}_i \sim f_{\mathrm{clk},i}$), so frequency can be set as a function of voltage deviation. The grid model is the LinDistFlow approximation $v = Rp + Xq + \mathbf{1}$, with simulation validated against nonlinear branch flow equations.

What would settle it

Measure DVFS command-to-power-response time on a GPU cluster and compare with the one-second control interval; if the realized power lags the commanded setpoint by more than a second, the simulated voltage deviations and the reported 12.8% gain over inverter-only control would not materialize in hardware.

Watch

Extended reading notes

Core claim

The central claim is that a communication-free, local droop controller—active power proportional to voltage deviation via DVFS, reactive power via rectifier injection—can keep distribution voltages near 1 p.u. even when data center loads swing violently. The paper demonstrates this in the IEEE 123-bus system with real LLM inference traces: with five data center buses, average voltage deviation stays at 0.024 p.u. versus 0.034 p.u. for inverter-only control, and bus 13, which leaves the [0.95, 1.05] p.u. band under inverter-only control, is pulled back inside when data centers participate. The control respects a total energy constraint designed to preserve quality of service, and it requires no communication: each bus acts on its own voltage measurement.

Load-bearing premise

The load-bearing premise is that a data center can predict its next-step compute demand and that GPU frequency changes actually realize the commanded power within the one-second control interval, without added latency or hardware delay.

Editorial extensions

If this is right

  • Data centers can serve as primary voltage regulators without needing real-time communication, acting on local measurements only.
  • The scheme shifts LLM query load in time, so voltage regulation comes at the cost of mild inference delay; the gain $k_p$ controls the tradeoff.
  • Incorporating data center active-power control reduces the need for fast inverter-based reactive reserves in distribution feeders.
  • The approach extends to other controllable computing loads beyond GPUs, as long as power draw scales with a controllable clock frequency.

Reading between the lines

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

  • If DVFS response time is slower than the one-second interval, the effective droop gain is lower than designed, so a hardware-aware gain tuning or faster control interval would be needed in practice.
  • The same local droop logic could be applied to data center UPS/battery systems or cooling loads to provide transient voltage support during dips.
  • A testable extension is to run the controller on hardware-in-the-loop with actual GPU frequency ramps and compare the measured voltage profile against the simulated one.
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

5 major / 5 minor

Summary. The paper proposes a distributed, communication-free voltage regulation scheme for distribution systems with AI data center loads. Data centers adjust active power via GPU DVFS and reactive power via rectifiers, using local voltage measurements and simple droop laws. A cumulative energy-balance term is introduced to enforce that total data center energy consumption remains unchanged, and a reactive droop loop is used for inverter buses. The scheme is evaluated on an IEEE 123-bus system with real LLM inference power traces, reporting that more than 97% of buses stay within 0.05 p.u. voltage deviation and that the approach outperforms inverter-only droop control by more than 12.8%. The LLM power data and simulation code are open sourced.

Significance. If the results hold, the paper offers a timely and practical mechanism for using AI data centers as fast voltage-support resources, addressing a growing concern given the rapid deployment of large data centers. The use of real LLM inference traces and the open-sourcing of both data and code are notable strengths that aid reproducibility. The contribution is, however, primarily simulation-based, and the paper's central quantitative claims depend on resolving several technical inconsistencies in the controller equations and on clarifying assumptions about actuation speed and QoS modeling. With those clarifications, the work could be a useful reference for distribution grid operators and data center designers.

major comments (5)
  1. [Section II-B, Eqs. (1), (2), (5)] The sign convention in the voltage model and the active-power droop law is internally inconsistent. Eqs. (1) and (2) write v = Rp + Xq + 1 and v_{t+1} = R u^P_t + X u^Q_t + 1, while Section II-B and Eq. (3b) explicitly describe u^P as active power consumption. Under this convention, decreasing u^P when V > 1, as Eq. (5) does, would lower V further, producing positive feedback rather than regulation. If instead u^P is intended to be a net injection (with consumption entered as negative injection), then the notation P^ref as consumption and the energy-balance constraint (3g) must be reconciled. This is load-bearing because the reported voltage improvements are attributed to Eq. (5); the authors must correct either the voltage model or the control law and state the convention used in the simulation.
  2. [Section III, Eq. (6)] The stated role of S_t and its update equation do not match. The text says S_{i,t} keeps track of the power not served up to time t, which requires a cumulative recurrence such as S_{t+1} = S_t + P^ref_t - u^P_{t+1}. Eq. (6), as printed, defines S_{i,t+1} = P^ref_{i,t} - u^P_{i+1,t}, which is only the one-step tracking error and cannot enforce the cumulative energy-balance constraint (3g). This discrepancy undermines the explanation of the query delays in Table I, which are attributed to the integral action. Please provide the exact recurrence used in the simulator and align Eq. (6) with it.
  3. [Section III, Eq. (5)] The adaptive gain alpha_{i,t} is not specified. The text gives only a qualitative description ("if S_{i,t} is large ... alpha_{i,t} increased" and conversely), with no update rule, initial value, or numerical values used in the simulations. Since alpha multiplies the cumulative energy term and directly influences the load-delay tradeoff reported in Table I, the results are not reproducible without this information. Please provide the adaptation law and the parameter values for each scenario.
  4. [Section II-B and Section IV-A] The controller assumes that P^ref_{i,t+1} is exactly predictable at time t and that the DVFS command is realized within the 1-second control interval with no latency or tracking error. No measurement or model of GPU frequency-ramp time, scheduler delay, or actuation lag is provided. Because the simulated voltage improvements rely on the commanded active power being delivered at the next timestep, this idealization is load-bearing for the quantitative claims; the paper should either add a latency model or explicitly state this assumption as a limitation and discuss its impact.
  5. [Table I and Section IV-B] The reported "Average Delay (s)" is a key part of the claimed tradeoff between voltage regulation and QoS, but the manuscript does not describe how this delay is computed from the LLM workload and the control action. No model, formula, or simulator description is given, and the code location is not referenced in the text. Please define the delay metric and explain how it is derived, so that the tradeoff can be evaluated independently.
minor comments (5)
  1. [Section III, Eq. (5)] The index on u^P_{i+1,t} appears to be a typo; to match Eq. (3b), the update should be u^P_{i,t+1}.
  2. [Section IV-B, Table I] The numerical entries in Table I contain formatting artifacts (e.g., "0 .3" and spacing before "3"), which should be cleaned up.
  3. [Section IV-B] The word "signficant" is misspelled and should be "significant."
  4. [Section III] The sentence "We control in turn control set the clock frequency as a linear function of the voltage deviation" is garbled and should be rewritten for clarity.
  5. [Section IV-B] The claim that "more than 97% of the buses always have a voltage deviation within 0.05 p.u." would be easier to verify if a quantile or empirical CDF of the voltage deviations were plotted or tabulated; the current figures do not directly support this specific number.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the voltage-regulation claims are simulation outputs, not fits, and the cited prior work is background only.

full rationale

The paper's central claim is an empirical simulation result: applying local droop laws (Eqs. (5)-(7)) to data-center active and reactive injections in an IEEE 123-bus nonlinear DistFlow simulator yields reported voltage deviations around 0.02-0.03 p.u. and more than 12.8% improvement over inverter-only droop. Nothing in the derivation fits a parameter to the reported voltage outcome: the droop gains kp and kq are swept (kp = 1, 10, 20; standard reactive droop), alpha_i,t is an auxiliary tuning term for the energy-balance goal rather than a parameter identified from the voltage data, and the voltage is evaluated with independent nonlinear DistFlow equations rather than with the linear model (1) used to motivate the controller. The self-citations ([15], [17], [30]) support background statements about inverter voltage control and LLM power transients; none is used as the load-bearing justification for the proposed controller or for the simulation result. The apparent sign and indexing inconsistencies in Eqs. (5)-(6) noted in review are correctness concerns about whether the printed law matches the implementation, not circular reasoning, so they do not enter this score.

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

The ledger shows the central claim rests on standard power-flow and hardware assumptions plus a small set of hand-scaled control gains. No exotic entities are introduced, but the unspecified alpha tuning and load-scaling factor are the main ad hoc elements.

free parameters (4)
  • kp,i (active power droop gain) = swept over 1, 10, 20
    Voltage deviation and delay results in Table I depend directly on this gain; it is chosen by the authors, not derived.
  • alpha_i,t (energy-balance integrator gain) = not specified
    Equation (5) uses alpha_i,t to enforce the total energy constraint; the paper says it is tuned to encourage satisfaction but gives no update law or values, so it is an ad hoc tuning parameter.
  • kq,i (reactive droop gain) = not specified
    Reactive droop Eq. (7) uses kq,i; its value is not reported, yet it affects the inverter and data center reactive response.
  • GPU load scaling factor = 8xA100 traces scaled to 280 kW peak
    The mapping from measured cluster power to data center load is a scenario choice; the paper does not give the multiplier or the snippet selection procedure, which affects all reported deviations.
assumptions (5)
  • domain assumption LinDistFlow linear model v = Rp + Xq + 1 accurately represents voltage magnitudes in the distribution feeder
    Used to design the droop controller (Eq. (1)-(2)); simulation actually uses nonlinear DistFlow, so the design model is an approximation.
  • domain assumption GPU power consumption scales linearly with clock frequency (P ~ fclk)
    Invoked in Section III before Eq. (4) to justify the active-power droop; real chips have a nonlinear DVFS curve and other power components.
  • domain assumption Next-step data center load Pref is known at time t
    Assumed in Section II-B with the justification that compute workloads are predictable over short timescales; this predictability underpins the feedforward-like use of Pref in Eq. (5).
  • domain assumption Voltage measurements are locally available and accurate enough for droop control
    The distributed scheme uses Vi,t at each bus; no measurement noise or communication delay is modeled.
  • domain assumption Total active energy of each data center is preserved over the horizon (Eq. (3g))
    The QoS constraint is asserted as a requirement; the paper does not prove the droop law actually satisfies it, relying on the tuned alpha term.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Voltage Regulation in Distribution Systems with Data Center Loads." pith.science (2026). https://pith.science/paper/4KAQIUEB

@misc{pith2026250706416,
  author       = {Pith},
  title        = {Pith review of: Voltage Regulation in Distribution Systems with Data Center Loads},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4KAQIUEB}},
  note         = {Machine review of arXiv:2507.06416}
}
read the original abstract

Recent boom in foundation models and AI computing have raised growing concerns on the power and energy trajectories of large-scale data centers. This paper focuses on the voltage issues caused by volatile and intensity of data center power demand, which also aligns with recent observations of more frequent voltage disturbances in power grids. To address these data center integration challenges, we propose a dynamic voltage control scheme by harnessing data center's load regulation capabilities. By taking local voltage measurements and adjusting power injections at each data center buses through the dynamic voltage and frequency scaling (DVFS) scheme, we are able to maintain safe voltage magnitude in a distributed fashion with higher data center computing load. Simulations using real large language model (LLM) inference load validate the effectiveness of our proposed mechanism. Both the LLM power data and proposed control scheme are open sourced.

Figures

Figures reproduced from arXiv: 2507.06416 by the authors.

Figure 1
Figure 1. Power consumption curve for one GPU during the model training [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Power profiles of the four LLM inference tasks reveal distinct power [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Voltages in a 6-bus toy example. Without voltage regulation, the [PITH_FULL_IMAGE:figures/full_fig_p002_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: IEEE 123-bus system with AI data center load buses. [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Average voltage magnitude deviation (p.u.) and standard deviation [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 7
Figure 7. Figure 7: Data center original load curve and load curve after implementing [PITH_FULL_IMAGE:figures/full_fig_p005_7.png]
Figure 6
Figure 6. Figure 6: 2-minute voltage curves for selected two data center buses, a nominal [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]

Discussion (0). Continue with ORCID 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. GPU-to-Grid: Voltage Regulation via GPU Utilization Control

    eess.SY 2026-02 conditional novelty 6.0 of 10

    GPU batch-size control, driven by real-time voltage and latency feedback, can serve as a fast distribution-grid voltage regulation resource.

Reference graph

Works this paper leans on

33 extracted references · 27 canonical work pages · cited by 1 Pith paper

  1. [1]

    On the opportunities and risks of foundation models,

    R. Bommasani, D. A. Hudson, E. Adeli, R. Altman, S. Arora, S. von Arx, M. S. Bernstein, J. Bohg, A. Bosselut, E. Brunskill et al. , “On the opportunities and risks of foundation models,” arXiv preprint arXiv:2108.07258, 2021

    arXiv 2021

  2. [2]

    Sparks of artificial general intelligence: Early experiments with gpt-4,

    S. Bubeck, V . Chandrasekaran, R. Eldan, J. Gehrke, E. Horvitz, E. Ka- mar, P. Lee, Y . T. Lee, Y . Li, S. Lundberg et al. , “Sparks of artificial general intelligence: Early experiments with gpt-4,” arXiv preprint arXiv:2303.12712, 2023

    arXiv 2023

  3. [3]

    The llama 3 herd of models,

    A. Dubey, A. Jauhri, A. Pandey, A. Kadian, A. Al-Dahle, A. Letman, A. Mathur, A. Schelten, A. Yang, A. Fan et al. , “The llama 3 herd of models,” arXiv preprint arXiv:2407.21783 , 2024

    arXiv 2024

  4. [4]

    Electricity 2024,

    International Energy Agency, “Electricity 2024,” https://www.iea.org/ reports/electricity-2024, 2024

    work page 2024

  5. [5]

    Reducing the carbon impact of generative ai inference (today and in 2035),

    A. A. Chien, L. Lin, H. Nguyen, V . Rao, T. Sharma, and R. Wijayawar- dana, “Reducing the carbon impact of generative ai inference (today and in 2035),” in Proceedings of the 2nd workshop on sustainable computer systems , 2023, pp. 1–7

    work page 2023

  6. [6]

    Thermal Analysis for NVIDIA GTX480 Fermi GPU Architecture

    S. Nagendra, “Thermal analysis for nvidia gtx480 fermi gpu architec- ture,” arXiv preprint arXiv:2403.16239 , 2024

    work page Pith review arXiv 2024

  7. [7]

    A 1500-a/48-v-to-1-v switching bus converter for next-generation ultra-high-power processors,

    Y . Zhu, J. Zou, and R. C. Pilawa-Podgurski, “A 1500-a/48-v-to-1-v switching bus converter for next-generation ultra-high-power processors,” IEEE Transactions on Power Electronics , 2024

    work page 2024

  8. [8]

    Data centers and energy: Reaching sustainability, the vital role of heat reuse,

    J. J. Jensen, “Data centers and energy: Reaching sustainability, the vital role of heat reuse,” Danfoss, 2024, accessed: 2025-04-30

    work page 2024

Show all 33 references
  1. [9]

    PG&E expects 3.5 GW data center growth by 2029,

    S. Wolfe, “PG&E expects 3.5 GW data center growth by 2029,” Re- newable Energy World , 2024, accessed: 2025-04-30

    2024

  2. [10]

    Green- ing geographical load balancing,

    Z. Liu, M. Lin, A. Wierman, S. H. Low, and L. L. Andrew, “Green- ing geographical load balancing,” ACM SIGMETRICS Performance Evaluation Review, vol. 39, no. 1, pp. 193–204, 2011

    2011

  3. [11]

    A manifesto for future generation cloud computing: Research directions for the next decade,

    R. Buyya et al. , “A manifesto for future generation cloud computing: Research directions for the next decade,” ACM computing surveys (CSUR), vol. 51, no. 5, pp. 1–38, 2018

    2018

  4. [12]

    Virginia narrowly avoided power cuts when 60 data centers dropped off the grid at once,

    M. Gooding, “Virginia narrowly avoided power cuts when 60 data centers dropped off the grid at once,” Data Center Dynamics , 2025

    2025

  5. [13]

    Incident review considering simultaneous voltage-sensitive load reductions,

    NERC, “Incident review considering simultaneous voltage-sensitive load reductions,” 2025, accessed: 2025-04-30

    2025

  6. [14]

    Adaptive var control for distribution circuits with photovoltaic generators,

    H.-G. Yeh, D. F. Gayme, and S. H. Low, “Adaptive var control for distribution circuits with photovoltaic generators,” IEEE Transactions on Power Systems , vol. 27, no. 3, pp. 1656–1663, 2012

    2012

  7. [15]

    An optimal and distributed method for voltage regulation in power distribution systems,

    B. Zhang, A. Y . Lam, A. D. Dom´ınguez-Garc´ıa, and D. Tse, “An optimal and distributed method for voltage regulation in power distribution systems,” IEEE Transactions on Power Systems , vol. 30, no. 4, pp. 1714–1726, 2014

    2014

  8. [16]

    V oltage regulation in distribution grids: A survey,

    P. Srivastava, R. Haider, V . J. Nair, V . Venkataramanan, A. M. An- naswamy, and A. K. Srivastava, “V oltage regulation in distribution grids: A survey,” Annual Reviews in Control , vol. 55, pp. 165–181, 2023

    2023

  9. [17]

    The unseen ai disruptions for power grids: Llm-induced transients,

    Y . Li, M. Mughees, Y . Chen, and Y . R. Li, “The unseen ai disruptions for power grids: Llm-induced transients,” arXiv preprint arXiv:2409.11416 , 2024

    2024 arXiv

  10. [18]

    Static and dynamic frequency scal- ing on multicore cpus,

    W. Bao, C. Hong, S. Chunduri, S. Krishnamoorthy, L.-N. Pouchet, F. Rastello, and P. Sadayappan, “Static and dynamic frequency scal- ing on multicore cpus,” ACM Transactions on Architecture and Code Optimization (TACO), vol. 13, no. 4, pp. 1–26, 2016

    2016

  11. [19]

    Dynamic gpu energy optimization for machine learning training workloads,

    F. Wang, W. Zhang, S. Lai, M. Hao, and Z. Wang, “Dynamic gpu energy optimization for machine learning training workloads,” IEEE Transactions on Parallel and Distributed Systems , vol. 33, no. 11, pp. 2943–2954, 2021

    2021

  12. [20]

    Benchmarking and dissecting the nvidia hopper gpu architecture,

    W. Luo, R. Fan, Z. Li, D. Du, Q. Wang, and X. Chu, “Benchmarking and dissecting the nvidia hopper gpu architecture,” in 2024 IEEE International Parallel and Distributed Processing Symposium (IPDPS) . IEEE, 2024, pp. 656–667

    2024

  13. [21]

    Agarwal and J

    A. Agarwal and J. Lang, F oundations of analog and digital electronic circuits. Elsevier, 2005

    2005

  14. [22]

    J. M. Rabaey, A. Chandrakasan, and B. Nikolic, Digital integrated circuits. Prentice hall Englewood Cliffs, 2002, vol. 2

    2002

  15. [23]

    Energy efficiency and renewable energy integration in data centres. strategies and modelling review,

    E. Or ´o, V . Depoorter, A. Garcia, and J. Salom, “Energy efficiency and renewable energy integration in data centres. strategies and modelling review,” Renewable and Sustainable Energy Reviews , vol. 42, pp. 429– 445, 2015

    2015

  16. [24]

    Ensuring renewable energy utilization with quality of service guarantee for energy-efficient data center operations,

    S. Kwon, “Ensuring renewable energy utilization with quality of service guarantee for energy-efficient data center operations,” Applied Energy , vol. 276, p. 115424, 2020. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0306261920309363

    2020

  17. [25]

    A survey on green-energy-aware power manage- ment for datacenters,

    F. Kong and X. Liu, “A survey on green-energy-aware power manage- ment for datacenters,” ACM Computing Surveys (CSUR) , vol. 47, no. 2, pp. 1–38, 2014

    2014

  18. [26]

    Dy- namollm: Designing llm inference clusters for performance and energy efficiency,

    J. Stojkovic, C. Zhang, ´I. Goiri, J. Torrellas, and E. Choukse, “Dy- namollm: Designing llm inference clusters for performance and energy efficiency,” arXiv preprint arXiv:2408.00741 , 2024

    2024

  19. [27]

    N. E. M. Association et al. , American National Standard for Electric Power Systems and Equipment-V oltage Ratings (60 Hertz) . National Electrical Manufacturers Association, 1996

    1996

  20. [28]

    A note on branch flow models with line shunts,

    F. Zhou and S. H. Low, “A note on branch flow models with line shunts,” IEEE Transactions on Power Systems , vol. 36, no. 1, pp. 537–540, 2020

    2020

  21. [29]

    Fast local voltage control under limited reactive power: Optimality and stability analysis,

    H. Zhu and H. J. Liu, “Fast local voltage control under limited reactive power: Optimality and stability analysis,” IEEE Transactions on Power Systems, vol. 31, no. 5, pp. 3794–3803, 2015

    2015

  22. [30]

    Decentralized safe reinforcement learning for inverter-based voltage control,

    W. Cui, J. Li, and B. Zhang, “Decentralized safe reinforcement learning for inverter-based voltage control,” Electric Power Systems Research , vol. 211, p. 108609, 2022

    2022

  23. [31]

    Options for control of reactive power by distributed photovoltaic generators,

    K. Turitsyn, P. Sulc, S. Backhaus, and M. Chertkov, “Options for control of reactive power by distributed photovoltaic generators,” Proceedings of the IEEE , vol. 99, no. 6, pp. 1063–1073, 2011

    2011

  24. [32]

    Geometry of the feasible output regions of grid-interfacing inverters with current limits,

    L. Streitmatter, T. Joswig-Jones, and B. Zhang, “Geometry of the feasible output regions of grid-interfacing inverters with current limits,” IEEE Control Systems Letters , 2025

    2025

  25. [33]

    Optimal sizing of capacitors placed on a radial distribution system,

    M. Baran and F. F. Wu, “Optimal sizing of capacitors placed on a radial distribution system,” IEEE Transactions on power Delivery , vol. 4, no. 1, pp. 735–743, 2002

    2002

Pith tools

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