Pith. sign in

REVIEW 1 major objections 1 minor 53 references

Grid-Interactive Thermal Management of AI Data Centers via Contextual Distributionally Robust Optimization

T0 review · 1 major / 1 minor · reviewed 2026-07-02 · grok-4.3

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

Pith's one-line read A contextual distributionally robust controller maintains near-zero thermal violations in AI data centers while cutting the cost of robustness by 13.7 percentage points versus min-max MPC.

desk verdict The paper's contextual Wasserstein radius adaptation for DRO cooling control is the claimed step forward, but the safety guarantees do not clearly survive the switch from fixed to dynamic radius. read the letter →

arxiv 2607.00099 v1 pith:GLOILLJK submitted 2026-06-30 eess.SY cs.SY

Jiachen Shen , Jian Shi , Yijie Yang , Chenye Wu , Dan Wang , Ju Bin Song , Zhu Han This is my paper · ORCID
classification eess.SYcs.SY
keywords datacenterthermalmanagementdistributionallyrobustoptimizationgrid-interactivecontrolAIworkloaduncertaintydemandresponsemodelpredictiveWassersteinambiguityset
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 develops a Contextual Distributionally Robust Optimization framework that adjusts the size of the uncertainty set in real time using AI workload and grid signals. This replaces fixed conservative bounds or purely forecast-based control with an adaptive approach that shrinks uncertainty when conditions are stable. The method is cast as an inf-sup problem, then turned into a tractable form via an exact reformulation of the worst-case cost and a conservative deterministic version of the DR-CVaR safety constraint. High-fidelity EnergyPlus co-simulations show the resulting controller keeps thermal violations near zero even under extreme spikes while lowering the extra operating cost of robustness. A reader would care because the result directly improves the ability of large AI facilities to provide flexible demand response without sacrificing equipment safety.

What carries the argument

Context-dependent Wasserstein radius that dynamically tightens or loosens the ambiguity set inside a distributionally robust MPC problem, carrying the safety-flexibility tradeoff.

What would settle it

A high-fidelity co-simulation run in which a workload spike outside the observed context variables produces a measurable thermal violation under the CDRO policy but not under a fixed-radius DRO policy with the same nominal radius.

Watch

Extended reading notes

Core claim

By making the Wasserstein radius depend on real-time context, the CDRO controller produces a scalable ADMM solution whose DR-CVaR thermal constraint remains safe while the expected-cost term is exactly reformulated, yielding near-zero violations and a 13.7-percentage-point reduction in the robustness cost premium relative to standard Min-Max MPC.

Load-bearing premise

Real-time AI and grid context can be used to shrink the Wasserstein radius without leaving out uncertainty that would produce actual thermal violations.

Editorial extensions

If this is right

  • The controller can be implemented at scale using a nested ADMM algorithm without losing the safety guarantee.
  • Demand-side flexibility increases during stable regimes because uncertainty bounds are allowed to contract.
  • Thermal safety is preserved under extreme spikes that defeat forecast-driven MPC.
  • The operational cost of maintaining robustness drops by roughly 13.7 percentage points compared with min-max MPC.

Reading between the lines

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

  • The same contextual-radius idea could be tested on other grid-interactive loads whose uncertainty also varies with observable signals, such as EV fleets or commercial HVAC.
  • If the context variables prove sufficient, the method might reduce the conservatism that currently limits data-center participation in real-time ancillary services.
  • A natural next measurement would be the sensitivity of the 13.7-point saving to the choice of context features and to the frequency of radius updates.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

1 major / 1 minor

Summary. The paper proposes a Contextual Distributionally Robust Optimization (CDRO) framework for grid-interactive thermal management of AI data centers. It dynamically adapts the Wasserstein radius based on real-time AI and grid context to formulate an infinite-dimensional inf-sup problem. The authors derive an exact tractable reformulation for the Wasserstein worst-case expected-cost term and a tractable conservative deterministic counterpart for the DR-CVaR thermal safety constraint. The problem is solved using a nested ADMM algorithm. High-fidelity EnergyPlus co-simulations are used to demonstrate near-zero thermal violations under extreme workload spikes and a 13.7 percentage point reduction in the operational cost premium of robustness compared to standard Min-Max MPC.

Significance. If the safety guarantees hold under the proposed dynamic radius adaptation, the framework could enable greater demand-side flexibility in data center cooling while maintaining thermal safety, potentially reducing costs associated with conservative robustness. The combination of theoretical reformulations and high-fidelity simulations provides a solid basis for practical impact in systems and control applications for energy systems.

major comments (1)
  1. [CDRO formulation and theoretical derivations] The exact reformulations and conservative DR-CVaR counterpart are presented for a fixed Wasserstein radius. The introduction of a context-dependent rule to shrink the radius in stable regimes lacks a demonstration that this adaptation preserves the ambiguity-set inclusion and the conservatism margin required by the DR-CVaR reformulation. This is load-bearing for the central safety claim of near-zero thermal violations, as noted in the abstract.
minor comments (1)
  1. [Abstract] The abstract would benefit from referencing specific equation numbers for the reformulations to allow readers to trace the claims more easily.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful review and for highlighting this important theoretical point. We respond to the major comment below.

read point-by-point responses

  1. Referee: [CDRO formulation and theoretical derivations] The exact reformulations and conservative DR-CVaR counterpart are presented for a fixed Wasserstein radius. The introduction of a context-dependent rule to shrink the radius in stable regimes lacks a demonstration that this adaptation preserves the ambiguity-set inclusion and the conservatism margin required by the DR-CVaR reformulation. This is load-bearing for the central safety claim of near-zero thermal violations, as noted in the abstract.

    Authors: We agree that the manuscript presents the exact reformulations and the conservative DR-CVaR counterpart under a fixed Wasserstein radius, and that the context-dependent adaptation rule is introduced without an explicit proof that the dynamic choice preserves ambiguity-set inclusion and the required conservatism margin. This is a valid observation. In the revised version we will insert a new subsection (Section 3.4) that formally proves the preservation property: we show that the context-dependent radius rule is constructed to ensure the adapted ambiguity set remains a valid subset of the original Wasserstein ball for any realized context, and that the resulting DR-CVaR upper bound remains conservative with respect to the fixed-radius case. The proof relies on the monotonicity of the Wasserstein distance with respect to the radius and on the Lipschitz continuity of the context-to-radius mapping. This addition will directly strengthen the safety guarantees supporting the near-zero violation claims. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity detected in derivation chain.

full rationale

The abstract outlines an inf-sup formulation followed by an exact reformulation of the Wasserstein worst-case term and a conservative deterministic counterpart for the DR-CVaR constraint. These steps are presented as standard derivations for fixed-radius DRO and do not reduce to the inputs by construction or via self-citation. The contextual radius adaptation is described as an extension that shrinks bounds in stable regimes, but the provided text contains no equations or claims showing that this adaptation is defined in terms of the safety guarantee itself or that any prediction is statistically forced by a fitted parameter. Simulation performance claims are empirical outcomes rather than derivations that collapse to the model inputs. No load-bearing self-citations or ansatz smuggling appear in the given content, so the derivation chain remains self-contained against external benchmarks.

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

Abstract supplies no information on free parameters, background axioms, or new entities; all such elements remain unknown.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Grid-Interactive Thermal Management of AI Data Centers via Contextual Distributionally Robust Optimization." pith.science (2026). https://pith.science/paper/GLOILLJK

@misc{pith2026260700099,
  author       = {Pith},
  title        = {Pith review of: Grid-Interactive Thermal Management of AI Data Centers via Contextual Distributionally Robust Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GLOILLJK}},
  note         = {Machine review of arXiv:2607.00099}
}
read the original abstract

Thermal management in AI data centers is increasingly challenged by bursty workloads and uncertain heat generation. To prevent thermal violations, existing cooling strategies either enforce conservative, rigid bounds that severely limit grid responsiveness, or rely on forecast-driven controllers that perform poorly under AI workload uncertainty and distribution shifts. To overcome the above challenges, this paper proposes a Contextual Distributionally Robust Optimization (CDRO) framework for grid-interactive cooling control. Unlike standard DRO with fixed ambiguity sets, the proposed approach dynamically adapts the Wasserstein radius using real-time AI and grid context. This safely shrinks uncertainty bounds during stable regimes, unlocking deep demand-side flexibility. Theoretically, we formulate the control as an infinite-dimensional inf-sup problem, derive an exact tractable reformulation for the Wasserstein worst-case expected-cost term, and then derive a tractable conservative deterministic counterpart for the Distributionally Robust Conditional Value at Risk (DR-CVaR) thermal safety constraint. Solved via a scalable nested Alternating Direction Method of Multipliers (ADMM) algorithm, the CDRO controller achieves near-zero thermal violations under extreme workload spikes in high-fidelity EnergyPlus co-simulations. Simultaneously, it reduces the operational cost premium of robustness by approximately 13.7 percentage points relative to standard Min-Max Model Predictive Control (MPC).

Figures

Figures reproduced from arXiv: 2607.00099 by the authors.

Figure 1
Figure 1. System Framework: The integrated control loop from external [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Thermal response during an AI workload spike. CDRO [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Pareto trade-off for Heatwave (TCO vs. EVP). CDRO outper [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (2 more)
Figure 6
Figure 6. Figure 6: Risk-Cost Pareto frontier. Varying the DR-CVaR tail parameter [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Empirical violation probability under forecast degradation. [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 53 canonical work pages

  1. [1]

    AI inference and training workloads are characterized by extreme burstiness

    Timescale Mismatch between Thermal Shock and Cooling Dynamics: The first and most formidable challenge is the significant gap between the volatility of AI thermal loads and the response speed of industrial cooling infrastructure. AI inference and training workloads are characterized by extreme burstiness. A sudden surge in matrix multiplication operations...

    work page

  2. [2]

    Grid-Interactive Thermal Management of AI Data Centers via Contextual Distributionally Robust Optimization

    Over-Cooling Trap vs. Demand Flexibility: Due to the aforementioned thermal risks, current industry practices rely on extreme conservatism, known as over-cooling. This in- volves maintaining setpoints far lower than necessary to create a safety buffer against unforeseen load spikes. While physi- cally safe, this static conservatism rigidly locks the cooli...

    work page Pith review arXiv 2026

  3. [3]

    It lacks the foresight to manage heat accumulation, often resulting in oscillatory behavior and significant energy waste [8]

    Limitations of Existing Cooling Control: The industry standard, PID control, is purely reactive and ill-suited for the high-density cooling requirements of AI clusters. It lacks the foresight to manage heat accumulation, often resulting in oscillatory behavior and significant energy waste [8]. Model Predictive Control advances this by utilizing physics-ba...

    work page

  4. [4]

    Optimization under Uncertainty and Distributional Ro- bustness: Uncertainty in power and energy systems is com- monly handled via stochastic programming (SP), which typi- cally assumes known probability distributions and can become sensitive to distribution shift, and via robust optimization (RO), which enforces feasibility under bounded uncertainty sets ...

    work page

  5. [5]

    We model its dynamics using a first- order Resistor-Capacitor (RC) thermal circuit model [20]

    Control-Oriented Thermal Entity Dynamics: The core temperature of a controlled thermal entity (rack/zone) ( Tcore,z) is a critical state. We model its dynamics using a first- order Resistor-Capacitor (RC) thermal circuit model [20]. The continuous-time dynamics are discretized using a forward Euler method, resulting in the following linear state-space rep...

    work page

  6. [6]

    Room Air and Cooling Loop Dynamics: We model the cold aisle as a single, well-mixed air volume. Its average temperature, which serves as the shared inlet temperature Tin[t], is determined by an energy balance between the heat removed by the Computer Room Air Conditioning (CRAC) units and the heat recirculated from the hot aisle, following established lump...

    work page

  7. [7]

    CRAC Unit Thermal-Fluid Dynamics: The performance of each CRAC unit i ∈ I is described by the following relationships. The mass flow rate of air, mair,i[t], is directly proportional to the fan speed decision si[t]: mair,i[t] = mrated air,i · si[t], (4) where mrated air,i is the fan’s rated mass flow rate at full speed. The supply air temperature, Tsup,i[t...

    work page

  8. [8]

    Their power consumption is highly non-linear

    Cooling Plant Energy Model: The cooling plant consists of CRAC fans, chillers, cooling towers, and pumps. Their power consumption is highly non-linear. • CRAC Fans: The power of the fan in CRAC unit i, Pfan,i[t], follows the fan laws, scaling cubically with its normalized speed si[t]: Pfan,i[t] = P rated fan,i · (si[t])3. (7) • Chillers: The chiller power...

    work page

Show all 53 references
  1. [9]

    Contextual Ambiguity Set Construction: To rigorously define the online decision boundaries and prevent information leakage, we first specify the available information set It at decision epoch t: Definition 1 (Online Information Set) . Let It denote the filtration containing al...

  2. [10]

    To ensure dimensional consis- tency, we explicitly monetize carbon emissions and thermal risks using conversion factors

    Objective Function: The objective is to minimize the total operational cost in USD. To ensure dimensional consis- tency, we explicitly monetize carbon emissions and thermal risks using conversion factors. The objective is expressed as: J (u) = sup Pξ∈P(c) EPξ "X t∈T Cost[t] # ...

  3. [11]

    System Constraints: To formulate a clear control prob- lem, we organize the system constraints into three categories: (i) deterministic physics and operational constraints (state evolution, actuator bounds, energy balances); (ii) a robust objective term minimizing the worst-ca...

  4. [12]

    We define c[t] to include real-time values and indicators of volatility

    Context Features and k-NN Retrieval: The context vec- tor c[t] is important for capturing the distribution of residuals. We define c[t] to include real-time values and indicators of volatility. All features come from the online information set It. For example, we use the varia...

  5. [13]

    Standard theory often fails because AI residuals have heavy tails

    Radius Calibration via Safety-Driven Backtesting: A major challenge is choosing the Wasserstein radius ρ(c[t]). Standard theory often fails because AI residuals have heavy tails. We use a data-driven approach instead. We define the ambiguity set as: Pρ(c[t]) := n P ∈ M(Ξ) | W1...

  6. [14]

    The primal problem seeks to minimize the expected cost under the worst-case distribution within the Wasserstein ball

    Step 1: Strong Duality Reformulation: We first con- vert the stochastic worst-case expectation into a deterministic convex problem. The primal problem seeks to minimize the expected cost under the worst-case distribution within the Wasserstein ball. To address this infinite-di...

  7. [15]

    To solve this efficiently, we decompose the global problem into two smaller distinct 7 subproblems

    Step 2: Non-Convex ADMM Decomposition: Even after the deterministic reformulation, the problem remains a non- convex MISOCP due to the underlying physics of the cooling plant, specifically the cubic fan power laws ( P ∝ s3) and the bi-quadratic chiller COP curves. To solve thi...

  8. [16]

    However, our problem structure aligns with the class of non- convex problems analyzed by Wang et al

    Convergence of Non-Convex ADMM: Standard ADMM convergence proofs typically rely on the convexity of the ob- jective functions, which does not hold for the cubic fan power laws and bi-quadratic chiller curves inherent to our model. However, our problem structure aligns with the...

  9. [17]

    Computational Complexity and Scalability: The online execution solves two MISOCP subproblems per ADMM it- eration. For a realistic deployment, the optimization size depends on the prediction horizon H, CRAC count |I|, thermal entities |Lz|, k-NN sample size k, and the num- ber...

  10. [18]

    First, dis- tributional robustness is essential for safety

    Attribution of Performance Gains: Table II highlights two primary sources of CDRO’s performance gains. First, dis- tributional robustness is essential for safety. Nominal methods like CVaR-MPC exhibits a substantially higher out-of-sample EVP than CDRO under distribution shift...

  11. [19]

    As illustrated in Fig

    Robustness to Workload Uncertainty (Scenario 1): To qualitatively explain these statistical results, Scenario 1 exposes the transient fragility of forecast-dependent controls during a sudden workload burst. As illustrated in Fig. 2, deterministic MPC fails to anticipate the bu...

  12. [20]

    High Cost

    Resilience Under Physical Constraints (Scenario 2): Fig. 3 presents the trade-off between cost (TCO) and risk (EVP) for all methods during a heatwave scenario. CDRO consistently occupies the ideal region. Static methods like RO and Min-Max MPC fall into the “High Cost” region....

  13. [21]

    Unlike Min-Max MPC, which maintains a rigid safety buffer, CDRO’s context- aware ambiguity set allows it to identify periods of internal sta- bility

    Grid-Interactive Decision Making (Scenario 3): Figure 4 illustrates power response to a price spike. Unlike Min-Max MPC, which maintains a rigid safety buffer, CDRO’s context- aware ambiguity set allows it to identify periods of internal sta- bility. During these windows, it t...

  14. [22]

    Figure 5 illustrates the trade-off between conservatism (Cost) and safety (Risk) as the Wasserstein radius ρ varies

    Radius Calibration and Sensitivity: We validated the effectiveness of our data-driven calibration approach. Figure 5 illustrates the trade-off between conservatism (Cost) and safety (Risk) as the Wasserstein radius ρ varies. The solid lines represent the Context-Aware approach...

  15. [23]

    We performed a sensitivity analysis by varying the DR-CVaR risk level ε ∈ { 0.01,

    Sensitivity to Risk Preference ( ε): A key contribution of our work is the ability to explicitly tune the safety-cost trade-off. We performed a sensitivity analysis by varying the DR-CVaR risk level ε ∈ { 0.01, . . . ,0.25}. As shown in Fig. 6, the Risk-Cost Pareto Frontier ex...

  16. [24]

    De- terministic MPC and Nominal Risk methods (e.g., CVaR- MPC) degrade rapidly as forecasts worsen

    Robustness under Forecast Degradation: Figure 7 com- pares performance under varying forecast noise levels. De- terministic MPC and Nominal Risk methods (e.g., CVaR- MPC) degrade rapidly as forecasts worsen. In contrast, both CDRO and NC-DRO retain low empirical violation rate...

  17. [25]

    The nominal configuration requires a median solve time of only 12.4 seconds

    Real-time Feasibility and Scalability: To ensure the CDRO controller is viable for standard 5-minute ( ∆t) dis- patch intervals, we evaluated its computational scalability. The nominal configuration requires a median solve time of only 12.4 seconds. A small scaling sweep confi...

  18. [26]

    Artificial intelligence index report 2025,

    N. Maslej et al. , “Artificial intelligence index report 2025,” arXiv preprint arXiv:2504.07139, 2025

    2025

  19. [27]

    Electricity 2024: Analysis and forecast to 2026,

    International Energy Agency, “Electricity 2024: Analysis and forecast to 2026,” International Energy Agency, Tech. Rep., Jan. 2024

    2024

  20. [28]

    Towards carbon-free electricity: A flow-based framework for power grid carbon accounting and decar- bonization,

    X. Chen, H. Chao, W. Shi, and N. Li, “Towards carbon-free electricity: A flow-based framework for power grid carbon accounting and decar- bonization,” Energy Conversion and Economics , vol. 5, no. 6, pp. 396– 418, Dec. 2024

    2024

  21. [29]

    Enhance low-carbon power system operation via carbon- aware demand response,

    X. Chen, “Enhance low-carbon power system operation via carbon- aware demand response,” Energy Internet, vol. 1, no. 2, pp. 141–149, Nov. 2024

    2024

  22. [30]

    Data centres as a source of flexibility for power systems,

    M. T. Takci, M. Qadrdan, J. Summers, and J. Gustafsson, “Data centres as a source of flexibility for power systems,” Energy Reports, vol. 13, pp. 3661–3671, Jun. 2025

    2025

  23. [31]

    Data-driven flexibility assessment for internet data center towards periodic batch workloads,

    Y . Cao, M. Cheng, S. Zhang, H. Mao, P. Wang, C. Li, Y . Feng, and Z. Ding, “Data-driven flexibility assessment for internet data center towards periodic batch workloads,” Applied Energy, vol. 324, p. 119665, Oct. 2022

    2022

  24. [32]

    Green ai,

    R. Schwartz, J. Dodge, N. A. Smith, and O. Etzioni, “Green ai,” Communications of the ACM , vol. 63, no. 12, pp. 54–63, Dec. 2020

    2020

  25. [33]

    Optimization control strategies and evaluation metrics of cooling systems in data centers: A review,

    Q. Chang, Y . Huang, K. Liu, X. Xu, Y . Zhao, and S. Pan, “Optimization control strategies and evaluation metrics of cooling systems in data centers: A review,” Sustainability, vol. 16, no. 16, p. 7222, Aug. 2024

    2024

  26. [34]

    Data center cooling using model-predictive control,

    N. Lazic, C. Boutilier, T. Lu, E. Wong, B. Roy, M. K. Ryu, and G. Imwalle, “Data center cooling using model-predictive control,” in Advances in Neural Information Processing Systems 31 , Montr ´eal, Canada, Dec. 2018, pp. 3814–3823

    2018

  27. [35]

    Joint optimization of cooling parameters and workload distributions based on model predictive control for rack-based data centers,

    J. Wang, W. Deng, C. Yue, W. Su, and X. Bai, “Joint optimization of cooling parameters and workload distributions based on model predictive control for rack-based data centers,” Journal of Building Engineering , vol. 100, p. 111801, Apr. 2025

    2025

  28. [36]

    Reinforcement learning for data center energy efficiency optimization: A systematic literature review and research roadmap,

    H. Kahil, S. Sharma, P. V ¨alisuo, and M. Elmusrati, “Reinforcement learning for data center energy efficiency optimization: A systematic literature review and research roadmap,” Applied Energy , vol. 389, p. 125734, Jul. 2025

    2025

  29. [37]

    Green data center cooling control via physics-guided safe reinforcement learning,

    R. Wang, Z. Cao, X. Zhou, Y . Wen, and R. Tan, “Green data center cooling control via physics-guided safe reinforcement learning,” ACM Transactions on Cyber-Physical Systems , vol. 8, no. 2, pp. 1–26, May 2024

    2024

  30. [38]

    Tube-based model predictive control of energy storage systems for enhancing transient stability of power systems,

    I. Kiaei and S. Lotfifard, “Tube-based model predictive control of energy storage systems for enhancing transient stability of power systems,” IEEE Transactions on Smart Grid , vol. 9, no. 6, pp. 6438–6447, Nov. 2018

    2018

  31. [39]

    Mixed-stage energy management for decentralized microgrid cluster based on enhanced tube model predictive control,

    P. Xie, Y . Jia, H. Chen, J. Wu, and Z. Cai, “Mixed-stage energy management for decentralized microgrid cluster based on enhanced tube model predictive control,” IEEE Transactions on Smart Grid , vol. 12, no. 5, pp. 3780–3792, Sep. 2021

    2021

  32. [40]

    A chance- constraints-based control strategy for microgrids with energy storage and integrated electric vehicles,

    A. Ravichandran, S. Sirouspour, P. Malysz, and A. Emadi, “A chance- constraints-based control strategy for microgrids with energy storage and integrated electric vehicles,” IEEE Transactions on Smart Grid , vol. 9, no. 1, pp. 346–359, Jan. 2018

    2018

  33. [41]

    Risk-averse model predictive control design for battery energy storage systems,

    D. Rosewater, R. Baldick, and S. Santoso, “Risk-averse model predictive control design for battery energy storage systems,” IEEE Transactions on Smart Grid , vol. 11, no. 3, pp. 2014–2022, May 2020

    2014

  34. [42]

    Distributionally robust model predictive control for smart electric vehicle charging station with v2g/v2v capabil- ity,

    H. T. Nguyen and D.-H. Choi, “Distributionally robust model predictive control for smart electric vehicle charging station with v2g/v2v capabil- ity,” IEEE Transactions on Smart Grid , vol. 14, no. 6, pp. 4621–4633, Nov. 2023

    2023

  35. [43]

    A distribution- ally robust model predictive control for static and dynamic uncertainties in smart grids,

    Q. Li, Y . Shi, Y . Jiang, Y . Shi, H. Wang, and H. V . Poor, “A distribution- ally robust model predictive control for static and dynamic uncertainties in smart grids,” IEEE Transactions on Smart Grid , vol. 15, no. 5, pp. 4890–4902, Sep. 2024

    2024

  36. [44]

    Distributed robust model predictive control-based energy management strategy for islanded multi-microgrids considering uncertainty,

    Z. Zhao, J. Guo, X. Luo, C. S. Lai, P. Yang, L. L. Lai, P. Li, J. M. Guerrero, and M. Shahidehpour, “Distributed robust model predictive control-based energy management strategy for islanded multi-microgrids considering uncertainty,” IEEE Transactions on Smart Grid , vol. 13, ...

    2022

  37. [45]

    Simplified server model to simulate data center cooling energy consumption,

    S.-W. Ham, M.-H. Kim, B.-N. Choi, and J.-W. Jeong, “Simplified server model to simulate data center cooling energy consumption,” Energy and Buildings, vol. 86, pp. 328–339, Jan. 2015

    2015

  38. [46]

    Thermal-aware task scheduling for data centers through minimizing heat recirculation,

    Q. Tang, S. K. S. Gupta, and G. Varsamopoulos, “Thermal-aware task scheduling for data centers through minimizing heat recirculation,” in Proceedings of the 2007 IEEE International Conference on Cluster Computing, Austin, TX,, Sep. 2007, pp. 129–138

    2007

  39. [47]

    W. M. Kays and A. L. London, Compact Heat Exchangers, 3rd ed. New York, NY: McGraw-Hill, 1984

    1984

  40. [48]

    Development and validation of a new variable refrigerant flow system model in EnergyPlus,

    T. Hong, K. Sun, R. Zhang, R. Hinokuma, S. Kasahara, and Y . Yura, “Development and validation of a new variable refrigerant flow system model in EnergyPlus,” Energy and Buildings , vol. 117, pp. 399–411, Apr. 2016

    2016

  41. [49]

    Data-driven distributionally robust optimization using the Wasserstein metric: Performance guarantees and tractable reformulations,

    P. Mohajerin Esfahani and D. Kuhn, “Data-driven distributionally robust optimization using the Wasserstein metric: Performance guarantees and tractable reformulations,” Mathematical Programming, vol. 171, no. 1– 2, pp. 115–166, Sep. 2018

    2018

  42. [50]

    Global convergence of ADMM in nonconvex nonsmooth optimization,

    Y . Wang, W. Yin, and J. Zeng, “Global convergence of ADMM in nonconvex nonsmooth optimization,” Journal of Scientific Computing , vol. 78, no. 1, pp. 29–63, Jan. 2019

    2019

  43. [51]

    LMPs by Resource Nodes, Load Zones and Trading Hubs (NP6-788-CD),

    Electric Reliability Council of Texas (ERCOT), “LMPs by Resource Nodes, Load Zones and Trading Hubs (NP6-788-CD),” ERCOT Market Information System (MIS) Data Product Details, 2026

    2026

  44. [52]

    WattTime Data API (V3) Documentation,

    WattTime, “WattTime Data API (V3) Documentation,” Online documen- tation, 2026

    2026

  45. [53]

    SIGNAL: Average CO2 (AOER),

    ——, “SIGNAL: Average CO2 (AOER),” WattTime Data Signals doc- umentation, 2026

    2026

Pith tools

Reviewed July 2, 2026 · model on record in the stance chip above.