Pith. sign in

REVIEW 2 major objections 2 minor 1 cited by

Watts vs. Bytes: Turning Data Centers into Grid Assets via Storage Compute Co-Optimization

T0 review · 2 major / 2 minor · reviewed 2026-05-20 · grok-4.3

Pith's one-line read Co-optimizing data center workloads with battery storage meets grid limits and doubles BESS value under stress.

desk verdict Co-optimization of workloads, DVFS, and BESS under grid limits yields doubled BESS value in stressed case studies, but real-time execution costs remain unmodeled. read the letter →

arxiv 2605.16190 v2 pith:NGVAIFE5 submitted 2026-05-15 eess.SY cs.SYmath.OC

classification eess.SYcs.SYmath.OC
keywords datacentersbatteryenergystorageco-optimizationgridinterconnectionlimitsworkloadschedulingDVFSancillaryservicespeakloadmanagement
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 co-optimization framework for data centers with co-located battery energy storage systems under utility interconnection limits on peak load and ramping. It jointly manages deadline-constrained computing workloads through scheduling and dynamic voltage and frequency scaling together with degradation-aware BESS dispatch to reduce costs and enable ancillary service participation. Case studies using real-world market and workload data confirm that the framework produces feasible day-ahead schedules across operating conditions. Benefits increase substantially when interconnection constraints bind, with daily BESS value rising by a factor of two or more as storage actions reduce the risk of schedulable workload incompletion while respecting limits.

What carries the argument

The day-ahead co-optimization model that jointly optimizes deadline-constrained workload scheduling and DVFS with degradation-aware BESS dispatch subject to peak-load and ramping interconnection limits.

What would settle it

Field implementation of the optimized day-ahead schedules in an operating data center that measures actual workload completion rates, energy costs, grid compliance, and any performance or reliability deviations from model predictions.

Watch

Extended reading notes

Core claim

The authors establish that a robust day-ahead co-optimization model integrating workload scheduling, DVFS, and BESS dispatch yields feasible schedules that optimize operations and grid services, with BESS value increasing by a factor of two or more under stressed peak-load and ramping limits primarily by mitigating potential incompletion in schedulable workloads while complying with constraints; under baseline conditions value stems from ancillary participation and improved management, while tight peak caps make workload composition matter such that non-schedulable jobs raise costs by more than 25 percent and DVFS emerges as an additional flexibility lever.

Load-bearing premise

Workload scheduling and DVFS can be executed in real time with negligible unmodeled performance or reliability costs, and the chosen real-world market and workload traces represent conditions where interconnection limits actually bind.

Editorial extensions

If this is right

  • Feasible day-ahead schedules are produced across a range of operating conditions.
  • BESS daily value increases by a factor of two or more when peak-load and ramping constraints bind.
  • Under tight peak-load caps a higher share of non-schedulable jobs raises operating cost by more than 25 percent relative to flexible mixes.
  • DVFS serves as a material flexibility lever when load limits are tight.
  • Coordinated compute-storage flexibility expands operational headroom and grid value of data centers.

Reading between the lines

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

  • Data centers could more readily locate in regions with constrained grid capacity if the approach is adopted.
  • The same co-optimization logic could extend to other flexible loads paired with storage such as EV charging hubs.
  • Real-time adjustments beyond the day-ahead horizon might yield further gains provided performance costs stay low.
  • Grid operators could treat data centers as more reliable providers of flexibility services.
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

2 major / 2 minor

Summary. The manuscript develops a robust co-optimization framework for day-ahead operation of data centers with co-located BESS under utility-imposed interconnection limits on peak load and ramping. It jointly optimizes deadline-constrained workloads through scheduling and DVFS together with degradation-aware BESS dispatch to minimize operating costs while enabling ancillary-service participation. Case studies using real-world market and workload data report feasible schedules, with BESS value rising by a factor of two or more under stressed peak-load and ramping conditions (driven by reduced schedulable-workload incompletion) and position DVFS as a material flexibility lever under tight limits.

Significance. If the central modeling assumptions hold, the work demonstrates that compute-storage co-optimization can materially expand data-center operational headroom and grid value precisely when interconnection capacity is scarcest. The use of real-world traces strengthens practical relevance and the finding that benefits amplify under binding constraints offers a concrete, falsifiable prediction for operators. These elements would be notable contributions to the literature on flexible demand and storage co-location if the quantitative claims survive additional validation.

major comments (2)
  1. [Abstract and Case Studies] Abstract and Case Studies section: The claim that daily BESS value increases by a factor of two or more under stressed peak-load and ramping limits rests on the assumption that jointly optimized workload scheduling and DVFS decisions can be executed in real time with negligible unmodeled performance, delay, or reliability penalties. When interconnection constraints bind, the model relies on these levers to create headroom; any real-time cost that increases incompletion or forces extra BESS cycling would directly erode the reported benefit and the assertion that processor-level control is a material flexibility lever.
  2. [Case Studies] Case Studies section: The quantitative results (feasible schedules, factor-of-two value increase, >25% cost impact from workload composition) are presented without error bars, sensitivity tables on key parameters such as DVFS cost coefficients or workload flexibility ratios, or explicit out-of-sample validation. This makes it difficult to assess whether the reported benefits are robust to the modeling choices that underpin the central claim.
minor comments (2)
  1. [Abstract] Abstract: The text contains visible LaTeX revision commands (e.g., “revise{by BESS actions…}” and “revise{Additionally, DVFS studies…}”) that should be removed before final submission.
  2. [Model formulation] Model formulation: Notation distinguishing schedulable versus non-schedulable workloads and the precise definition of interconnection ramping limits could be introduced earlier and used consistently to improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. The comments highlight important considerations regarding model assumptions and result robustness. We address each major comment below and outline the revisions we will make to strengthen the manuscript.

read point-by-point responses
  1. Referee: [Abstract and Case Studies] Abstract and Case Studies section: The claim that daily BESS value increases by a factor of two or more under stressed peak-load and ramping limits rests on the assumption that jointly optimized workload scheduling and DVFS decisions can be executed in real time with negligible unmodeled performance, delay, or reliability penalties. When interconnection constraints bind, the model relies on these levers to create headroom; any real-time cost that increases incompletion or forces extra BESS cycling would directly erode the reported benefit and the assertion that processor-level control is a material flexibility lever.

    Authors: We agree that the day-ahead co-optimization framework implicitly assumes that the computed workload schedules and DVFS decisions can be implemented in real time without significant unmodeled penalties or reliability impacts. The model focuses on generating feasible day-ahead schedules under interconnection constraints, with BESS actions explicitly reducing potential workload incompletion. Real-time execution dynamics and associated costs lie outside the current scope. In the revised manuscript we will add an explicit discussion of this modeling assumption and its implications in the Case Studies section, and we have updated the abstract to qualify the reported BESS value increase under the modeled conditions. revision: partial

  2. Referee: [Case Studies] Case Studies section: The quantitative results (feasible schedules, factor-of-two value increase, >25% cost impact from workload composition) are presented without error bars, sensitivity tables on key parameters such as DVFS cost coefficients or workload flexibility ratios, or explicit out-of-sample validation. This makes it difficult to assess whether the reported benefits are robust to the modeling choices that underpin the central claim.

    Authors: We concur that additional sensitivity analysis would improve assessment of robustness. In the revised Case Studies section we will include sensitivity tables for the DVFS cost coefficients and workload flexibility ratios. The presented results derive from deterministic optimization using real-world traces; we will add a short discussion noting the deterministic formulation and identifying stochastic extensions or out-of-sample testing as valuable future directions. These changes will better support the quantitative claims. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

Forward optimization model with external data traces shows no circularity

full rationale

The paper develops a co-optimization framework for day-ahead data-center scheduling that jointly optimizes workload scheduling, DVFS, and BESS dispatch subject to interconnection limits. Case-study results are generated by applying this model to real-world market and workload traces; the reported benefits (including the factor-of-two BESS value increase under binding constraints) are simulation outputs rather than algebraic identities or parameters fitted inside the same equations. No self-definitional steps, fitted-input predictions, or load-bearing self-citations appear in the derivation chain. The framework remains self-contained against external benchmarks and does not reduce its central claims to its own inputs by construction.

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

Abstract-only review supplies no explicit free parameters, axioms, or invented entities; standard assumptions of convex optimization and representative historical traces are implicit but not enumerated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Watts vs. Bytes: Turning Data Centers into Grid Assets via Storage Compute Co-Optimization." pith.science (2026). https://pith.science/paper/NGVAIFE5

@misc{pith2026260516190,
  author       = {Pith},
  title        = {Pith review of: Watts vs. Bytes: Turning Data Centers into Grid Assets via Storage Compute Co-Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NGVAIFE5}},
  note         = {Machine review of arXiv:2605.16190}
}
read the original abstract

Data center interconnections increasingly face tighter peak-demand and ramp-rate limits while being expected to support grid operations. Satisfying these requirements calls for coordinated computing and energy controls, yet their joint operational and economic implications remain poorly understood. To tackle this problem, we formulate a robust day-ahead co-optimization of computing load scheduling, server dynamic voltage and frequency scaling (DVFS), and co-located battery energy storage system (BESS) dispatch. The resulting mixed-integer linear program hedges against uncertainty in fixed load and ancillary service deployment while enforcing interconnection limits on peak demand and ramp rate, ancillary service capacity commitments in reserve and flexible ramping, and workload execution constraints. Case studies using CAISO and PJM market data of a 100~MW data center with a 36~MWh/12~MW BESS show that workload scheduling, DVFS, and storage provide complementary flexibility. Under binding peak-load limits, increasing the schedulable workload share reduces mean daily operating cost by up to 20.7\%, and the daily value of storage more than doubles relative to operation under less restrictive limits. Under normal conditions, optimal BESS sizing is driven more by capital cost and cycling allowance than by energy duration alone. An 8~MW aggregate ancillary service commitment increases operational cost by only 0.4\%, whereas reserve-only requirements become infeasible at commitments as small as 4~MW. These findings show that coordinated computing and storage controls can support grid services economically under binding interconnection constraints while protecting workload delivery.

Figures

Figures reproduced from arXiv: 2605.16190 by the authors.

Figure 1
Figure 1. Overview of the robust day-ahead co-optimization framework for a data center with co-located BESS. The coordinator jointly schedules flexible compute [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Representative daily dispatch for a 100 MW data-center case with a 36 MWh/12 MW co-located BESS under hard interconnection constraints. The left panel shows the fixed load, internal data-center load with storage, and grid-facing net load with and without the co-located BESS. The right panel shows the robust SOC trajectory and scenario spread, together with charge/discharge and ancillary-service power components. The… view at source ↗
Figure 3
Figure 3. Battery sizing results for the 100 MW data center with real PJM [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Optimal installed BESS energy capacity versus cycling limit for discount rates r=7% and r=10%. Megapack 3 remains at or near the search boundary, while Megapack 2XL is more sensitive to cycling and discount-rate assumptions because its higher power density is paired wi…
Figure 5
Figure 5. Figure 5: Interconnection sensitivity under alternative load compositions over 31 daily instances. [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Hourly LP shadow prices for the base interconnection case over 31 [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Battery value under stressed interconnection limits over 31 daily instances. The 36 MWh/12 MW BESS has modest value under loose interconnection [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Multiday discrete-DVFS case study under the 100 MW load-limit configuration. The cases compare no DVFS with 3-level ±5%, 5-level ±10%, 7-level ±15%, and 9-level ±20% operating sets. Overall, allowing DVFS improves the mean daily operating value and reduces day-to-day v…
Figure 9
Figure 9. Figure 9: Multiday fixed-load and schedulable-demand ablation. In panel (a), scaling the fixed-load trace changes the residual headroom below the load cap; in [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Multiday out-of-sample robustness. The first-stage schedule from each 50-scenario training problem is evaluated on 500 fresh scenarios per day. [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]

Discussion (0). Continue with ORCID to comment.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

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. PowerAtlas: Towards Electricity-Computing Co-Scheduling for Power Systems

    cs.LG 2026-07 conditional novelty 6.0 of 10

    SFT plus feasibility-aware GRPO lets 3B LLMs produce electricity–computing co-schedules that are far more grid-feasible and cheaper than untrained or frontier zero-shot baselines on ECBench.

Pith tools

Reviewed May 20, 2026 · model on record in the stance chip above.