pith. sign in

arxiv: 2605.17723 · v1 · pith:GMZZWFKEnew · submitted 2026-05-18 · 📡 eess.SY · cs.SY· econ.GN· q-fin.EC

Residential Battery Pooling Under Backup Commitments

Jerry Anunrojwong , Baosen Zhang This is my paper

Pith reviewed 2026-05-19 22:04 UTC · model grok-4.3

classification 📡 eess.SY cs.SYecon.GNq-fin.EC
keywords residential batteriesbattery poolingbackup commitmentsmodel predictive controlfirm marginERCOTarbitrageoutage protection
0
0 comments X

The pith

Pooling batteries across homes still improves margins even when each keeps its own backup reserve, though gains shrink as backup obligations grow stricter.

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

The paper compares independent control of each residential battery against coordinated pooling while enforcing per-home state of charge and a household-specific backup energy requirement for outages. Using model predictive control on 15-minute intervals with real household data and ERCOT prices, it evaluates the 543 homes able to support backup on their own across caps from two to twenty-four hours. Pooling raises weekly firm margins above the standalone level, but the added value falls smoothly from 1.49 to 1.27 dollars per home as the cap lengthens. This matters because many providers want to sell both market services and backup protection; the results show coordination remains helpful without fully relaxing individual guarantees.

Core claim

When dispatch is coordinated across homes yet each battery retains its own state of charge and must satisfy its own backup commitment, the pooled model predictive controller still achieves higher firm margins than standalone independent control. The pooling benefit declines from 1.49 to 1.27 dollars per home per week as the backup cap rises from two to twenty-four hours, amounting to 13.5 percent down to 11.8 percent of the standalone margin which itself stays near 11 dollars per home per week.

What carries the argument

Model predictive control optimization that maximizes expected market revenue while enforcing individual battery dynamics and household-specific backup energy reserves.

If this is right

  • Providers can still capture coordination gains while offering household-level backup products.
  • The relative benefit of pooling is larger when backup caps are shorter.
  • Standalone margins remain stable near eleven dollars per home per week across the cap range.
  • Coordination value declines smoothly rather than disappearing at longer backup durations.

Where Pith is reading between the lines

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

  • The same per-home constraint approach could be tested on other shared resources such as vehicle-to-grid fleets with minimum range guarantees.
  • Relaxing the perfect-forecast assumption inside the MPC might reveal how robust the pooling advantage remains under price or load uncertainty.
  • If scaled, the modest but persistent benefit could influence contract design for combined market-service and backup offerings.

Load-bearing premise

The 543 homes already able to meet backup needs on their own, together with the two-to-twenty-four-hour caps, represent the relevant range for assessing pooling value under service constraints.

What would settle it

Applying the same controller to homes that cannot meet backup needs independently, or measuring real outage events where pooled homes lose backup power more often than the model predicts.

Figures

Figures reproduced from arXiv: 2605.17723 by Baosen Zhang, Jerry Anunrojwong.

Figure 1
Figure 1. Figure 1: ERCOT load zones. The empirical analysis uses ERCOT real-time prices at settlement [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Base-week net load profiles for four representative homes. Net load is defined as household [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Total pooled state-of-charge trajectories under the backup-cap experiment. Each curve [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
read the original abstract

Residential batteries increasingly serve two roles: they can earn money by arbitraging wholesale prices and providing grid services, and they provide backup power during outages. This dual use creates a basic tradeoff between earning market value and preserving outage readiness. Coordination across many batteries can help, but a provider cannot treat the fleet as a single virtual battery when each household is promised its own backup protection. We compare standalone control, in which each home is dispatched independently, with pooling, in which homes are coordinated while each battery retains its own state of charge and household-specific backup requirement. Both regimes are implemented as model predictive control problems with 15-minute decision intervals and evaluated using household telemetry together with ERCOT market inputs. The empirical design focuses on the 543 homes in our sample that can support at least one backup product in standalone operation and studies backup caps ranging from 2 to 24 hours. Lower caps relax backup obligations, while the 24-hour cap coincides with assigning each home its own longest feasible backup tier. Pooling remains beneficial in this service-constrained setting, but its value declines smoothly as backup obligations tighten. Standalone firm margin ranges from \$11.06 per home per week at the 2-hour cap to \$10.79 at the 24-hour cap, while pooling benefit falls from \$1.49 to \$1.27 per home per week. Relative to standalone firm margin, pooling is worth about 13.5% at the 2-hour cap and about 11.8% at the 24-hour cap. Coordination therefore still helps after preserving household-level backup guarantees, but its value declines as backup obligations tighten.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The paper claims that pooling residential batteries under household-specific backup commitments remains beneficial compared to standalone control, even as backup obligations tighten. Using model predictive control with 15-minute intervals on telemetry from 543 homes (selected as those able to support at least one backup product standalone) and ERCOT market data, standalone firm margins range from $11.06 to $10.79 per home per week across 2- to 24-hour caps, while pooling adds $1.49 to $1.27 (13.5% down to 11.8% relative benefit).

Significance. If the results hold, they demonstrate that coordination value persists under realistic per-home backup constraints, which is relevant for virtual power plant and residential storage service design. The forward simulation on independent price and load data provides a non-circular empirical basis, and the smooth decline in pooling benefit with increasing caps is a clear, policy-relevant pattern.

major comments (1)
  1. [Empirical Design (sample selection described in abstract and methods)] The headline quantitative claims (pooling benefit of $1.49–$1.27 per home per week and 13.5–11.8% relative to standalone) rest exclusively on the 543-home subset that can already support standalone backup. The manuscript reports no results on the unfiltered population and no sensitivity check that re-includes homes unable to meet standalone requirements. This filtering choice is load-bearing for the reported percentages and the smooth-decline pattern, because pooling could enable backup services or produce different marginal value once the constraint is relaxed to fleet level.
minor comments (1)
  1. [Abstract] The abstract could briefly state the size of the original dataset before the 543-home filter to clarify the selection fraction and its potential impact on representativeness.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the major comment on empirical design below and will revise the text to strengthen the justification for our sample selection.

read point-by-point responses

  1. Referee: The headline quantitative claims (pooling benefit of $1.49–$1.27 per home per week and 13.5–11.8% relative to standalone) rest exclusively on the 543-home subset that can already support standalone backup. The manuscript reports no results on the unfiltered population and no sensitivity check that re-includes homes unable to meet standalone requirements. This filtering choice is load-bearing for the reported percentages and the smooth-decline pattern, because pooling could enable backup services or produce different marginal value once the constraint is relaxed to fleet level.

    Authors: We appreciate the referee's observation on the sample selection. The 543-home subset consists of those homes in our telemetry data that can support at least one backup product under standalone operation, as stated in the abstract and methods. This choice is intentional: it permits a direct comparison of standalone versus pooled control while holding the household-specific backup commitments constant across both regimes. Homes unable to meet standalone requirements would receive no backup service under independent control, so including them would compare different service levels rather than isolating the incremental value of coordination under per-household constraints—the central question of the paper. We will revise the methods and discussion sections to state this rationale more explicitly and to note the implications for generalizability to the broader population. revision: yes

Circularity Check

0 steps flagged

No circularity: results from direct forward simulation on external data

full rationale

The paper evaluates standalone versus pooled battery control via model predictive control applied to real household telemetry and ERCOT price series. All reported margins and pooling benefits (e.g., $1.49 to $1.27 per home per week) are computed outputs of these simulations rather than quantities obtained by fitting parameters to the target statistics or by reducing equations to self-referential definitions. No self-citations, uniqueness theorems, or ansatzes are invoked to justify the central quantitative claims; the derivation chain consists of standard optimization and data-driven evaluation that remains independent of the reported percentages.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work relies on standard model predictive control formulations and external market/household data inputs without introducing new free parameters, axioms beyond domain-standard assumptions, or invented entities.

axioms (1)
  • domain assumption Model predictive control with 15-minute decision intervals accurately represents optimal battery dispatch under price and load uncertainty for both standalone and pooled regimes.
    Invoked in the implementation and evaluation of both control regimes described in the abstract.

pith-pipeline@v0.9.0 · 5833 in / 1366 out tokens · 37807 ms · 2026-05-19T22:04:47.368109+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

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.

Reference graph

Works this paper leans on

54 extracted references · 54 canonical work pages

  1. [1]

    and Y\"ucel,

    Agrawal, Vishal V. and Y\"ucel,. Management Science , volume =

    work page

  2. [2]

    Aguiar, Nayara and Dubey, Anamika and Gupta, Vijay , journal =

    work page

  3. [3]

    and Besbes, Omar and Xu, Bolun , journal =

    Anunrojwong, Jerry and Balseiro, Santiago R. and Besbes, Omar and Xu, Bolun , journal =

    work page

  4. [4]

    Bae, Joonho and Kapuscinski, Roman and Silberholz, John , journal =

    work page

  5. [5]

    Beck, Amir , publisher =

    work page

  6. [6]

    , journal =

    Ben-Ayed, Omar and Blair, Charles E. , journal =. 1990 , month = jun, publisher =

    work page 1990

  7. [7]

    Ben-Tal, Aharon and Nemirovski, Arkadi , journal =

    work page

  8. [8]

    Bertsimas, Dimitris and Sim, Melvyn , journal =

    work page

  9. [9]

    Borenstein, Severin , journal =

    work page

  10. [10]

    2007 , month = sep, publisher =

    Colson, Benoit and Marcotte, Patrice and Savard, Gilles , journal =. 2007 , month = sep, publisher =

    work page 2007

  11. [11]

    Cruise, James and Flatley, Lisa and Gibbens, Richard and Zachary, Stan , journal =

    work page

  12. [12]

    2018 , editor =

    Franceschi, Luca and Frasconi, Paolo and Salzo, Saverio and Grazzi, Riccardo and Pontil, Massimiliano , booktitle =. 2018 , editor =

    work page 2018

  13. [13]

    , journal =

    Gao, Zuguang and Alshehri, Khaled and Birge, John R. , journal =

    work page

  14. [14]

    and Baik, Sunhee and Miller, Cesca and Carvallo, Juan Pablo , journal =

    Gorman, Will and Barbose, Galen L. and Baik, Sunhee and Miller, Cesca and Carvallo, Juan Pablo , journal =

    work page

  15. [15]

    , journal =

    Jeroslow, Robert G. , journal =. 1985 , publisher =

    work page 1985

  16. [16]

    2021 , editor =

    Ji, Kaiyi and Yang, Junjie and Liang, Yingbin , booktitle =. 2021 , editor =

    work page 2021

  17. [17]

    and Powell, Warren B

    Jiang, Daniel R. and Powell, Warren B. , journal =

    work page

  18. [18]

    Kaps, Christian and Netessine, Serguei , journal =

    work page

  19. [19]

    Khodaei, Amin , journal =

    work page

  20. [20]

    , journal =

    Kim, Jae Ho and Powell, Warren B. , journal =

    work page

  21. [21]

    and Anderson, Kate and DiOrio, Nicholas A

    Laws, Nicholas D. and Anderson, Kate and DiOrio, Nicholas A. and Li, Xiangkun and McLaren, Joyce , journal =

    work page

  22. [22]

    Liu, Guodong and Starke, Michael and Xiao, Bailu and Tomsovic, Kevin , journal =

    work page

  23. [23]

    Marqusee, Jeffrey and Becker, William and Ericson, Sean , journal =

    work page

  24. [24]

    Mohsenian-Rad, Amir-Hamed and Leon-Garcia, Alberto , journal =

    work page

  25. [25]

    2004 , series =

    Nesterov, Yurii , publisher =. 2004 , series =

    work page 2004

  26. [26]

    Nottrott, Anders and Kleissl, Jan and Washom, Byron , journal =

    work page

  27. [27]

    and Tan, Burcu and Kazan, Osman , journal =

    Parker, Geoffrey G. and Tan, Burcu and Kazan, Osman , journal =

    work page

  28. [28]

    2016 , editor =

    Pedregosa, Fabian , booktitle =. 2016 , editor =

    work page 2016

  29. [29]

    and Mitter, Sanjoy K

    Roozbehani, Mardavij and Dahleh, Munther A. and Mitter, Sanjoy K. , journal =

    work page

  30. [30]

    Secomandi, Nicola , journal =

    work page

  31. [31]

    , journal =

    Sunar, Nur and Swaminathan, Jayashankar M. , journal =

    work page

  32. [32]

    and Wang, Derek D

    Wu, Owen Q. and Wang, Derek D. and Qin, Zhenwei , journal =

    work page

  33. [33]

    Zhou, Yangfang (Helen) and Scheller-Wolf, Alan and Secomandi, Nicola and Smith, Stephen , journal =

    work page

  34. [34]

    Liu, Guodong and Ollis, Ben and Stenvig, Nils and Xu, Yao and Zhang, Yichen and Tomsovic, Kevin , booktitle =

    work page

  35. [35]

    and Byrne, Raymond H

    Nguyen, Tu A. and Byrne, Raymond H. , booktitle =

    work page

  36. [36]

    Mirletz, Brian , year =

    work page

  37. [37]

    Mirletz, Brian and Guittet, Darice , booktitle =

    work page

  38. [38]

    2024 , month = oct, url =

    Gorman, Will and Barbose, Galen and Baik, Sunhee and Miller, Cesca and Carvallo, Juan Pablo , institution =. 2024 , month = oct, url =

    work page 2024

  39. [39]

    2025 , month = may, note =

    work page 2025

  40. [40]

    , booktitle =

    Mather, Barry A. , booktitle =

    work page

  41. [41]

    2024 , month = may, note =

    work page 2024

  42. [42]

    2024 , month = jun, note =

    work page 2024

  43. [43]

    2025 , month = mar, note =

    work page 2025

  44. [44]

    2024 , month = feb, note =

    work page 2024

  45. [45]

    Operations Research , volume =

    L\". Operations Research , volume =

    work page

  46. [46]

    and Kapuscinski, Roman and Suresh, Santhosh , journal =

    Wu, Owen Q. and Kapuscinski, Roman and Suresh, Santhosh , journal =

    work page

  47. [47]

    Wu, Hongyu and Pratt, Annabelle and Munankarmi, Prateek and Lunacek, Monte and Balamurugan, Sivasathya Pradha and Liu, Xuebo and Spitsen, Paul , journal =

    work page

  48. [48]

    , journal =

    Sun, Chao and Sun, Fengchun and Moura, Scott J. , journal =

    work page

  49. [49]

    Cai, Jie and Zhang, Hao and Jin, Xing , journal =

    work page

  50. [50]

    Zou, Bin and Peng, Jinqing and Yin, Rongxin and Luo, Zhengyi and Song, Jiaming and Ma, Tao and Li, Sihui and Yang, Hongxing , journal =

    work page

  51. [51]

    and Subkhankulova, Dina , journal =

    Zakeri, Behnam and Castagneto Gissey, Giorgio and Dodds, Paul E. and Subkhankulova, Dina , journal =

    work page

  52. [52]

    , journal =

    Parra, David and Patel, Martin K. , journal =

    work page

  53. [53]

    Applied Energy , volume =

    Tobajas, Javier and Garcia-Torres, Felix and Roncero-S. Applied Energy , volume =

    work page

  54. [54]

    Engineering Applications of Artificial Intelligence , volume =

    Cai, Wenqi and Kordabad, Arash Bahari and Gros, S. Engineering Applications of Artificial Intelligence , volume =

    work page