View in EDS HTML Full Text PDF Full Text

Bridging the Information Gap: A Mechanism Design Approach to Forecasting AI's Power Grid Load.

Saved in:
Bibliographic Details
Title: Bridging the Information Gap: A Mechanism Design Approach to Forecasting AI's Power Grid Load.
Authors: Cai, Xinlei1 (AUTHOR), Chen, Kexin2 (AUTHOR), Jiang, Lizhou1 (AUTHOR), Xu, Ruichen2 (AUTHOR) xuruichen@cuhk.edu.cn, Dong, Kai1 (AUTHOR), Meng, Zijie1 (AUTHOR)
Source: Energies (19961073). Jun2026, Vol. 19 Issue 11, p2553. 19p.
Subject Terms: *Information asymmetry, *Data centers, *Electric power consumption, *Energy consumption forecasting, *Language models, *Load forecasting (Electric power systems), *Incentive (Psychology), *Game theory
Abstract: The rapid proliferation of Large Language Models (LLMs) is increasing electricity demand from data centers, creating new challenges for power-demand forecasting and grid planning. A key difficulty is that architecture- and deployment-related information that affects inference load is often private to LLM providers. This paper proposes a two-stage, mechanism-assisted forecasting framework under information asymmetry. In the first stage, a stylized incentive mechanism elicits verifiable reduced-form demand parameters from LLM providers at a chosen reporting precision. In the second stage, the elicited parameters are incorporated into forecasting models as architecture- and deployment-informed features. Using calibrated synthetic scenarios constructed from public data-center energy reports, open LLM-inference energy benchmarks, and secondary public estimates, we find that incorporating elicited parameters reduces the mean squared error (MSE) of the ResNet forecasting backbone by 65.1% relative to an architecture-agnostic ResNet baseline. Similar improvements are observed for a gradient-boosting model, indicating that the main empirical value comes from procuring informative provider-side demand features rather than from a specific neural architecture. The results should be interpreted as a proof-of-concept demonstration rather than a full operational model of LLM serving or power-system dispatch. [ABSTRACT FROM AUTHOR]
Database: Energy & Power Source
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
Text:
  Availability: 1
Header DbId: enr
DbLabel: Energy & Power Source
An: 194587941
AccessLevel: 6
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Bridging the Information Gap: A Mechanism Design Approach to Forecasting AI's Power Grid Load.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Cai%2C+Xinlei%22">Cai, Xinlei</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Chen%2C+Kexin%22">Chen, Kexin</searchLink><relatesTo>2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Jiang%2C+Lizhou%22">Jiang, Lizhou</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Xu%2C+Ruichen%22">Xu, Ruichen</searchLink><relatesTo>2</relatesTo> (AUTHOR)<i> xuruichen@cuhk.edu.cn</i><br /><searchLink fieldCode="AR" term="%22Dong%2C+Kai%22">Dong, Kai</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Meng%2C+Zijie%22">Meng, Zijie</searchLink><relatesTo>1</relatesTo> (AUTHOR)
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%22Energies+%2819961073%29%22">Energies (19961073)</searchLink>. Jun2026, Vol. 19 Issue 11, p2553. 19p.
– Name: Subject
  Label: Subject Terms
  Group: Su
  Data: *<searchLink fieldCode="DE" term="%22Information+asymmetry%22">Information asymmetry</searchLink><br />*<searchLink fieldCode="DE" term="%22Data+centers%22">Data centers</searchLink><br />*<searchLink fieldCode="DE" term="%22Electric+power+consumption%22">Electric power consumption</searchLink><br />*<searchLink fieldCode="DE" term="%22Energy+consumption+forecasting%22">Energy consumption forecasting</searchLink><br />*<searchLink fieldCode="DE" term="%22Language+models%22">Language models</searchLink><br />*<searchLink fieldCode="DE" term="%22Load+forecasting+%28Electric+power+systems%29%22">Load forecasting (Electric power systems)</searchLink><br />*<searchLink fieldCode="DE" term="%22Incentive+%28Psychology%29%22">Incentive (Psychology)</searchLink><br />*<searchLink fieldCode="DE" term="%22Game+theory%22">Game theory</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: The rapid proliferation of Large Language Models (LLMs) is increasing electricity demand from data centers, creating new challenges for power-demand forecasting and grid planning. A key difficulty is that architecture- and deployment-related information that affects inference load is often private to LLM providers. This paper proposes a two-stage, mechanism-assisted forecasting framework under information asymmetry. In the first stage, a stylized incentive mechanism elicits verifiable reduced-form demand parameters from LLM providers at a chosen reporting precision. In the second stage, the elicited parameters are incorporated into forecasting models as architecture- and deployment-informed features. Using calibrated synthetic scenarios constructed from public data-center energy reports, open LLM-inference energy benchmarks, and secondary public estimates, we find that incorporating elicited parameters reduces the mean squared error (MSE) of the ResNet forecasting backbone by 65.1% relative to an architecture-agnostic ResNet baseline. Similar improvements are observed for a gradient-boosting model, indicating that the main empirical value comes from procuring informative provider-side demand features rather than from a specific neural architecture. The results should be interpreted as a proof-of-concept demonstration rather than a full operational model of LLM serving or power-system dispatch. [ABSTRACT FROM AUTHOR]
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=enr&AN=194587941
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.3390/en19112553
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 19
        StartPage: 2553
    Subjects:
      – SubjectFull: Information asymmetry
        Type: general
      – SubjectFull: Data centers
        Type: general
      – SubjectFull: Electric power consumption
        Type: general
      – SubjectFull: Energy consumption forecasting
        Type: general
      – SubjectFull: Language models
        Type: general
      – SubjectFull: Load forecasting (Electric power systems)
        Type: general
      – SubjectFull: Incentive (Psychology)
        Type: general
      – SubjectFull: Game theory
        Type: general
    Titles:
      – TitleFull: Bridging the Information Gap: A Mechanism Design Approach to Forecasting AI's Power Grid Load.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Cai, Xinlei
      – PersonEntity:
          Name:
            NameFull: Chen, Kexin
      – PersonEntity:
          Name:
            NameFull: Jiang, Lizhou
      – PersonEntity:
          Name:
            NameFull: Xu, Ruichen
      – PersonEntity:
          Name:
            NameFull: Dong, Kai
      – PersonEntity:
          Name:
            NameFull: Meng, Zijie
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 06
              Text: Jun2026
              Type: published
              Y: 2026
          Identifiers:
            – Type: issn-print
              Value: 19961073
          Numbering:
            – Type: volume
              Value: 19
            – Type: issue
              Value: 11
          Titles:
            – TitleFull: Energies (19961073)
              Type: main
ResultId 1