View in EDS HTML Full Text PDF Full Text

XGBoost-Based 52-Week Peak-Load Forecasting Model with Monthly Adaptive Training and Sequential Prediction.

Saved in:
Bibliographic Details
Title: XGBoost-Based 52-Week Peak-Load Forecasting Model with Monthly Adaptive Training and Sequential Prediction.
Authors: Kim, Kyeong-Hwan1 (AUTHOR), Kim, Tae-Geun2 (AUTHOR), Kwon, Bo-Sung3 (AUTHOR), Song, Kyung-Bin4 (AUTHOR) kbsong@ssu.ac.kr
Source: Energies (19961073). Jun2026, Vol. 19 Issue 11, p2683. 27p.
Subject Terms: *Load forecasting (Electric power systems), *Boosting algorithms, *Time, *Demand forecasting, *Photovoltaic power generation, *Feature selection, *Electric power production forecasting
Abstract: The operational resilience and strategic infrastructure planning of modern power grids are fundamentally anchored in the precision of mid-term load forecasting (MTLF). However, accurate forecasting over a 52-week horizon is increasingly challenging due to the growing variability in electricity demand driven by extreme weather events and the expansion of behind-the-meter (BTM) photovoltaic generation. In response to these difficulties, a 52-week forecasting framework for weekly peak load is established in this study, leveraging the Extreme Gradient Boosting (XGBoost) algorithm. The primary contribution of this study lies not in the architectural modification of the XGBoost algorithm itself, but in the systematic integration of (i) a reproducible feature-screening protocol, (ii) month-specific training-set construction, and (iii) a sequential rolling prediction architecture validated under both actual- and forecast-input conditions. The proposed framework was validated using data from the Korean power system (2020–2024) and compared with a Long Short-Term Memory (LSTM) benchmark. In the Actual-Input scenario, the average Mean Absolute Percentage Error (MAPE) for the proposed framework was 2.90%, demonstrating superior precision over the LSTM model, which exhibited a 3.73%. Under the Forecast-Input scenario, the framework maintained high robustness with an average MAPE of 3.86%. These results demonstrate that the integrated framework-level approach provides a more practical and stable solution for mid-term power system operations than individual baseline models within the studied context. [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: 194588071
AccessLevel: 6
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: XGBoost-Based 52-Week Peak-Load Forecasting Model with Monthly Adaptive Training and Sequential Prediction.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Kim%2C+Kyeong-Hwan%22">Kim, Kyeong-Hwan</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Kim%2C+Tae-Geun%22">Kim, Tae-Geun</searchLink><relatesTo>2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Kwon%2C+Bo-Sung%22">Kwon, Bo-Sung</searchLink><relatesTo>3</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Song%2C+Kyung-Bin%22">Song, Kyung-Bin</searchLink><relatesTo>4</relatesTo> (AUTHOR)<i> kbsong@ssu.ac.kr</i>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%22Energies+%2819961073%29%22">Energies (19961073)</searchLink>. Jun2026, Vol. 19 Issue 11, p2683. 27p.
– Name: Subject
  Label: Subject Terms
  Group: Su
  Data: *<searchLink fieldCode="DE" term="%22Load+forecasting+%28Electric+power+systems%29%22">Load forecasting (Electric power systems)</searchLink><br />*<searchLink fieldCode="DE" term="%22Boosting+algorithms%22">Boosting algorithms</searchLink><br />*<searchLink fieldCode="DE" term="%22Time%22">Time</searchLink><br />*<searchLink fieldCode="DE" term="%22Demand+forecasting%22">Demand forecasting</searchLink><br />*<searchLink fieldCode="DE" term="%22Photovoltaic+power+generation%22">Photovoltaic power generation</searchLink><br />*<searchLink fieldCode="DE" term="%22Feature+selection%22">Feature selection</searchLink><br />*<searchLink fieldCode="DE" term="%22Electric+power+production+forecasting%22">Electric power production forecasting</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: The operational resilience and strategic infrastructure planning of modern power grids are fundamentally anchored in the precision of mid-term load forecasting (MTLF). However, accurate forecasting over a 52-week horizon is increasingly challenging due to the growing variability in electricity demand driven by extreme weather events and the expansion of behind-the-meter (BTM) photovoltaic generation. In response to these difficulties, a 52-week forecasting framework for weekly peak load is established in this study, leveraging the Extreme Gradient Boosting (XGBoost) algorithm. The primary contribution of this study lies not in the architectural modification of the XGBoost algorithm itself, but in the systematic integration of (i) a reproducible feature-screening protocol, (ii) month-specific training-set construction, and (iii) a sequential rolling prediction architecture validated under both actual- and forecast-input conditions. The proposed framework was validated using data from the Korean power system (2020–2024) and compared with a Long Short-Term Memory (LSTM) benchmark. In the Actual-Input scenario, the average Mean Absolute Percentage Error (MAPE) for the proposed framework was 2.90%, demonstrating superior precision over the LSTM model, which exhibited a 3.73%. Under the Forecast-Input scenario, the framework maintained high robustness with an average MAPE of 3.86%. These results demonstrate that the integrated framework-level approach provides a more practical and stable solution for mid-term power system operations than individual baseline models within the studied context. [ABSTRACT FROM AUTHOR]
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=enr&AN=194588071
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.3390/en19112683
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 27
        StartPage: 2683
    Subjects:
      – SubjectFull: Load forecasting (Electric power systems)
        Type: general
      – SubjectFull: Boosting algorithms
        Type: general
      – SubjectFull: Time
        Type: general
      – SubjectFull: Demand forecasting
        Type: general
      – SubjectFull: Photovoltaic power generation
        Type: general
      – SubjectFull: Feature selection
        Type: general
      – SubjectFull: Electric power production forecasting
        Type: general
    Titles:
      – TitleFull: XGBoost-Based 52-Week Peak-Load Forecasting Model with Monthly Adaptive Training and Sequential Prediction.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Kim, Kyeong-Hwan
      – PersonEntity:
          Name:
            NameFull: Kim, Tae-Geun
      – PersonEntity:
          Name:
            NameFull: Kwon, Bo-Sung
      – PersonEntity:
          Name:
            NameFull: Song, Kyung-Bin
    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