配電ネットワークにおける二次変電所の短期負荷予測
Short-Term Load Forecasting for Secondary Substations in Electrical Distribution Networks (原題)
Andreotti D, Spiller M, Rancilio G, Merlo M
🤖 gxceed AI 要約
日本語
DERと脱炭素化に伴う新規電力需要の増加を受け、配電系統運用者（DSO）には中低圧網の能動的管理が求められている。本論文は、二次変電所（SS）レベルの予測と接続点の集約予測を組み合わせた階層的予測フレームワークを提案し、Random Forestによる1〜5日先予測とMinT調整で整合性を確保する。消費主体の変電所は発電主体より予測精度が高く、階層調整は受動的変電所で最大6.4%のMAE削減をもたらすが、分散電源が増えるほど効果は減衰する。
English
As DERs and decarbonization-driven loads grow, DSOs need proactive MV/LV network management. This paper proposes a hierarchical forecasting framework combining secondary substation (SS) predictions with aggregated connection-point forecasts, using horizon-specific Random Forests and MinT reconciliation for coherence. Consumption-dominated substations are far more predictable than generation-dominated ones; reconciliation yields up to 6.4% day-ahead MAE reduction for passive substations but fades as local generation grows.
Unofficial AI-generated summary based on the public title and abstract. Not an official translation.
📝 gxceed 編集解説 — Why this matters
日本のGX文脈において
日本では再エネ大量導入と配電網の制約顕在化が進み、一般送配電事業者の運用高度化・レジリエンス強化に直結する。SSBJや有報の開示論点ではないが、電力会社の設備計画・系統運用の実務に資する知見を提供する。
In the global GX context
Fits the global energy-transition agenda where DSOs face rising DER penetration and electrification loads. While not a disclosure paper, it informs TCFD/ISSB-adjacent adaptation and grid-resilience planning, and offers transferable methods for utilities worldwide.
👥 読者別の含意
🔬研究者:階層的予測とMinT調整を配電網の異種変電所に適用した実証的知見を提供する。
🏢実務担当者:変電所の運用条件別に予測・調整戦略を適応させる実務的指針として活用できる。
🏛政策担当者:配電網の能動的管理と再エネ大量導入下での系統運用政策の設計に示唆を与える。
📄 Abstract（原文）
The increasing penetration of distributed energy resources (DERs) and new electric loads associated with decarbonization is pushing Distribution System Operators (DSOs) towards more proactive management of Medium Voltage (MV) and Low Voltage (LV) networks. In this context, short-term load forecasting (STLF) at the secondary substation (SS) level is becoming increasingly relevant for network operation and planning. However, conventional approaches typically use exclusively the data at SS level. without exploiting information available across the distribution network. This paper proposes a data-driven hierarchical forecasting framework that combines substation level predictions with aggregated forecasts of the underlying connection points. SS are characterized according to their operating conditions, based on the balance between annually consumed and produced energy, to investigate how these affect predictability. Forecasts from one to five days ahead are obtained using horizon-specific Random Forest (RF) models combining autoregressive, meteorological, and calendar information. Minimum Trace (MinT) reconciliation then ensures coherence between substation level and connection point forecasts. The analysis reveals markedly different forecasting behavior across operating conditions, with consumption-dominated substations proving considerably more predictable than their generation-dominated counterparts. Hierarchical reconciliation follows the same pattern, delivering its most consistent gains for passive substations, with an average day-ahead MAE reduction of 3.4%, reaching up to 6.4%, while its benefit gradually fades as local generation grows. The results provide DSOs with practical indications for adapting forecasting and reconciliation strategies across heterogeneous distribution networks.
🔗 Provenance — このレコードを発見したソース
- Research Square https://doi.org/10.20944/preprints202609.1999.v1first seen 2026-09-26 04:19:35 · last seen 2026-09-29 04:21:27
🔔 こうした論文の新着を逃したくない方は キーワードアラート に登録（無料・3キーワードまで）。
gxceed は公開メタデータに基づく研究支援データセットです。要約・翻訳・解説は AI 支援で生成されています。 最終的な解釈・検証は利用者が原典資料に基づいて行うことを前提とします。