AI-Driven Carbon Intensity Forecasting with Green Sliding Window Detection
AI駆動のカーボンインテンシティ予測とグリーンスライディングウィンドウ検出 (AI 翻訳)
M. D R, S. S, S. M, Anup P Vaidya, Pranav Srinivas, Anagh Bheemsenrao Deshpande
🤖 gxceed AI 要約
日本語
本論文は、LSTMネットワークを用いて都市グリッドの短期的な炭素強度（CI）を予測するAIフレームワークを提案する。予測値をCarbon Score Indexに変換することで、産業施設が低CI運転ウィンドウを特定し、排出量を15%以上削減できることを実証した。従来手法と比較して22〜30%の性能向上を示し、軽量深層学習が持続可能な都市エネルギーシステムに貢献する可能性を示す。
English
This paper presents an AI framework using LSTM networks to forecast short-term carbon intensity (CI) of urban grids, converting forecasts into a Carbon Score Index. Industrial facilities can identify low-CI operating windows, achieving 15%+ emission reductions. Performance metrics show 22-30% improvement over classical baselines, demonstrating lightweight deep learning for sustainable urban energy systems.
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
日本の工場やマイクログリッドがカーボンインテンシティの変動を活用し、運用調整による排出削減を実現する手法として有用。SSBJや有報でのGHG排出量管理にも応用可能。
In the global GX context
This framework advances carbon-aware operations globally by enabling real-time, data-driven emission reductions. Aligns with TCFD/ISSB expectations for forward-looking climate metrics and supports corporate net-zero strategies.
👥 読者別の含意
🔬研究者:LSTM-based carbon forecasting with actionable index opens new directions for carbon-aware decision-making research.
🏢実務担当者:Carbon Score Index provides a practical tool to schedule industrial processes during low-carbon hours, reducing operational emissions.
🏛政策担当者:Supports demand-side flexibility policies and urban grid decarbonization targets by translating forecasts into actionable insights.
📄 Abstract（原文）
Urban grids account for approximately 7 0% of global energy emissions, but the carbon intensity (CI) of urban grids varies significantly throughout the day, meaning there is no clear guidance available for cities or microgrids to operate in a lowcarbon fashion. Current systems typically provide cities with CI information for historical periods but do not provide an accurate forecast of short-term CI along with actionable insight, which limits progress toward achieving the United Nations' Sustainable Development Goals. The work described here presents a framework based on AI technology that predicts short-term CI for an urban grid using Long Short-Term Memory (LSTM) networks trained on multi-year intervals of demand, renewable generation, weather conditions, and interconnector data. Short-Term Carbon Intensity Forecasts are converted to a Carbon Score Index to enable urban and industrial energy consumers to identify low-CI operating windows. In the model validation process, the forecast's performance metrics yielded a Scaled MAE of 0.0875 and a RMSE of 0.1066, representing a 22% to 30% performance improvement compared to classical baselines. Experimentation indicates that by using the identified low-carbon operating windows, industrial facilities may reduce their emissions by 15% or more. Overall, the findings of this research support the premise that contemporary lightweight deep learning architectures can facilitate practical and data-informed carbon-aware decision-making for developing sustainable urban energy systems.
🔗 Provenance — このレコードを発見したソース
- semanticscholar https://doi.org/10.23919/indiacom70271.2026.11525648first seen 2026-05-29 05:44:43 · last seen 2026-06-03 05:21:01
🔔 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。