Modeling Carbon Emission Allowance Prices Using Multi-Scale Decomposition and Integrated Deep Learning Approaches
炭素排出枠価格のモデリング：マルチスケール分解と統合深層学習アプローチ (AI 翻訳)
Kai Sun, Xitao Fan
🤖 gxceed AI 要約
日本語
本論文は、中国炭素排出取引所（CCET）の2021年7月から2025年3月までのデータを用い、VMD-CEEMDAN分解、GARCH-MIDAS、CNN-BiLSTM-Attentionモデルを統合したハイブリッド予測モデルを構築。マクロ経済、類似商品、エネルギー構造、気候環境などの多周波影響因子を考慮し、MAE、MSE、RMSEで比較モデルを上回る精度と安定性を示した。
English
This paper develops a hybrid forecasting model for carbon emission allowance prices using China Carbon Emissions Trading Exchange data from July 2021 to March 2025. It integrates VMD-CEEMDAN decomposition, GARCH-MIDAS, and CNN-BiLSTM-Attention, incorporating multi-frequency influencing factors. The model outperforms benchmark models in MAE, MSE, and RMSE, demonstrating improved accuracy and stability.
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
日本でも炭素価格メカニズムの検討が進む中、中国市場のデータを用いた高精度な価格予測手法は、日本の排出量取引制度設計や企業のカーボンプライシング対応に示唆を与える。
In the global GX context
As carbon pricing expands globally, accurate price forecasting is crucial. This model, tested on China's growing ETS, provides a methodology that can be adapted to other markets like EU ETS or California, supporting transition finance and risk management.
👥 読者別の含意
🔬研究者:This paper introduces a novel hybrid model (VMD-CEEMDAN-GARCH-MIDAS-CNN-BiLSTM-Attention) for carbon price forecasting that outperforms existing models, offering a new benchmark for future research.
🏢実務担当者:Traders and risk managers in carbon markets can use this model to improve price prediction and inform trading strategies.
🏛政策担当者:Regulators can leverage such forecasting to monitor market stability and assess the effectiveness of carbon pricing policies.
📄 Abstract（原文）
Carbon trading, as the critical market mechanism to achieve greenhouse gas emotion reduction, have received more attention around the world. The target that China establish and improve Carbon Emissions Trading Exchange (CCET) in recent year is to help the achievement of “peak carbon dioxide emissions”, “carbon neutrality”. However, Considering the characteristics of nonlinearness, strong fluctuation of carbon price, the accuracy of data prediction reduced significantly. Based on China Carbon Emissions Trading Exchange (CCET) data from July 2021 to March 2025, this paper build a multi-factor hybrid forecasting model that integrates VMD-CEEMDAN decomposition with GARCH-MIDAS and CNN-BiLSTM-Attention model throughout the ARCH and Sample Entropy as testing methods. Meanwhile, this model introduce influence factor including Macro Economy, Similar Products, Energy Structure and Climate Environment. Combine the random forest method with the MIDAS approach to select and integrate multi-frequency influencing factors, and evaluate model performance using indicators such as MAE, MSE, and RMSE. The experimental results show that the proposed model is better than the other model in accuracy and stability. The indicators including MAE, MSE and RMSE all are less than other comparison models.
🔗 Provenance — このレコードを発見したソース
- semanticscholar https://doi.org/10.54097/enkwnq29first seen 2026-05-15 17:07:47
🔔 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。