Hybrid Feature‐Selection and Multi‐Scale Deep Learning for Carbon Price Forecasting in Emerging Carbon Markets
International Journal of Energy Research · 2026-01 · 2026 (1)
DOI: 10.1155/er/1314913 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Energy Research, online erschienen am 13.08.2026. Verfasst von Sensheng Li und 5 weiteren Autor:innen. Die Arbeit nennt 73 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Natural Science Foundation of Sichuan Province.
Abstract (Originalsprache)
As global attention to CO 2 intensifies, accurate carbon‐price forecasting has become a core topic in carbon‐market research. We propose a hybrid forecasting framework, extremely randomized trees (ET)–multivariate variational mode decomposition (MVMD)–gated recurrent unit (GRU), that integrates three advanced components to predict carbon prices for China’s six pilot emissions trading systems (ETSs)—Hubei, Shenzhen, Guangdong, Beijing, Shanghai, and Fujian. First, ET are used to assess feature importance and select key predictors. Second, after the observations are chronologically divided into an 80% training set and a 20% test set, MVMD is applied separately to the two subsets to decompose the carbon‐price series and the selected predictors into eight regularized intrinsic mode functions (IMFs), thereby achieving denoising and multi‐scale representation. Third, a GRU model is employed to learn the decomposed components and generate the final carbon‐price forecast. Empirical results across the six pilots show that ET–MVMD–GRU achieves the lowest MAE, RMSE, and MAPE among the 10 competing models, including the newly added ET–VMD–GRU and ET–ICEEMDAN–GRU, and remains in the superior model set in the model confidence set (MCS) tests. The framework offers a new methodological avenue for carbon‐price forecasting and provides evidence‐based guidance for policymakers on developing carbon markets in developing countries, strengthening risk management, and advancing the internationalization of carbon markets.
Förderung
- Natural Science Foundation of Sichuan Province
Zitieren
APA
Li, S., Huang, M., Wang, F., Hu, J., Wang, H., & Li, H. (2026). Hybrid Feature‐Selection and Multi‐Scale Deep Learning for Carbon Price Forecasting in Emerging Carbon Markets. International Journal of Energy Research, 2026(1). https://doi.org/10.1155/er/1314913
Vancouver
Li S, Huang M, Wang F, Hu J, Wang H, Li H. Hybrid Feature‐Selection and Multi‐Scale Deep Learning for Carbon Price Forecasting in Emerging Carbon Markets. International Journal of Energy Research. 2026;2026(1). doi:10.1155/er/1314913
BibTeX
@article{Li2026,
author = {Li, Sensheng and Huang, Minjun and Wang, Fuping and Hu, Junkun and Wang, Hanyue and Li, Houjian},
title = {Hybrid Feature‐Selection and Multi‐Scale Deep Learning for Carbon Price Forecasting in Emerging Carbon Markets},
journal = {International Journal of Energy Research},
year = {2026},
volume = {2026},
number = {1},
doi = {10.1155/er/1314913}
}
EUVIMED