Year: 2025
Type: article
Abstract: Accurate carbon price forecasting enables the steady operation of the carbon trading market and optimal resource allocation while also empowering market participants to understand dynamics and make informed decisions, ultimately supporting sustainable development in the carbon market. While early research primarily focused on point forecasting of single-value carbon price, recent studies have shifted towards interval prediction, although there is still a lack of research dedicated to developing models for interval-valued predictions. The importance of interval-valued forecasting lies in its ability to better capture the upper and lower bounds of the carbon price range across different time dimensions, thereby revealing the intrinsic patterns and trends of price fluctuations and assisting in point forecasting to comprehensively capture carbon market volatility. This study offers a novel approach based on a CEEMDAN-CNN-BiLSTM-SENet hybrid model, providing a framework for both point and interval-valued carbon price predictions. The model makes a more comprehensive analysis of the carbon market possible by combining the predictions from these two approaches. In the case study using Hubei market’s data, the mean absolute percentage error for carbon pricing was 0.8125%, with the MAPE for the highest and lowest prices being 1.8898% and 1.7852%, respectively—both outperforming other comparative models. The results demonstrate that this model can measure trends of carbon pricing effectively.
Source: Energies
DOI: 10.3390/en18051054
Authors: Haoyu Chen, Qunli Wu, Chonghao Han
Institutions: CHN Energy (China), North China Electric Power University
Language: English
Field: Decision Sciences
Domain: Social Sciences
SDG: Climate action
Open Access status: gold
Estimated APC paid: $3,137
OpenAlex ID: W4407842955