A Deep Learning Framework for Predicting Carbon Trading Price Dynamics

Amidst the pressing challenges posed by global climate change, carbon emissions have become a critical concern across numerous sectors. Carbon emission trading, as a market-based mechanism, stands out as a pivotal policy instrument for curbing emissions. The accurate prediction of carbon trading prices is crucial for effectively achieving dual-carbon objectives, positioning it as a central theme in academic research. This paper presents a novel interval prediction framework, CEEMDAN-BiLSTM-Attention-KDE, designed to address the inherent volatility of price series. By integrating a deep learning architecture with kernel density estimation, this approach provides interval predictions that offer substantial informational value. Empirical analysis based on data from the Shanghai carbon trading market demonstrates that the proposed prediction model delivers superior performance and predictive accuracy for carbon emission trading prices.