Carbon Emission Prediction for Energy-Intensive Industries Based on a TCN-Transformer Hybrid Model

This research presents a systematic framework for high-precision carbon emission forecasting in energy-intensive industries, integrating emission accounting methodologies with a Temporal Convolutional Network (TCN)-Transformer hybrid architecture. First, direct and indirect emissions from representative enterprises in the steel, concrete, and electrolytic aluminum sectors were rigorously quantified using the Intergovernmental Panel on Climate Change (IPCC)emission factor approach. Subsequently, a TCN-Transformer model was developed to capture temporal patterns in annual emission data, where the TCN module leverages dilated convolutions and residual connections to extract multi-scale features, while the Transformer component employs multi-head self-attention to model long-range dependencies. Empirical results demonstrate that the proposed hybrid model outperforms standalone Transformer architectures by 35% in prediction accuracy, achieving a Mean Absolute Percentage Error (MAPE) of 3.21%. This significant improvement underscores the model's efficacy in capturing complex emission dynamics, providing a robust tool for proactive carbon management strategies.