Accurate multi-objective prediction of CO₂ emission performance indexes and industrial structure optimization using...

Area of Science:

• Environmental Science
• Computer Science
• Data Science

Background:

• Global carbon peaking and neutrality targets necessitate accurate prediction of CO2 emission performance indexes (CEPIs) and industrial structure optimization.
• Accurate multi-objective prediction of CEPIs remains a significant challenge in environmental and economic policy-making.

Purpose of the Study:

• To propose a novel Multihead Attention-based Convolutional Neural Network (MHA-CNN) model for accurate multi-objective prediction of CEPIs.
• To provide data-driven suggestions for industrial structure optimization to enhance carbon emission efficiency.
• To evaluate the performance of the MHA-CNN model against existing methods like CNN and LSTM.

Main Methods:

• Development and application of a MHA-CNN model incorporating deep learning strategies for feature extraction and model adaptability.
• Utilizing a multihead attention (MHA) mechanism to interpret variable influence weights and enhance prediction accuracy.
• Comparative analysis of MHA-CNN against CNN and Long Short-Term Memory (LSTM) models using 8 influence variables.

Main Results:

• The MHA-CNN model demonstrated superior performance in multi-objective CEPI prediction compared to traditional CNN and LSTM models.
• MHA analysis effectively identified variable contributions to CEPIs, showing high consistency with geographical distribution analyses.
• The study identified that increasing the tertiary industry while decreasing the first and secondary industries aids carbon emission efficiency.

Conclusions:

• The MHA-CNN model offers a robust solution for accurate CEPI prediction and variable importance analysis.
• The findings support optimizing industrial structures to improve total-factor carbon emission efficiency and energy utilization.
• This research provides critical insights for achieving global carbon reduction targets through advanced modeling and strategic industrial adjustments.