A factorial-analysis-based Bayesian neural network method for quantifying China's CO₂ emissions under dual-carbon...

Area of Science:

• Environmental Science and Policy
• Computational Science
• Energy Economics

Background:

• Urgent need for energy-structure transformation and CO2 emission reduction globally.
• China's dual carbon targets: carbon peak before 2030 and carbon neutrality by 2060.
• Existing methods lack comprehensive analysis of factors influencing CO2 emissions.

Purpose of the Study:

• Introduce a novel Factorial-Analysis-based Bayesian Neural Network (FABNN) method.
• Quantify and predict CO2 emissions in China (CEC) considering multiple factors.
• Analyze individual and interactive effects of energy, economic, and social factors on CEC.

Main Methods:

• Development and application of the FABNN method for CO2 emission analysis.
• Investigation of factors including natural gas consumption (CONG), GDP, and rate of urbanization (ROU).
• Simulation and prediction using 512 designed scenarios to meet dual carbon targets.

Main Results:

• FABNN demonstrates superior performance over conventional machine learning methods for CEC simulation.
• Key factors influencing CEC are identified as GDP (34.6%), CONG (43.5%), and ROU (21.9%).
• CEC projected to peak between 2027-2032 and achieve carbon neutrality between 2053-2057 across all scenarios.

Conclusions:

• The FABNN method is effective for simulating and predicting China's CO2 emissions.
• Energy consumption, economic growth, and urbanization significantly impact CO2 emission trajectories.
• Optimal scenarios show substantial CO2 reduction potential and decreased fossil fuel consumption, supporting national climate goals.