Synergistic effect evaluation method of atmospheric emission reduction based on deep learning fusion model

Area of Science:

• Environmental Science
• Data Science
• Chemical Engineering

Background:

• Effective air quality management is hindered by the lack of nonlinear models linking industrial emissions to pollution levels.
• Accurate simulation of emission impacts is crucial for mitigating severe pollution peaks and informing policy.

Purpose of the Study:

• To propose a novel method for evaluating synergistic effects of atmospheric emission reduction.
• To simulate the impact of industrial emissions on air quality and develop targeted control policies.

Main Methods:

• Developed a deep learning fusion model, GR-BILSTM, integrating Generative Adversarial Network (GAN) and ResNet-BILSTM.
• Employed perturbation analysis to quantify the impact of industrial park emissions on PM2.5 concentrations.
• Validated model performance against LSTM, BILSTM, and CNN-BILSTM, assessing fitting accuracy.

Main Results:

• The GR-BILSTM model demonstrated superior fitting accuracy compared to existing models.
• Perturbation analysis quantified the influence of emission reduction ratios on subsequent PM2.5 concentrations.
• Identified SO2 (16%), NOx (15%), and TSP (18%) as major contributors to industrial park PM2.5 pollution.

Conclusions:

• The GR-BILSTM model provides an accurate simulation of emission-air quality relationships.
• The findings offer a quantitative basis for formulating targeted emission reduction policies in industrial parks.
• Understanding the specific contributions of pollutants like SO2, NOx, and TSP is vital for effective environmental management.