Deep learning-based screening approach for priority pollutants: a case study on retired power battery recycling

Area of Science:

• Environmental Science
• Materials Science
• Computer Science

Background:

• Growing volumes of retired power batteries pose environmental risks during recycling.
• Urgent need for effective methods to identify and assess these recycling-related environmental hazards.

Purpose of the Study:

• To develop and validate a novel screening framework for pollutant prioritization in retired power battery recycling.
• To apply deep learning and hierarchical clustering for accurate risk assessment.

Main Methods:

• Construction of an integrated pollutant screening model (McA) using five deep learning methods with performance-based weighting.
• Application of hierarchical clustering analysis for pollutant categorization.
• Utilizing SHapley Additive exPlanations (SHAP) for risk factor visualization.

Main Results:

• The McA model demonstrated significantly improved accuracy and reliability over traditional machine learning (R² = 0.9999).
• Identified 13 pollutants from retired power battery recycling, categorized into four priority levels (I-IV).
• SHAP analysis revealed key influencing factors: acute toxicity, irritation/corrosivity, and endocrine disruption.

Conclusions:

• The developed framework provides effective pollutant prioritization for retired power battery recycling.
• Deep learning methods show high potential for enhancing pollutant screening and risk management in battery recycling.
• Results offer insights for developing safer and more sustainable battery recycling processes.