Insight into VOCs Source Profiles by Machine Learning: Role of Commonalities in Synergistic pollution Controls

Under the trend of low-carbon and cost-reduction, achieving efficient control requires identifying the commonalities in volatile organic compounds (VOCs) source profiles and implementing collaborative emissions reduction strategies. This study focuses on the analysis of common pollution characteristics in chemical industrial clusters, examining the emission behaviors of VOCs from nearly 200 emission outlets across 14 industries. A total of 593 VOCs were identified, including 488 new species. The highest concentration of newly discovered VOCs is 240 × 10³ μg/m³, accounting for 91%. The identical emission behavior of different components and isomers of industrial sources in several industries is revealed. The dominant species were redefined based on three dimensions. Using machine learning (ML), the maximum incremental reactivity (MIR) values of 488 VOCs were simulated, and based on the common characteristics of VOCs and photochemistry, VOC factor groups were identified that represent 75%-80% of the emission sources in the chemical industrial cluster. The average percentage of oxygenated volatile organic compounds (OVOCs) in this study was 28% higher than in other studies. This study follows the trend of synergistic emission reduction, reduces the blindness of large-scale establishment and updating of source profiles, and provides an efficient control method of VOCs.