Prediction and correlation study of dioxin emissions classifications from municipal solid waste incinerators

Existing studies on the correlation of dioxins mainly focus on dioxin indicators such as chlorobenzenes and phenols. To investigate the correlation between dioxin emissions and conventional pollutants as well as incineration conditions, this study compares the performance of decision trees, random forests, gradient boosting trees, and artificial neural networks for dioxin emissions classification based on 12,164 CEMS monitoring data from municipal solid waste incinerator across China. The results show that the random forest model outperforms the other models in terms of prediction accuracy, precision, F1 score, and recall rate, with values of 89.26 %, 94 %, 93.16 %, and 93.58 %, respectively. Furthermore, the correlation between dioxin emission and conventional pollutants is explored through statistical analysis, feature contributions analysis, and principal component analysis. The results indicate a very close association between dioxin emissions and CO concentration. Compared to the meet condition, there is an average increase of 101 % in CO concentration when dioxin concentration exceeds the limit. Additionally, dioxin emissions also show a high correlation with temperature and particulate matter concentration. These findings provide a theoretical basis and scientific guidance for dioxin reduction and regulation in municipal solid waste incinerators.