Prediction of dioxins emissions from modern WtE plants with machine learning: in view of capacities, operation, and age of incinerators

As urbanization accelerates and consumption patterns shift, the global municipal solid waste (MSW) quantity continues to soar. In this context, incineration, a primary environmentally sound treatment method, has long been under the spotlight due to its associated dioxins emissions. Traditional methods for detecting dioxins are costly and cannot cover the entire operating time of the MSW incinerators, hence the growing interest in using machine learning to predict dioxins emissions. This study aims to predict dioxins emissions from waste-to-energy (WtE) plants using machine learning algorithms. In this study, we initially divided global data into three groups: China, Europe and the United States, and Japan. Then, we constructed prediction models based on deep forest regression and XGBoost regression, considering the age, capacity, and daily operation time of incinerators as influencing factors to forecast dioxins concentrations. The results indicate that age, capacity, and daily operation time are significantly correlated with dioxins emission concentrations (p < 0.01). Notably, the Japan-model performed the best, suggesting that age and daily operation time of incinerators significantly impact dioxins emissions. Furthermore, the study discovered that measures such as MSW classification, upgrades of air pollution control devices, optimized operating time, and constructing larger-scale WtE facilities can reduce dioxins concentrations. Moreover, we found that implementing any single one of these measures could lower dioxins concentrations by 20–50 %. This study offers a novel perspective for understanding and predicting dioxins emissions from WtE plants and provides a scientific foundation for the development and implementation of dioxins emission reduction policies.