Modelling secondary waste composition using optimization and machine learning techniques: Case of the Czech Republic

Modelling secondary waste composition using optimization and machine learning techniques: Case of the Czech Republic

WoS Article

To support the shift toward a circular economy in waste management, it is essential to monitor progress using measurable indicators. However, the growing volume of secondary waste from pre-treatment processes highlights the need to assess its composition, as it can represent a diverse mixture and complicates the evaluation of individual waste streams. The proposed approach aims to estimate the composition of secondary waste by using a combination of machine learning and optimization techniques. The cornerstone for evaluation is data from waste management monitoring. Machine learning based on linear or Bayesian linear regression allows for the efficient processing of large datasets and the identification of key relationships in the system. The optimization model developed for a special form of data reconciliation maintains insight into the results and ensures the preservation of mass balances. In a case study in the Czech Republic, the model identified a 3 % reduction in the material recovery of municipal waste, as this waste is used for energy recovery or landfilled after transformation into secondary waste. Mixed secondary waste consists of 46 % plastic waste, with only 20 % being truly recycled. A significant portion is landfilled, which represents a potential for at least energy recovery from the waste. With refined waste management indicators and potential for recovery, the results can contribute to improvements in terms of technology and regional focus.

To support the shift toward a circular economy in waste management, it is essential to monitor progress using measurable indicators. However, the growing volume of secondary waste from pre-treatment processes highlights the need to assess its composition, as it can represent a diverse mixture and complicates the evaluation of individual waste streams. The proposed approach aims to estimate the composition of secondary waste by using a combination of machine learning and optimization techniques. The cornerstone for evaluation is data from waste management monitoring. Machine learning based on linear or Bayesian linear regression allows for the efficient processing of large datasets and the identification of key relationships in the system. The optimization model developed for a special form of data reconciliation maintains insight into the results and ensures the preservation of mass balances. In a case study in the Czech Republic, the model identified a 3 % reduction in the material recovery of municipal waste, as this waste is used for energy recovery or landfilled after transformation into secondary waste. Mixed secondary waste consists of 46 % plastic waste, with only 20 % being truly recycled. A significant portion is landfilled, which represents a potential for at least energy recovery from the waste. With refined waste management indicators and potential for recovery, the results can contribute to improvements in terms of technology and regional focus.

Waste management indicators; Machine learning; Data reconciliation; Quadratic optimization; Secondary waste composition; Material recovery

Waste management indicators; Machine learning; Data reconciliation; Quadratic optimization; Secondary waste composition; Material recovery

ŠOMPLÁK, R.; PLUSKAL, J.

01.08.2025

Waste Management

205

1

United States of America

115019

13

https://www.sciencedirect.com/science/article/pii/S0956053X25004301