Spatially distributed production data for supply chain models - Forecasting with hazardous waste

Spatially distributed production data for supply chain models - Forecasting with hazardous waste

WoS Article

This paper introduces a novel approach to forecasting future commodity production in hundreds of nodes, which represents a key input for many applications of supply-chain models. A mathematical model was proposed to handle the problem of forecasting with spatially distributed and uncertain data. It is derived from the principle of regression analysis and extended by a data reconciliation technique. Additional areal constraints guarantee mass conservation in a tree-like structure, which reflects the organisational arrangement of an investigated region. The proposed model was tested through a case study, where future production of hazardous waste suitable for thermal treatment was forecasted in 206 base-nodes, 14 superior nodes and one apex. Based on an extensive investigation of historical data, it was revealed that extrapolations carried out at different levels of the hierarchical organisational structure lead to inconsistent forecasts. The differences between forecasts reached up to 50%. In addition to this, mass conservation was violated. Significant corrections were performed by computations utilising the formulated model. The corrections ranged from between 0% and 12% for 90% of nodes. There were 17 nodes, where massive adjustments of up to 30% were inevitable.

This paper introduces a novel approach to forecasting future commodity production in hundreds of nodes, which represents a key input for many applications of supply-chain models. A mathematical model was proposed to handle the problem of forecasting with spatially distributed and uncertain data. It is derived from the principle of regression analysis and extended by a data reconciliation technique. Additional areal constraints guarantee mass conservation in a tree-like structure, which reflects the organisational arrangement of an investigated region. The proposed model was tested through a case study, where future production of hazardous waste suitable for thermal treatment was forecasted in 206 base-nodes, 14 superior nodes and one apex. Based on an extensive investigation of historical data, it was revealed that extrapolations carried out at different levels of the hierarchical organisational structure lead to inconsistent forecasts. The differences between forecasts reached up to 50%. In addition to this, mass conservation was violated. Significant corrections were performed by computations utilising the formulated model. The corrections ranged from between 0% and 12% for 90% of nodes. There were 17 nodes, where massive adjustments of up to 30% were inevitable.

Supply chain; forecasting; extrapolation; short time series; hazardous waste; thermal treatment

Supply chain; forecasting; extrapolation; short time series; hazardous waste; thermal treatment

PAVLAS, M.; ŠOMPLÁK, R.; SMEJKALOVÁ, V.; NEVRLÝ, V.; SZÁSZIOVÁ, L.; KŮDELA, J.; POPELA, P.

2018

14.07.2017

Elsevier Ltd

UK

0959-6526

Journal of Cleaner Production

161

United States of America

1317

1328

11

Pavlas2017Spatially distributed production data for supply chain models