2023-03-15
Short-term time series forecasting for multi-site municipal solid waste management
Publication
Publication
Procedia Computer Science , Volume 220 p. 170- 179
Smart containers equipped with ultrasonic sensors at waste and recycle facilities allow waste and recycling companies to build a more efficient and data-driven approach for the collection of municipal solid waste (MSW). In this paper, we propose three time series algorithms that predict the MSW generation of six waste types, using data obtained from smart sensors placed inside 3,640 containers at facilities in six municipalities in the Netherlands. Per neighborhood and per waste type, three models are developed: a Seasonal NaIve Benchmark model, ensemble models of Error, trend, seasonality models with external variables (ETSX), and Quantile Regression models with external variables. According to the RMSE, the ETSX model is the outperforming model for 74% of the time. It is also found that poor weather conditions such as precipitation, wind gusts and thunderstorms result in less waste disposal. The proposed prediction models can be used for more efficient waste collection, in order to collect waste before the fll rate percentage exceeds 100%. In future studies the inclusion of spatial variables and clustering of the containers can be considered.
| Additional Metadata | |
|---|---|
| , , , , , , , | |
| doi.org/10.1016/j.procs.2023.03.024 | |
| Procedia Computer Science | |
| 14th International Conference on Ambient Systems, Networks and Technologies Networks, ANT 2023 | |
| Organisation | Centrum Wiskunde & Informatica, Amsterdam (CWI), The Netherlands |
|
Fokker, E., Koch, T., & Dugundji, E. (2023). Short-term time series forecasting for multi-site municipal solid waste management. In Procedia Computer Science, Special Issue 14th International Conference on Ambient Systems, Networks and Technologies Networks, ANT 2023 (Vol. 220, pp. 170–179). doi:10.1016/j.procs.2023.03.024 |
|
