Sequential algorithm of building the regression-classification model for total nitrogen simulation: application of machine learning
Artykuł w czasopiśmie
MNiSW
100
Lista 2023
| Status: | |
| Autorzy: | Barbusiński Krzysztof, Szeląg Bartosz, Białek Anita, Łazuka Ewa, Popławska Emilia, Szulżyk-Cieplak Joanna, Babko Roman, Łagód Grzegorz |
| Dyscypliny: | |
| Aby zobaczyć szczegóły należy się zalogować. | |
| Rok wydania: | 2023 |
| Wersja dokumentu: | Drukowana | Elektroniczna |
| Język: | angielski |
| Wolumen/Tom: | 301 |
| Strony: | 106 - 114 |
| Web of Science® Times Cited: | 0 |
| Bazy: | Web of Science |
| Efekt badań statutowych | NIE |
| Materiał konferencyjny: | NIE |
| Publikacja OA: | NIE |
| Abstrakty: | angielski |
| Total nitrogen (TN) concentration is one of important indications of wastewater quality and also a parameter important for wastewater treatment plant performance evaluation. Since the variability of total nitrogen in the effluent from the wastewater treatment plant is the result of the processes taking place in the bioreactor, the processes can be described by mechanistic models, for example, activated sludge models. However, calibration of many parameters is required in such models, and can leads to problems in identifying their proper numerical values. The paper proposes a novel way to deal with this problem by presenting a methodology for building a model for simulating TN, based on sequential structure. In the applied approach, regression models for simulation of TN are first created using Extreme Gradient Boosting (XGBoost), and random forest (RF) methods. In the case of unsatisfactory predictive ability, a division of the dependent variable into a classifier form is made. In the next stage, classification models are created by RF and XGBoost methods and sensitivity analysis is performed by calculating Shapley indices. Two classification models were built that allow for the identification of TNeff variability ranges. The new approach using two models instead of one is preferable because it allows control and optimization of the bioreactor operation. |
