An algorithm for selecting a machine learning method for predicting nitrous oxide emissions in municipal wastewater treatment plants - Publikacja - MOST Wiedzy

Wyszukiwarka

An algorithm for selecting a machine learning method for predicting nitrous oxide emissions in municipal wastewater treatment plants

Abstrakt

This study presents an advanced algorithm for selecting machine learning (ML) models for nitrous oxide (N2O) emission prediction in wastewater treatment plants (WWTPs) employing the activated sludge process. The examined ML models comprised multivariate adaptive regression spline (MARS), support vector machines (SVM), and extreme gradient boosting (XGboost). The study explores the concept that involves new criteria to select the most appropriate ML model: (i) fitting the results of model prediction to the measurements taking into account the range of variability of the input data, and (ii) model verification applying a reference MCM to identify the input-output relationship using the global sensitivity analysis (GSA). Using the k-means method, it was shown that the relative errors (%e) of N2O prediction by ML models depend on the range of variability of the input data (nitrogen compounds concentration in the bioreactor compartments, influent flowrate, air flowrate). The smallest relative errors of N2O prediction (%e = 0.13 for MARS, %e = 0.12 for SVM and %e=0.10 for XGboost) were found for the concentrations: NH4-N = 24.14 mg N/L (anaerobic compartment), NO3-N = 5.40 mg N/L (aerobic compartment), and the largest (%e > 0.35) for the concentrations: NH4-N = 29.43 mg N/L (anaerobic compartment), NO3-N = 7.90 mg N/L (aerobic compartment). Calculations using the GSA method confirmed that the XGboost model was the only one that showed identical relationships between all the considered input variables and N2O emission rate. The ML model obtained in this way can be used as an alternative to the MCM for estimating N2O emission as a significant contributor to the carbon footprint of WWTPs.

Cytowania

  • 1 2

    CrossRef

  • 0

    Web of Science

  • 1 4

    Scopus

Cytuj jako

Pełna treść

pełna treść publikacji nie jest dostępna w portalu

Słowa kluczowe

Informacje szczegółowe

Kategoria:
Publikacja w czasopiśmie
Typ:
artykuły w czasopismach
Opublikowano w:
Journal of Water Process Engineering nr 54,
ISSN: 2214-7144
Język:
angielski
Rok wydania:
2023
Opis bibliograficzny:
Szeląg B., Zaborowska E., Mąkinia J.: An algorithm for selecting a machine learning method for predicting nitrous oxide emissions in municipal wastewater treatment plants// Journal of Water Process Engineering -Vol. 54,iss. 103939 (2023), s.1-10
DOI:
Cyfrowy identyfikator dokumentu elektronicznego (otwiera się w nowej karcie) 10.1016/j.jwpe.2023.103939
Źródła finansowania:
  • Publikacja bezkosztowa
Weryfikacja:
Politechnika Gdańska

wyświetlono 101 razy

Publikacje, które mogą cię zainteresować

Meta Tagi