Application of a hybrid mechanistic/machine learning model for prediction of nitrous oxide (N2O) production in a nitrifying sequencing batch reactor

Nitrous oxide (N2O) is a key parameter for evaluating the greenhouse gas emissions from wastewater treatment plants. In this study, a new method for predicting liquid N2O production during nitrification was developed based on a mechanistic model and machine learning (ML) algorithm. The mechanistic model was first used for simulation of two 15-day experimental trials in a nitrifying sequencing batch reactor. Then, model predictions (NH4-N, NO2-N, NO3-N, MLSS, MLVSS) along with the recorded online measurements (DO, pH, temperature) were used as input data for the ML models. The data from the experiments at 20 degrees C and 12 degrees C, respectively, were used for training and testing of three ML algorithms, including artificial neural network (ANN), gradient boosting machine (GBM), and support vector machine (SVM). The best predictive model was the ANN algorithm and that model was further subjected to the 95% confidence interval analysis for calculation of the true data probability and estimating an error range of the data population. Moreover, Feature Selection (FS) techniques, such as Pearson correlation and Random Forest, were used to identify the most relevant parameters influencing liquid N2O predictions. The results of FS analysis showed that NH4-N, followed by NO2-N had the highest correlation with the liquid N2O production. With the proposed approach, a prompt method was obtained for enhancing prediction of the liquid N2O concentrations for short-term studies with the limited availability of measured data. (C) 2022 Published by Elsevier Ltd on behalf of Institution of Chemical Engineers.

Automated summary

A new method combining a mechanistic model and machine learning algorithms was developed to predict liquid N2O production during nitrification. The mechanistic model was used to simulate experimental trials, and the machine learning algorithms were used for prediction. Feature selection techniques were employed to identify the most relevant parameters for liquid N2O predictions. The proposed method enables fast and accurate prediction of liquid N2O concentrations with limited availability of measured data.