Integrating Bio-Electrochemical Sensors and Machine Learning to Predict the Efficacy of Biological Nutrient Removal Processes at Water Resource Recovery Facilities

Abstract

Monitoring biological nutrient removal (BNR) processes at water resource recovery facilities (WRRFs) with data-driven models is currently limited by the data limitations associated with the variability of bioavailable carbon (C) in wastewater. This study focuses on leveraging the amperometric response of a bio-electrochemical sensor (BES) to wastewater C variability, to predict influent shock loading events and NO 3 − removal in the first-stage anoxic zone (ANX1) of a five-stage Bardenpho BNR process using machine learning (ML) methods. Shock loading prediction with BES signal processing successfully detected 86.9% of the influent industrial slug and rain events of the plant during the study period. Extreme gradient boosting (XGBoost) and artificial neural network (ANN) models developed using the BES signal and other recorded variables provided a good prediction performance for NO 3 − removal in the ANX1, particularly within the normal operating range of WRRFs. A sensitivity analysis of the XGBoost model using SHapley Additive exPlanations indicated that the BES signal had the strongest impact on the model output and current approaches to methanol dosing that neglect C availability can negatively impact nitrogen (N) removal due to cascading impacts of overdosing on nitrification efficacy.

Machine Learning Application for Nutrient Removal Rate Coefficient Analyses in Horizontal Flow Constructed Wetlands

“…The model evaluation indicators used were correlation coefficient ( R 2 , eq ), mean squared error (MSE, eq ), and root-mean-square error (RMSE, eq ). While R 2 indicates how well the datasets fit on the regression model, MSE and RMSE are the standard deviation (SD) of the predicted error and are indicators of how well the data points are distributed around the line of best fit. The R 2 , MSE, and RMSE were calculated using eqs , , and , respectively …”

“…For the evaluation of the models, the datasets were divided into training (80%) and testing (20%) datasets (Figure ). To avoid overfitting in limited datasets, threefold cross-validation was applied on the training dataset. , Three equally sized folds were made from the training datasets, two of which were utilized for training with one for validation. In order to prevent overfitting in single-training datasets, each fold was trained and validated.…”

A Critical Review of Data Science Applications in Resource Recovery and Carbon Capture from Organic Waste

“…Contrary to the frequently utilized but small primary experimental data sets, primary time-series data sets representing relatively larger sample sizes from long-term runs in pilot- or full-scale facilities were rarely used (less than 10%) for developing predictive (70%) and optimization (30%) models of only two technologies: anaerobic digestion (median N = 233) and gasification (median N = 3,831) (Figure a; gray box and whisker plots). Time-series data can provide important insights into the real-time nonlinear effects of organic waste input and varying operational conditions of a technology on RRCC output . Considering that certain RRCC technologies have achieved commercial implementation at large scales (e.g., more than 1,700 anaerobic digestors in the USA alone and close to 100 industrial pyrolysis plants worldwide), , limited use of time-series data for data-driven modeling in the reviewed literature reflects the difficulty of researchers to access long-term data from existing RRCC facilities for modeling and analysis.…”

“…Time-series data can provide important insights into the real-time nonlinear effects of organic waste input and varying operational conditions of a technology on RRCC output. 704 Considering that certain RRCC technologies have achieved commercial implementation at large scales (e.g., more than 1,700 anaerobic digestors in the USA alone and close to 100 industrial pyrolysis plants worldwide), 705 , 706 limited use of time-series data for data-driven modeling in the reviewed literature reflects the difficulty of researchers to access long-term data from existing RRCC facilities for modeling and analysis. Such inaccessibility to time-series data could also have contributed to the limited applications of deep learning methods in RRCC; only one study was found through this literature review that applied long short-term memory for predicting biogas production in an anaerobic digestion facility.…”

Optimization of a Novel Engineered Ecosystem Integrating Carbon, Nitrogen, Phosphorus, and Sulfur Biotransformation for Saline Wastewater Treatment Using an Interpretable Machine Learning Approach

“…Biological removal of nitrogen (N) and phosphorus (P) from wastewater plays a crucial role in protecting receiving waters from eutrophication. 1 Meanwhile, sulfur (S) contamination, arising from seawater intrusion, volcanic eruptions, 2 acid mining wastewater, 3 and flue gas desulfurization wastewater, 4 raises concerns about corrosion and odor emissions. As a result, numerous saline sewage treatment technologies have been explored to target S, with the aim of promoting N and P removal via S conversion.…”