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

This study shows for the first time howBES signals canprovide an essential data stream to track bioavailable C for data-drivenmodeling and optimization of BNR processes. Monitoring biological nutrient removal (BNR) processesat waterresource recovery facilities (WRRFs) with data-driven models is currentlylimited by the data limitations associated with the variability ofbioavailable carbon (C) in wastewater. This study focuses on leveragingthe amperometric response of a bio-electrochemical sensor (BES) towastewater C variability, to predict influent shock loading eventsand NO3 (-) removal in the first-stage anoxiczone (ANX1) of a five-stage Bardenpho BNR process using machine learning(ML) methods. Shock loading prediction with BES signal processingsuccessfully detected 86.9% of the influent industrial slug and rainevents of the plant during the study period. Extreme gradient boosting(XGBoost) and artificial neural network (ANN) models developed usingthe BES signal and other recorded variables provided a good predictionperformance for NO3 (-) removal in the ANX1,particularly within the normal operating range of WRRFs. A sensitivityanalysis of the XGBoost model using SHapley Additive exPlanationsindicated that the BES signal had the strongest impact on the modeloutput and current approaches to methanol dosing that neglect C availabilitycan negatively impact nitrogen (N) removal due to cascading impactsof overdosing on nitrification efficacy.