Treatment and hybrid modeling of domestic reverse osmosis concentrate using biological activated carbon

Abstract

The simulation and removal of organics from municipal reverse osmosis concentrate (ROC) using biological activated carbon (BAC) was investigated. BAC treatment of ROC achieved a TOC removal of 25.5% and 28.1% for empty bed contact time (EBCT) of 40 min and 60 min, respectively. Batch studies revealed that the TOC removal followed a first order reaction with a rate constant of 2.735 at R2 > 0.98. Most of the organics were removed within 80 min of treatment, with an average non-biodegradable fraction of 0.6453. The results obtained from the batch studies were then used for simulation of the continuous lab-scale BAC reactors, yielding average relative deviations of 7%–8% and normalized root mean square errors (RMSE) of 0.09–0.1. Results from a third reactor with EBCT of 80 min were used to train and verify a machine learning model. The model was then used for calculating the non-biodegradable organics fraction of ROC, which will be fed into the first order reaction model. The resulting serial hybrid modeling further reduced the average relative deviations to 5%–6% and a normalized RMSE of 0.08.