Artificial intelligence‐based multiobjective optimization of reverse osmosis desalination pretreatment using a hybrid ZnO‐immobilized/photo‐Fenton process

Abstract

AbstractThe performance of desalination plants predominantly depends on the enhancement of membrane productivity through the effective removal of organic foulants from saline water prior to the membrane process. This research evaluates the performance of the ZnO‐immobilized solar nanophotocatalytic process integrated with Fe2+/H2O2 system for the removal of organics from reverse osmosis (RO) feed seawater. Machine‐learning and response surface methodology (RSM) models were used for optimizing the performance of such a hybrid system in terms of five input factors: initial TOC (mg/L), pH, H2O2 dosage (g/L), Fe2+ dosage (mg/L) and solar irradiation time (minutes). Both machine‐learning and RSM regression models were optimized using nondominated sorting genetic algorithm (NSGA‐III) for estimating optimum organic degradation performance in terms of residual Fe2+, total organic carbon (TOC) removal and chemical oxygen demand (COD) removal. The response values obtained from the experimental run conducted at the optimum settings of ANN‐NSGA‐III was found to be TOC removal = 81.4%, COD removal = 77.4% and residual Fe2+ = 1.95 mg/L. The pilot‐scale solar nanophotocatalytic reactor optimized in the present research is worthy of being upscaled for wide application in desalination plants.