The effect of feed water TDS on RO membrane rejection rates and performance

Abstract

Power generation is a major user of fresh water, which is becoming important for Eskom and South Africa due to the water-scarce status. Therefore, the water usage at a thermal power station is a key metric and reverse osmosis (RO) is a key technology that can be used to reduce water consumption. A double pass RO system at a thermal power station used to produce demineralised water for steam production is operating with a large variance in its brackish feed water with total dissolved salts (TDS) ranging between 80 and 1000 mg/L. The reference plant showed a significant decrease in observed salt rejection with lowering feed water concentration. The reason for this decrease in salt rejection was not clear to Eskom and required further investigation. An experimental test rig was used to study 4” spiral wound membrane modules to replicate the reference plant in a small-scale and controllable setup. The range of feed water concentration was replicated by treating NaCl solutions varying from 10 - 2500 mg/L. Polyamide RO Membranes from three different suppliers (Filmtec™, Toray and Hydranautics) were tested to compare the response in salt rejection to feed water concentration. The test rig was not temperature controlled, and therefore, more than one variable changed during the experimental data collection which requires the development of a model from literature and experimental data to enable the comparison between the membranes on an equal footing. The results produced by the standard salt diffusion equation significantly deviated from the experimental results below 100 mg/L NaCl feed concentrations. This deviation was attributed to the Donnan effect which showed a reduced effect at lower concentrations and subsequently lowers the salt rejection capability of the RO membrane in this region. Therefore, additional terms presented by Anisimov & Orlov (2018) were incorporated into the salt flux equation which then produced a model that predicts the membrane performance throughout the entire feed concentration range that was tested. The experimental input data is also replicated using the modelling software available from the three membrane suppliers. The results from the software correlated well with the experimental and developed model results. The recommended modifications from an optimisation exercise on the reference plant could allow the plant to be more flexible by achieving maximum recovery over the entire range of feed water qualities. The modifications proposed can potentially allow the reference plant, when operating on the lowest feed water concentration, to increase overall water recovery from 77% to 96%, reduce wastewater by 80% and increase the plant production by 23%.