Purifying IJssellake water

Abstract

The aim of this research was to evaluate the feasibility of direct hollow fiber nanofiltration membranes for drinking water purposes. The experiments were performed on the dNF40 pilot provided by NXF. The fouling potential and the ion retention of the dNF40 pilot were determined by continuous filtration experiments using raw IJssellake water under different operational conditions. The performances (ion retention and fouling potential) of the dNF40 on raw IJssellake water were compared with the dNF40 performances on pre-treated water from Waterwinstation Prinses Juliana (WPJ) (previously done at PWNT). The WPJ pre-treated water has undergone extensive pre-treatment consisting of drum screens, flocculation, sedimentation, rapid sand filtration (RSF) and granular activated carbon (GAC). The OMP retention of the dNF40 pilot was determined by full recirculation experiments using WPJ pre-treated water under two different operational conditions with elevated concentrations 'spiked solution'.
Limited to no fouling impact was observed on the membrane performance when feeding the pilot with raw IJssellake water. The membrane performance parameters (mass transfer coefficient (MTC), trans membrane pressure (TMP) and normalized pressure drop (NPD)) were stable over time. In addition, limited to no fouling impact was observed on the membrane when feeding the pilot with WPJ pre-treated water. However, membrane performance (i.e. MTC) was better for raw IJssellake water (1 year old membrane) compared to WPJ pre-treated water (virgin membrane). This implies that the active outer layer of the membrane has undergone a change in properties leading to these higher MTC values.
An increase in recovery, flux and crossflow velocity resulted in a decrease in ion retention. However, a decrease in ion retention with elevated crossflow velocity is unusual. Higher crossflow velocities should actually lead to an increase in ion retention due to reduced ion build-up next to the membrane surface (i.e. lower concentration polarization effect). However, the lower ion retention can be attributed to the elevated MTC during experiments. The removal of natural organic matter (NOM) was consistently above 90% and was not influenced by a change in operational condition. Ion retention was higher for WPJ pre-treated water (virgin membrane) compared to raw IJssellake water (1 year old membrane). This can be attributed to the increase in MTC of 1.5 LMH/bar in raw IJssellake water (compared to WPJ pre-treated water) potentially caused by a change in the properties of the active outer layer.
For determining the OMP retention of the dNF40 membrane, a spiked solution containing per- and polyfluoroalkyl substances (PFAS) and pharmaceutical compounds was analyzed. The PFAS compounds of the spiked solution were retained very well (above 80%). As expected, the retention increased with increasing MW. The adsorption percentage of PFAS was between 40%-90%. The pharmaceutical retention was around 30%, although all pharmaceuticals analyzed had a MW below the MWCO of the membrane.
A 5-stage full-scale dNF40 plant was designed based on a permeate flow of 15 M m3/year, a total hardness concentration in the permeate stream below 1.4 mmol/L and a recovery percentage of 85%. Based on the 5-stage full-scale dNF40 plant an economical analysis was performed and compared to the full-scale UF-RO in Heemskerk. The total cost (OPEX and CAPEX) were cheapest for the full-scale dNF40 plant fed with raw IJssellake water (12 ct/m3), followed by the dNF40 plant fed with WPJ pre-treated water (32 ct/m3) and most expensive for the UF-RO plant (35 ct/m3). The major factors in the OPEX was the membrane replacement cost for the dNF40 plant and the energy and chemical cost for the UF-RO plant.