Journal of Membrane Science 301. High-recovery reverse osmosis desalination using intermediate chemical demineralization, страница 7

An additional benefit of chemical demineralization was the moderate to low organics removal (quantified in terms of 11–42% DOC removal) over the tested pH range of 9.2–11.3. The DOC fraction in the RO concentrate included both natural organic matter (NOM) and antiscalant which was added prior to the primary RO. The observed DOC removal was consistent with other studies on precipitation softening water treatment, suggesting that DOC removal occurs possibly due to surface adsorption and/or complexation with mineral salts, as well as co-precipitation [25,28]. It is important to note that in both bench-scale studies [10,12] and the present study, antiscalant carryover(forthepresentantiscalanttypeanddose)didnotresult in significant reduction in chemical demineralization efficiency with respect to cation removal.

Previous lab-scale work [12] suggested that, for CRW PRO concentrate that was supersaturated with respect to CaSO₄, precipitation of CaSO₄ may be induced by CaCO₃ precipitation by providing mineral seeds for CaSO₄ crystallization. However, the PRO concentrate generated during the Phase I pilot testing periodwasconsistentlybelowsaturationwithrespecttogypsum. As a result, SO₄²− removal was less than 1–2% throughout the testing period, suggesting that sulfate removal by CaSO₄ precipitationwasnegligible.Also,Ba²⁺ concentrationinthewaterwas negligible(∼7×10−⁷ M)comparedtoSO₄²− (∼3.5×10−³ M) and therefore BaSO₄ precipitation resulted in a relatively minor reduction of SO₄²− concentration.

Phase I pilot testing suggested that significant removals of membrane scaling precursors from the PRO concentrate were achieved by inducing CaCO₃ precipitation in the SCR using NaOH (Fig. 3). It was also revealed that the SCR effluent pH was a key monitoring parameter for controlling the level of demineralization in the SCR. Based on thermodynamic analysis, the levels of SiO₂ and Ca²⁺ removal attained during Phase I pilot testing (at SCR effluent pH>10) were determined adequate for enabling 95% recovery RO desalination without oversaturating the final RO concentrate with respect to SiO₂, SrSO₄, and CaSO₄. In addition, the attained Ba²⁺ removal levels were above

90% at SCR effluent pH above 10, which would ensure that the SRO concentrate (at 68% SRO water recovery or 95% overall water recovery) would have a saturation index below ∼30, with respect to BaSO₄; this would be well below the saturation index of ∼62 in the RO concentrate (at 83% water recovery desalination) from the PRO stage in which BaSO₄ scaling was not encountered. Thermodynamic solubility analysis also suggested that, with the levels of removal achieved in the SCR (Fig. 3), RO desalting was feasible at an overall water recovery of 95%, which was the operational upper bound limit of the present pilot system, and could possibly be further increased up to 98%, as previously demonstrated in bench-scale studies [10,12].

4.2. Integration of SCR and secondary RO to achieve up to

95% overall recovery

The performance of the integrated pilot-scale PRO– ICD–SRO desalting process (Fig. 2) was evaluated and demonstratedduringPhaseIIpilottesting(∼3.3months).Inthisphase, the PRO concentrate was continually demineralized in the SCR and the filtered SCR effluent was desalted in the SRO pilot unit (up to 68% recovery), thus enhancing the overall desalination water recovery (up to 95%). In Phase II pilot testing, the PRO unit was operated at a water product recovery of 83%, while the SRO unit was operated at either 50% or 68% recovery during two different periods, resulting in an overall water recovery of 91% and 95%, respectively. During the course of Phase II testing, certain system units (or equipment) were not fully operational which forced periods of system shut down (shown as shaded regions in Figs. 4–6) due to various maintenance and operational issues. The periods with continuous operation of the complete train of PRO desalting, SCR, and SRO desalting are denoted by the roman numerals (i)–(iv) in Figs. 4–6.