In this study, nanofiltration (NF) was used to further treat the water from the disinfection process of a drinking water treatment plant (DWTP). The results indicate that the NF process reduced the dissolved organic carbon (DOC) by 92.87%, UV absorbance at 254 nm (UVA254) by 88.68%, conductivity by 94% and total dissolved solids by 94.49%. The removal efficiencies of trihalomethanes (THMs), 2-methylisoborneol (2-MIB) and trans-1,10-dimethyl-trans-9-decalol (geosmin) were also evaluated for both the DWTP and NF treatment processes. High-performance size-exclusion chromatography (HPSEC) analysis, three-dimensional excitation–emission matrix (3D-EEM) fluorescence spectroscopy and parallel factor (PARAFAC) analysis were applied to characterize the changes in different organic compounds in each process. A correlation analysis was carried out for samples of the raw water, feed water and permeate water to determine the correlations between the maximum fluorescence intensities (Fmax) of the PARAFAC components and the DOC, UVA254 and high specific UV absorbance (SUVA) of the samples. The results showed that a significant correlation (p < 0.01) existed between UVA254 and the Fmax of the three PARAFAC components as well as between DOC and Fmax.

Nanofiltration (NF) is a promising post-treatment technology for providing high-quality drinking water. However, membrane fouling remains a challenge to long-term NF in providing high-quality drinking water. Herein, we found that coupling pre-treatments (sand filtration (SF) and ozone–biological activated carbon (O₃-BAC)) and NF is a potent tactic against membrane fouling while achieving high-quality drinking water. The pilot results showed that using SF+O₃-BAC pre-treated water as the feed water resulted in a lower but a slowly rising transmembrane pressure (TMP) in NF post-treatment, whereas an opposite observation was found when using SF pre-treated water as the feed water. High-performance size-exclusion chromatography (HPSEC) and three-dimensional excitation–emission matrix (3D-EEM) fluorescence spectroscopy determined that the O₃-BAC process changed the characteristic of dissolved organic matter (DOM), probably by removing the DOM of lower apparent molecular weight (LMW) and decreasing the biodegradability of water. Moreover, amino acids and tyrosine-like substances which were significantly related to medium and small molecule organics were found as the key foulants to membrane fouling. In addition, the accumulation of powdered activated carbon in O₃-BAC pre-treated water on the membrane surface could be the key reason protecting the NF membrane from fouling.

Nanofiltration (NF) membranes, which can consistently offer safe and reliable water quality, have become increasingly popular in drinking water treatment. In this study, the conventional (coagulation-sedimentation-sand filtration) and ozonation-biologically activated carbon filtration (O₃-BAC) advanced treatment processes at a full-scale drinking water treatment plant (DWTP) were combined with a pilot-scale NF process for treatment of Taihu Lake water. The results showed that the “conventional + O₃-BAC + NF” combined processes had superior effects on removing natural organic matter (NOM), Br⁻, and other common water quality parameters (e.g., turbidity, conductivity, TDS, and total hardness) with efficiencies of 88.8–99.8%, for which the NF process played a critical role. The conventional plus O₃-BAC processes effectively removed formation potential of chlorinated disinfection by-products (Cl-DBPFPs, by 28.0–46.6%), but had poorer effect in reducing formation potential of brominated DBPs (Br-DBPFPs, by −2637.2–17.3%). NOM concentrations (characterized by dissolved organic carbon (DOC), ultraviolet absorbance at 254 nm (UV₂₅₄), and/or fluorescent components) were the driving factors for most DBPFP species, while elevation of [Br⁻]/[DOC] ratio likely resulted in enhanced formation of brominated trihalomethanes (THMs) during chlorination of the BAC effluent. By adding the pilot-scale NF process, the “conventional + O₃-BAC + NF” treatment train effectively controlled DBPFP, yielding the removal efficiencies of Cl-DBPFP and Br-DBPFP as 77.6–100% and 33.5–100%, respectively, with monochloroacetic acid, mono-bromo-acetic acid, and tribromomethane formation potentials (MCAA-FP, MBAA-FP, and TBM-FP) not detected in the final effluent. Low temperature in the winter season might be the primary reason for the rapid increase of transmembrane pressure when operating the NF membrane under flux of 25 L/(m²·h), which could be largely delayed by lowering the flux to 20 L/(m²·h). Characterization of the membrane cleaning solutions showed that macromolecular biopolymers (6000 Da–4000K Da) such as polysaccharides and proteins were the main contributors to membrane fouling.

For the treatment of lake water with algae, the coagulation–ultrafiltration–ozone–biologically activated carbon (CUF–O3–BAC) integrated process was first used to treat East Taihu Lake water in China, aiming at evaluating the removal efficiencies of algae, permanganate index (CODMn), UV254, NH3-N and disinfection by-products (DBPs) precursors. In addition, the long-term performance of the membrane operation under the fluxes of 60, 70, 80 and 90 L/(m2·h) was also investigated, and kinetic models were established. The experimental results showed that the integrated process had positive impaction of algae, CODMn, UV254 and NH3-N removal, and the removal rates were 95.89 ± 1.52, 76.18 ± 4.38, 72.06 ± 4.72 and 81.31 ± 6.71%, respectively. The CUF process was prone to increase the formation potentials of DBPs. Although ozone could reduce the formation risks of chlorinated trihalomethanes (THMs) to a certain extent, it is ineffective to reduce those of brominated THMs and haloacetic acids (HAA5). However, the CUF–O3–BAC process was an effective technology for the removal of THMs and HAA5 precursors in drinking water treatment. Finally, it was found that the relationship between transmembrane pressure (TMP) and time conformed to the first-order and second-order kinetic models, and the linear fitting coefficients were all above 90%.