Performance de la nanofiltration pour l'élimination de la matière organique naturelle: essais sur l'usine de Méry/Oise

L'intérêt croissant que les traiteurs d'eaux portent à l'élimination de la matière organique naturelle (MON) a abouti au développement de nouvelles technologies de traitement. Dans ce but, un prototype de nanofiltration à l'échelle industrielle (2 x 1400 m 3 j-¹) est installé à l'usine de Méry sur Oise depuis juillet 1992. Utilisé en traitement de finition après clarification et filtration sur sable, il alimente depuis février 1993 un réseau test de la commune d'Auvers sur Oise (6 000 hbts) en région parisienne.L'objectif de cette publication est de présenter quelques uns des résultats de caractérisation de la MON obtenus pendant 9 mois d'expérimentation (octobre 1992 à juillet 1993), et en particulier ceux concernant les rendements d'élimination de la matière organique naturelle et par voie de conséquence de la demande en chlore.Ces rendements sont généralement supérieurs à 90 % en termes de COD,CODB et d'absorbance UV à 254 et 270 nm éliminés. L'analyse des potentiels de réactivité avec le chlore (taux de chloration: 2,5 mg Cl2/mg C, temps de réaction: 72 heures, pH = 7,5, 20 °C) montre que le perméat est peu consommateur de chlore (demande en chlore < 0,2 mg l-¹ Cl2) et peu précurseur de chloroforme et de trihalométhanes (PFCHC13 < 3 µg l-¹, PFTHM < 11 µg l-¹). Les rendements d'élimination des PFTHM et PFTOX sont généralement supérieurs à 90 %.L'analyse spécifique des constituants majoritaires du perméat montre que les acides aminés totaux (hydrolyse acide puis dérivation à l'OPA/HPLC) constituent une proportion importante du COD (25 à 60 % selon les saisons). Ces composés représentent la quasi totalité de la demande en chlore du perméat si l'on se réfère aux données bibliographiques.Compte tenu de ces résultats, la nanofiltration apparaît comme un excellent procédé de traitement de finition des eaux à potabiliser. En effet, bien qu'elle constitue une barrière de sécurité contre les germes pathogènes, la très faible charge organique du perméat obtenu par nanofiltration (COD~0,15 à 0,3 mg l-¹ C, CODB<0,1 mg l-¹ C) rend plus aisée la maîtrise du résiduel de chlore (lorsqu'une chloration est nécessaire pour maintenir la qualité de l'eau dans les réseaux) et constitue une limitation importante de la formation des sous-produits de chloration.Increasing interest in removing natural organic matter (NOM) has lead to the development of new drinking water treatment technologies. Since July 1992, a nanofiltration demonstration plant (2 x 1400 m3 d-1) has been used to treat sandfiltered water from the Oise river. The permeate has been distributed since February 1993 to the 6000 inhabitants of Auvers/Oise in the Paris suburb. The purpose of this paper is to present and discuss some ofthe results obtained over nine months of operation of this full scale plant, particularly yields of NOM removal and consequently the decreasing of chlorine reactivity (chlorine demand, TTIM and TOX formation potentials).Dissolved organic carbon (DOC) and UV-absorbance were determined using DOC analyser and a spectrophotometer. Biodegradable dissolved organic carbon (BDOC), which represents the biologically assimilable portion of DOC, was determined using the method of JORET et LEVI (1986). Chlorine demand, trihalomethane and total organohalide formation potentials (THMFP and TOXFP) were carried out under the following experimental conditions: applied chlorine dose of 2.5 mg Cl2/mg DOC, pH = 7,5 72 h-contact time and 20°C. Ultrafiltration experiments involved the use ofa laboratory ultrafiltration cell, Total amino-acids were analysed by HPLC after hydrolysis and orthophtaldialdehyde (OPA) derivatization. Aldehyde and ketone determination was based on the method developedby GLAZE et al. (1989) involving pentafluorobenzyl hydroxyl amine (PFBHA) derivatization.Characterization of sand-filtered water (SFW): The sand-frltered water (SFW) upstream of the nanofiItraton membranes has a DOC between 2.4 and 4.l mg l-1, depending on the season (table 1). Its BDOC ranges from 0.7 to l.l mg l-1 C. In fact, a BDOC value higher than 0.3 mg l-1 C has been mentioned by several authors as the limit above which possible bacterial regrowth can take place in the distribution network.The chlorine consumption curves, shown in figure 2 for five sampling campaigns, indicate that the chlorine demand of the SFW can reach 3.4 to 5.2 mg l-1 depending on the season (table 2). The THMFP and the TOXFP are 108-149 ug l-1 and 344-446 ug l-1 Cl- respectively. Note that the ratio of chlorine demand over DOC varies from 1.0 to 1.7 mg Cl2/mg DOC while the THMFP/DOC and TOXFP/DOC ratios present average values of 47.5 ug/mg DOC and 160 ugCl-/mg DOC respectively.The distribution of the SFW (table 3) indicates that the fraction with apparent mo lecular weight Iess than 3 kilodatons contains the major compounds at this stage of the water treatment. This fraction presents the highest chlorine consumption. Specific total amino acids (TAA) analyses demonstrate that TAA represent 3 to 8% of the DOC of the sand-filtered water. The most abundant arnino acids are glycine, aspartic acid, glutamic acid, serine and alanine. The chlorine consumption attributed to these amino acids is evaluated as 1 mg l-1 Cl2, that is to say 1/5 to 1/3 of the SFW chlorine demand. Formaldehyde and acetaldehyde seem to be the major aldehydes present in the SFW with a level of 7 ug l-1 and 20 ug l-1 of formaldehyde and acetaldehyde respectively. They represent only about 0.5 to 0.6 % of the SFW DOC.Characterization of the permeate: The nanofiltration permeate presents a very low NOM level in terms of DOC, BDOC and UV absorbance at 270 nm, that is to say 0.14 to 0.34 mg l-1 C, < 0.1 mg l-1 C and < 0.006 cm-1-l respectively (table 6).The chlorine consumption curves, showt in figure 4 for five sampling campaigns, demonstrate the low permeate reactivity with chlorine. The chlorine demands (table 7) after 72 hours are between 0.12 and 0.32 mg l-1. Moreover chlorine demand/DOC ratios have a value from 0.46 to 0.93 mg Cl2/mg DOC, i.e. half the values measured for SFW. The THMFP and TOXFP (72hours) range from 7 to 11 ug l-1 and 26 to 31 ug l-1 Cl- respectively.Total amino acid (TAA) analyses showed that TAA represent 35 to 60% of the permeate DOC and can account almost entirely for the chlorine consumption. Formaldehyde and acetaldehyde (the major aldehydes analysed) represent 7 to 8% of the permeate DOC.According to the results presented in this paper, nanofiltration appears to be an excellent technolory as a polishing step in surface water treatment. Whereas the level of sand-fïltered water (SFW) DOC varies from 2.4 to 4.1 mg l-1 C (depending on the season), the permeate DOC is consistently lower than 0.3 mg l-1 C. The efficiency of nanofiltration is about 90% for DOC, BDOC and consequently for chlorine demand, THMFP and TOXFP. The high retention of NOM is probably in relation with the percentage (75%) of compounds with apparent molecular weight above 500 daltons in the SFW. In fact the low values of BDOC and chlorine demand justify the use of nanofiltration for the production of a water which represents a very low risk of bacterial regrowth and a low risk of formation of disinfection by product in the network when distributed with a low concentration of residual chlorine