Pilot Scale Comparison of Enhanced Coagulation with Magnetic

Pilot Scale Comparison of Enhanced Coagulation with Magnetic Resin Plus Coagulation Systems. Peter Jarvis, Max Mergen, Jenny Banks, Brian McIntosh, ...
0 downloads 0 Views 810KB Size
Environ. Sci. Technol. 2008, 42, 1276–1282

Pilot Scale Comparison of Enhanced Coagulation with Magnetic Resin Plus Coagulation Systems PETER JARVIS,† MAX MERGEN,† JENNY BANKS,‡ BRIAN MCINTOSH,† SIMON A. PARSONS,† AND B R U C E J E F F E R S O N * ,† Centre for Water Science, Building 39, Cranfield University, Cranfield, Bedford, MK43 0AL, United Kingdom, and Yorshire Water Services Ltd, Halifax Road, Bradford, BD6 2LZ, United Kingdom

Received June 26, 2007. Revised manuscript received October 9, 2007. Accepted October 22, 2007.

Previous work has shown that magnetic ion-exchange treatment before coagulation gives high natural organic matter (NOM) removal and reduced levels of disinfection byproduct when compared to conventional enhanced coagulation. The impact of the resin process on the downstream floc formation process after coagulation and the subsequent effect on clarification has not previously been shown. Water containing high concentrations of NOM were treated at pilot scale using (1) conventional enhanced coagulation and compared with (2) treatment using magnetic resin followed by coagulation at reduced doses of 50–70%. Bench scale testing was also carried out to determine floc properties for systems with and without resin pretreatment. It was demonstrated that pretreatment using magnetic resin was able to significantly reduce the turbidity load onto filters as a result of the formation of a large and more robust floc. Resin pretreatment also improved NOM removal and reduced disinfection byproduct formation when compared with conventional coagulation. The turbidity load on to the filters following resin pretreatment was 1.5 ( 0.7 NTU, whereas this value was 2.9 ( 0.3 NTU for conventional coagulation. Flocs produced with resin pretreatment were larger than those produced by conventional coagulation, with a median floc size of 1000 µm compared to 600 µm. The improvement in floc properties following magnetic resin pretreatment was proposed to be due to the removal of NOM that was characteristic of carboxylic acids before the coagulation stage.

Introduction Coagulation using metal salts is the standard method for removing natural organic matter (NOM) during drinking water treatment. This is normally achieved by lowering pH and increasing coagulant doses in a process commonly referred to as enhanced coagulation. During enhanced coagulation, NOM removal is increased thereby reducing the formation of disinfection byproduct (DBPs) (1–3). DBPs form when residual organic matter reacts with disinfectant (usually chlorine) to form potentially harmful and carcinogenic compounds, the most common by mass (and therefore * Corresponding author phone: +44 1234 7544813; fax: +44 1234 751671; e-mail: [email protected]. † Cranfield University. ‡ Yorshire Water Services Ltd. 1276

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 4, 2008

most widely studied) being the trihalomethanes (THMs) and haloacetic acids (HAAs) (4). One of the main disadvantages of enhanced coagulation is increased chemical addition and the production of more sludge that must be disposed of. In addition, coagulation has been shown to give poor removal of low molecular weight (MW) and hydrophilic NOM (5, 6). Magnetic ion-exchange resin has emerged as an effective alternative treatment technology for NOM removal. The resin is a strong base anion (SBA) exchange resin with ammonium functional groups, consisting of 150–180 µm beads of a macroporous, polyacrylic structure (7, 8). During the treatment process, NOM rich raw water is contacted with the magnetic resin, allowing ion-exchange of the negatively charged functional groups of the NOM with chloride ions on the resin. The resin matrix contains a magnetic iron oxide component that enables rapid aggregation and rapid settlement during clarification from the treated water stream following the contact stage. Many studies have shown the benefit of using magnetic resin for removing NOM and reducing DBP formation (1, 3, 9–12). For some types of water, using the resin on its own has been shown to be as effective or better than when compared directly to enhanced coagulation for dissolved organic carbon (DOC) removal and reduced DBP formation (1). However, for most water types, work has shown that pretreating organic rich water with magnetic resin followed by coagulation with reduced coagulant doses of up to 80% gives reduced dissolved organic carbon residuals and lower DBPs by up to 75% when compared to coagulation (3, 9–11). While removal efficiencies have been shown to be significantly improved by resin pretreatment, little research has been carried out on the impact of using magnetic resin on downstream processes, particularly given the reduced coagulant doses used. There is some bench-scale evidence showing that pretreatment of NOM laden waters with magnetic resin followed by coagulation results in an increase in floc size, strength, and settling rate when compared to just using coagulation (12). However, there is no evidence showing how downstream floc properties and solid–liquid separation processes may be affected in a continuous water treatment system. The principal objective of this work was to link the operational performance of a continuous pilotscale magnetic resin system and a coagulation pilot plant with raw water character and the properties of the flocs formed during downstream treatment. A direct comparison was made between treating water with conventional coagulation and a resin pretreated water followed by coagulation with reduced coagulant doses.

Materials and Methods (A) Pilot Trials. Raw water from a moorland reservoir feeding a water treatment works (WTW) in the north of England (Albert WTW, Halifax, UK) was treated using the following: (1) A pilot plant consisting of enhanced coagulation, dissolved air flotation (DAF), and filtration. Water was coagulated using doses as used by the WTW at the time of abstraction of the raw water. (2) Pilot plants using magnetic resin pretreatment followed by conventional coagulation, DAF and filtration at reduced coagulant doses (50, 60 and 70% reductions). Magnetic resin pilot plant trials were conducted at Albert water treatment works, Halifax, Yorkshire using a 1.0 m3 hr-1 pilot rig provided by Orica Watercare. The resin was mixed with raw water at a concentration of 30 mL L-1 in the contact tank for 13 min to give a resin service of 1000 bed volumes (BV) before the solid–liquid separation stage where resin was separated by rapid sedimentation. 96.7% of the resin was 10.1021/es071566r CCC: $40.75

 2008 American Chemical Society

Published on Web 01/09/2008

TABLE 1. Raw Water Characteristics and Removal Summary for the Treatment Systems Investigated post filter final water quality

DOC (mg L-1) UV254 (m-1) SUVA (L mg-1 m-1) turbidity (NTU) zeta potential (mV)

raw water

magnetic resin pretreated water

conventional coagulation

7.3 40.4 5.5 1.7 -16.9

3.7 21.5 5.8 1.3 -11.1

1.2 ( 0.3 2.4 ( 0.5 2 0.04 ( 0.01 -1.3 ( 3.0

returned back to the contact tank. The remaining portion was regenerated in brine before returning to the contact tank. Optimized resin concentrations and contact times had been established for this water using standard jar testing in preliminary studies. Water pretreated with magnetic resin was tankered to a pilot plant at Cranfield University (Bedford, UK) consisting of coagulation, DAF and filtration. The plant operated at a flow rate of 0.27 m3 hr-1. The pilot plant was operated for three hours for each coagulation condition. Each coagulation condition was run through the pilot plant twice. The performance results reported were taken from all of the data points collected from both runs. The experimental conditions run through the pilot plant were as follows: (1) Conventional coagulation: coagulation using ferric sulfate (Ferripol XL) at 7.7. mg L-1 as Fe at pH 4.2. (2) MR50%: coagulation of water treated with magnetic resin (conditions as above) using a 50% coagulant dose reduction as used for conventional coagulation (3.85 mg L-1 Fe, pH 4.2). (3) MR60%: coagulation of water treated with magnetic resin (conditions as above) using a 60% dose reduction as used for conventional coagulation (3.1 mg L-1 Fe, pH 4.2). (4) MR70%: coagulation of water treated with magnetic resin (conditions as above) using a 70% dose reduction as used for conventional coagulation (2.3 mg L-1 Fe, pH 4.2). Water was rapidly mixed for 3 min after the addition of ferric sulfate coagulant. The coagulation pH was kept at pH 4.2 (as used on site) by the addition of sodium hydroxide (0.01 M). Water was then flocculated for 33 min using picket fence stirrers. Water then passed to a DAF tank operating at 7 m3 m-2 hr-1 at a 22% recycle ratio and an air concentration of 88 mg L-1. Treated water then went on to a 0.3 m diameter filter column operating at 9 m3 m-2 hr-1 and containing 16/ 30 grade sand (1–0.5 mm diameter) at a depth of 1 m. After passing through the filter, the final water passed through an online turbidity meter. The start of each experimental run was taken from when water had started to pass through the filter. The saturated air was then turned on and samples were taken post DAF and filtration every 15 min for the duration of each run. Online turbidity was logged from the online turbidity meter every five minutes. (B) Sampling and Analysis. Raw water, magnetic resin pretreated water, and coagulated waters post DAF and filtration were analyzed for dissolved organic carbon (DOC) using a Shimadzu TOC-5000A TOC analyzer, ultraviolet absorbance at 254 nm (UV254) using a Jenway 6505 UV/vis spectrophotometer, turbidity using a Hach 2100 turbidimeter and zeta potential using a Malvern Zetasizer. Raw water and resin pretreated water were fractionated by XAD resin adsorption techniques into their hydrophobic (HPOA) and hydrophilic (HPIA) components using a published method in Sharp et al., 2006 (13). High performance size exclusion chromatography (HPSEC) analysis was undertaken using a high performance liquid chromatography device (Shimadzu VP series, Shimadzu) using UV254 detection (further details in the Supporting Information: HPSEC methodology). Water samples were taken every 15 min during each experimental

MR50% 0.9 ( 0.2 1.3 ( 0.3 1.4 0.03 ( 0.02 -4.4( 5.9

MR60%

MR70%

1.2 ( 0.3 1.6 ( 0.6 1.3 0.03 ( 0.02 -2.6 ( 3.4

1.1 ( 0.4 2.0 ( 0.9 1.8 0.06 ( 0.11 -10.2 ( 3.9

run after DAF and after filtration and analyzed for DOC, UV254, turbidity and zeta potential. Samples were taken every hour post filtration for HPSEC and total trihalomethane formation potential (THMFP). THMFP was carried out using a method adapted from 5710 in “Standard Methods for Treatment and Examination of Water and Wastewater (1992)” (14) (Supporting Information: THM methodology). Water was filtered through a 0.45 µm filter paper before DOC, UV254, HPSEC, fractionation and THM analysis. (C) Floc Characterization. Floc size and strength were compared for flocs formed after conventional enhanced coagulation and pretreatment followed by coagulation with 60% coagulant reduction using published techniques (15–17). Floc formation tests were carried out on a jar tester. After a rapid mix at 200 rpm and the slow stir at 30 rpm, the effect of increased shear rate was investigated by increasing the rpm on the jar tester for a further 15 min. Separate experiments were carried out and repeated three times for rpms of 30, 40, 50, 75, 100, 150, and 200. Dynamic floc size was measured during growth and breakage of the flocs using a laser diffraction instrument (Malvern Mastersizer 2000). The suspension was monitored by drawing water through the optical unit of the Mastersizer and back into the jar by a peristaltic pump on the return tube using 5 mm internal diameter peristaltic pump tubing at a flow rate of 1.5 L hr-1. Size measurements were taken every minute for the duration of the jar test and logged onto a PC.

Results (A) Natural Organic Matter Removal. The raw water used in this study was typical of a UK moorland water source being of high organic content (7.3 mg L-1 DOC), high UV254 absorbance (40.4 m-1) and low turbidity (1.7 NTU) (Table 1). As expected, this was water was amenable to treatment by magnetic resin (18). After resin pretreatment, the DOC had been reduced by 50% from 7.3 to 3.7 mg L-1 DOC; a similar trend as was seen for UV254 removal (Table 1). There was no significant change in the SUVA before and after resin pretreatment given that there was only a 5% difference between the two (0.3 L mg-1 m-1). Further information on the characterization of the organic material removed by the treatment options is given in the Supporting Information: NOM characterization and Figure B. The removal of NOM after coagulation for systems with and without resin pretreatment showed differences in residual DOC and UV254 absorbing compounds after treatment (Figure 1a and b). These curves show the percentile of DOC, UV254 under the value shown on the x axis and give an indication of the robustness of the system (the same data presented in time series is shown in the Supporting Information: Figures C and D). In the context of this paper, the definition of a robust system has been taken from Huck and Coffey (19) as “one that provides excellent performance under normal conditions and deviates minimally from this during periods of upset or challenge”. Here, we have evaluated robustness by looking at the mean residual values and the slope of the linear regression line between 10 and 90% of the VOL. 42, NO. 4, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1277

FIGURE 1. Robustness of the water quality for the different treatment options for DOC and turbidity after DAF and filtration. The straight lines represent the linear regression lines through the 10–90%iles. samples. A steeper slope through the data across this range indicated a more robust treatment around the average than for a system with a widespread of data around the average. It was evident from Figure 1a and b that the MR50% reduction consistently achieved lower DOC residuals than conventional coagulation. This was reflected by the lower mean DOC of 0.9 ( 0.2 mg L-1 after DAF for the MR50% compared to 1.5 ( 0.4 mg L-1 for conventional coagulation. Similarly, the DOC after filtration was 0.9 ( 0.2 mg L-1 for the MR50% reduction and 1.2 ( 0.3 mg L-1 for conventional coagulation. All of the other resin pretreated systems also showed consistently better removal after DAF than for conventional coagulation. After filtration there was less difference in DOC removal between enhanced coagulation and the MR60% MR70% coagulant reductions (Table 1). This was because the filter had a capacity for removing some residual DOC that was not removed during coagulation; this removal was greater when the residual was higher. The robustness of the different treatment systems for DOC removal was high for the MR50% treatment, as reflected by a higher slope of 207 when compared to a slope of 102 and 115 for conventional treatment after DAF and filtration respectively. The other resin treatment systems had similar robustness to the conventional system except the MIEX70% after DAF which was 33% more robust than MIEX60% and conventional coagulation. However, given the closeness of the DOC residuals, this was not considered a significant result. The UV254 removal after DAF was consistent with that seen for DOC removal, with resin pretreated systems outperforming conventional coagulation removal (Supporting 1278

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 4, 2008

Information: Figure E). The improved NOM removal for magnetic resin pretreated systems was also reflected by reduced THM formation following resin pretreatment (Supporting Information: Figure F and text description). After resin pretreatment, only a small part of the charged organic matter was removed as reflected by only a 35% reduction in the zeta potential (Table 1). The subsequent coagulation stage removed most of the charge. If insufficient charge is removed by coagulation, through insufficient addition of positively charged coagulant, process stability will be reduced. For the MR70% system the zeta potential of the treated water was -10.2 mV, whereas for the MR50% and MR60% the zeta potential was more positive (-4.4 to -2.6 mV). (B) Turbidity Removal. The turbidity measured in the flocculator for conventional coagulation was 19 NTU, while for resin pretreatment the value was higher at 38 NTU. The mean turbidity after DAF for conventional coagulation was 2.9 ( 0.3 NTU, while this was 1.50 ( 0.65 NTU after magnetic resin pretreatment regardless of the coagulant reduction used representing removal efficacy across the DAF of 85 and 96% respectively (Table 1). From Figure 1c, it was apparent that resin pretreatment gave significantly lower turbidities after DAF for all of the reduced coagulant doses used. The steep slopes of these lines for up to 90% of the cumulative samples (slopes of between 81 and 89) reflect a more robust treatment process because a narrow range of turbidities were observed (between 1.92 and 0.88 NTU), whereas the turbidity using conventional coagulation spanned a greater range of turbidities (2.56–5.66 NTU) and a more shallow slope of 27.

FIGURE 2. (a) Floc growth for flocs formed with and without magnetic resin pretreatment for MR60% and (b) Floc strength profiles for d10 and d50 floc sizes.

FIGURE 3. Cumulative floc size distributions with increasing breakage rpm for flocs formed with and without magnetic resin pretreatment based on volume (a) and (b) particle number. The deterioration in turbidity for the top 10% of samples of all systems was turbidity spikes experienced during the initial stabilization of the DAF. However, it was interesting to note that the performance of the conventional system deteriorated for the top 20% of samples which further indicated a less robust treatment system. For the final treated water, low mean turbidity residuals were achieved for conventional coagulation and MR50% and MR60% reductions (0.03–0.04 NTU) (Table 1). Some deterioration and a more unstable final turbidity was seen at the highest coagulant reduction, reflected by the increased residual turbidity of 0.06 NTU and high standard deviation of this turbidity ((0.11 NTU). This result was because of insufficient removal of charged NOM identified by a more negative zeta potential of -10.2 ( 3.9 mV of the coagulated suspension at the low coagulant dose compared to values of between -1.3 ( 3.0 and -4.4 ( 5.9 mV for all of the other coagulant systems. Filtered turbidity was much more robust and consistent for all of the treatment options (Figure 1d). This was particularly the case for MR50% and MR60% systems where the slope values were high (ranging between 639 and 985). The conventional coagulation and MR70% systems had more shallow slopes of between 425 and 480, reflecting a greater spread of turbidities and therefore less robust systems. The residuals above 80% deviated from the steep part of the slope for all treatment systems. This was a result of the filter ripening stage giving rise to initially higher turbidities. The higher range of turbidities seen for the MR70% coagulant

reduction was seen as a reflection of the extended filter ripening for this system because of poor charge neutralization. (C) Floc Characterization. The flocs formed for the conventional coagulation system were compared with those formed from the MIEX60% system. This was chosen because significantly improved turbidities were seen from the pilot plant at this dose reduction while similar DOC removals were seen when compared with conventional coagulation. The median volume based floc diameter (d50) was significantly different between systems with and without resin pretreatment (Figure 2a). The median floc size for conventional coagulation reached a maximum of 600 µm after 7 min of the jar test. The equivalent size for flocs formed after resin pretreatment was 1000 µm. The response of the flocs to increased shear rate was used to determine floc strength. There was a significant linear relationship (r2 ) 0.92–0.96) between log floc size and log rpm for the median and 10th percentile (d10) floc sizes (Figure 2b). As the shear rate increased, floc size decreased. It was apparent that until 100 rpm, magnetic resin pretreated flocs maintained a greater size than for conventional coagulation. At higher shear rates, there was convergence on a similar floc size of 150 µm for the median floc size and 50 µm for the d10 floc size. Analysis of the cumulative floc size distributions showed a large scale change across the whole of the floc size distribution after exposure to the highest shear rate (200 rpm) (Figure 3a). For conventional coagulation and resin pretreated systems less than 10% of flocs were smaller VOL. 42, NO. 4, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1279

than 200 µm before breakage, but after exposure to 200 rpm, 75% of the flocs were less than 200 µm. At the lower breakage shear rate (40 rpm), the floc size distribution did not significantly change before and after the increased shear rate. Software within the particle size instrument used in these studies was able to accurately convert the volume based particle diameter to a number frequency distribution (20). This was done to assess the number of particles before and after breakage in the system because this is of importance in both DAF and filtration (21, 22). The cumulative number distribution conversions in Figure 3b show some interesting patterns. First, there was a significant decrease in the median equivalent particle diameter when the distribution was based on number frequency indicating that there were a very high number of small particles in the system. The median floc size was 29 ( 13.9 µm for conventional coagulation and 39.7 ( 3.1 µm for resin pretreated flocs before breakage. There was an appreciable change in the number distribution after a slight increase in shear rate (40 rpm) between the different systems. For conventional coagulation, 20% of particles were between 10 and 20 µm, while for magnetic resin pretreatment no flocs were generated in this size class. Under the highest level of shear rate (200 rpm), 26% of the particles were