First full-scale combined MBBR, coagulation, flocculation, Discfilter plant with phosphorus removal in France

The moving bed biofilm reactor (MBBR) process is well known and reliable, and has been used extensively for nitrogen and carbon removal from wastewater (Ødegaard et al. 1994; Rusten & Ødegaard 2007; Ødegaard et al. 2010). Suspended solids (SS) removal by microscreens with woven media (either Discfilters or Drumfilters) downstream of an MBBR has been studied and industrialized since 2005 (Persson et al. 2006; Mattsson et al. 2009; Wilén et al. 2012; Gustavsson & Cimbritz 2013; Rossi et al. 2013). The SS concentrations from an MBBR process designed for biological oxygen demand (BOD) removal from municipal wastewater are typically in the 100–250 mg/L range (Ødegaard et al. 2010) and even up to 500 mg/L (Rossi et al. 2013). Therefore, chemical pre-treatment (flocculation alone or combined with coagulation) is needed upstream of microscreens with 40-micron pore openings to meet strict effluent discharge requirements. The advantages of combining a flocculation step with a 40-micron microscreen are increased SS removal efficiency and improved filtration flux, leading to decreased filtration area requirement, and hence lower investment cost and smaller footprint.

A coagulation-flocculation-filtration process configuration is necessary when chemical phosphorus precipitation is targeted (Väänänen 2017), and must be considered when there is a strict discharge requirement on effluent SS. Optimal dosages in total phosphorus (TP) removal applications are crucial to create flocs sufficiently strong to be retained by the sieving process (Väänänen 2017). Different wastewater characteristics (e.g. the influent chemical composition, particle characteristic size distribution, etc.) affect the optimum chemical dosage. Coagulant and polymer overdosing should be avoided, to keep the operating costs as low as possible, so careful dispersion and mixing design in the coagulation and flocculation stages, combined with precise process control, are necessary, especially when low TP concentrations are required (Väänänen et al. 2017).

Many small municipal WWTPs currently require upgrades due to population growth, legislation changes, and outdated treatment methods. Taninges WWTP was built in 1973 for 5,700 population equivalent (PE) using conventional activated sludge (AS) treatment. It was upgraded in 2015 and recommissioned in 2016 for up to 12,000 PE, with a pure MBBR process followed by 2-stage chemical pre-treatment and disc filtration to achieve <2 mg-TP/L and <10 mg-SS/L in the effluent.

Several pilot-scale studies have been carried out to gather comprehensive data for this process combination (Ødegaard et al. 2010), but it is important to show that long-term operation at full scale is successful. The aim of this work was to observe the performance of this first full-scale installation, and improve process knowledge on particle separation and TP removal efficiencies downstream of pure MBBR, without primary clarification upstream. An additional objective was to study whether lower effluent TP concentrations could be reached by adjusting coagulant dosing, thus preparing for stricter effluent TP requirements in the future.

Taninges WWTP is 2 km from the village centre, very close to popular ski resorts in south-eastern France. The only primary treatment methods for the wastewater are screening (15 and 3 mm), and fat and grit removal. During upgrading, the existing AS process at the WWTP was replaced by pure MBBR, in 2 lines of 2 reactors each in series (Figure 1). Carbon removal occurs in the first 193 m³ reactor (60% filling with AnoxKaldnes K5™ carriers, Veolia Water Technologies AB, Sweden) and nitrification in the subsequent 285 m³ reactor (50% filling with AnoxKaldnes K5™). The MBBR filling proportion enables treatment up to 12,000 PE, leaving the possibility to increase to 17,000 PE, if required, by adding a further MBBR line.

The MBBR is followed by 2-stage chemical pre-treatment for TP precipitation and a 40-micron Discfilter (HSF2212/11-1F, Veolia Water Technologies AB, Sweden), which has 11 vertically mounted discs. One more disc can be added to increase plant capacity, if required. The filter was dimensioned for a maximum flow of 280 m³/h. Part of the effluent is reused for backwashing the filter panels at 0.7 MPa without interrupting filtration, the rest is discharged to the river Le Giffre. Backwashing sludge is discharged to a buffer tank and then dewatered in a decanter centrifuge, whose centrate is returned upstream of the MBBR.

Before construction of the separation process after the MBBR and upgrading P removal with chemical treatment, coagulant and flocculation aids, dosing ranges for phosphorus removal were studied using conventional jar tests on site. Chemical doses of 3.7 mg-Al³⁺/L (polyaluminum chloride, PAX-18, Kemira Ltd., Finland) and 3 mg/L cationic emulsion polymer (Hydrex 6631 with 47% polymer content, Hydrex Veolia Ltd, France) were found to be suitable for initial operation of the recommissioned full-scale plant.

Once the plant was back in operation, a second set of bench-scale jar tests (Flocculator 2000, Kemira Kemi AB., Sweden) combined with filtration was required to optimise chemical dosing upstream of the filter. Effluent concentrations between 1 and 1.5 mg-TP/L were targeted, to ensure that the ≤2 mg-TP/L discharge requirement is fulfilled. During the three monitoring periods described here, the Al-based coagulant dose was varied between 3.7 and 5.5 mg-Al³⁺/L, on the basis of jar test results, and the cationic polymer was decreased from 3.0 to 1.9 mg/L.

Performance was monitored continuously using online instruments at 5 minute intervals, covering daily and seasonal variations from December 2016 to May 2017. Optical turbidity probes (Solitax SC, Hach Lange GmbH, Germany) and an online TP analyser (Phosphax Sigma, Hach Lange GmbH) were connected to a controller (SC1000, Hach Lange GmbH) to monitor SS and dissolved phosphorus (orthophosphate, PO₄³⁻) in the Discfilter influent and effluent.

Conversion factors were calculated from the correlations between turbidities measured online and SS. They were determined as 1.7 for the MBBR effluent and 2.5 for the filter effluent. Turbidity was also determined in the laboratory with a portable turbidity meter (Hach 2100Q, Hach Lange GmbH, Germany) to verify the accuracy of the values given by the online turbidity probes.

Hourly composite samples were also collected from the MBBR (before the coagulant dosing point) and the filter effluents, using two auto-samplers (AS950, Hach Lange GmbH) over three-day periods at three different locations, in December 2016, and April and May 2017. A grab sampling campaign was carried out parallel to the online measurement campaign (Table 1), and the online meters were calibrated regularly via the lab measurements.

The SS content was determined according to APHA Standard Methods (APHA 2005). Spectrophotometer based cuvette tests LCK348, LCK349 and LCK 350 (Hach Lange GmbH, Germany) were used for TP measurement. Cuvette tests (Hach Lange GmbH) LCK303 and LCK304 were used to measure ammonium (NH₄-N), LCK238 for total nitrogen (TN), and LCK114 and LCK314 for COD in the composite samples. All cuvette analyses were measured in a DR 2800 spectrophotometer (Hach Lange GmbH).

Typically, the post pure-MBBR SS concentration entering the solids separation unit is between 150 and 250 mg-SS/L when treating municipal sewage (van Haandel & van der Lubbe 2012). At Taninges WWTP (without primary treatment), the daily average SS concentrations in the MBBR effluent varied between 90 and 360 mg/L (Figure 2(a)). SS concentrations below 10 mg/L could be reached consistently using chemical pre-treatment (with 4.7 to 5.5 mg-Al³⁺/L and 1.9 mg/L polymer) and disc filtration. The Al³⁺ dose required was in the same range as in settling/flotation combined with chemical pre-treatment, where effluent SS is expected to be around 10–15 mg-SS/L (van Haandel & van der Lubbe 2012).

P removal from wastewater involves precipitating phosphate as SS and the subsequent removal of those solids (Tchobanoglous et al. 2003). The filter TP removal efficiency was between 50 and 97% (Figure 2(b)), depending on the Al³⁺ dose and the influent TP concentration. The filter effluent TP concentrations were below 2 mg/L, with Al³⁺ doses above 4.7 mg-Al³⁺/L for TP concentrations of 3–9 mg/L in the MBBR effluent.

The required molar ratios of Al³⁺ dosed/TP in the influent were between 0.7 and 2.3 to achieve <2 mg-TP/L in the filtrate (Figure 3). Previous studies indicated that a molar ratio of 5–7 mol-Me³⁺/mol-influent-TP was required to achieve <0.1 mg-TP/L consistently in the effluent from an activated sludge plant combining 2-stage chemical pre-treatment and disc filtration (Väänänen 2014, 2017). Such reported molar ratios were higher since the tests were performed on a different wastewater application with a lower influent TP concentration (0.2–0.5 mg-TP/L) and different target effluent quality (0.1 mg-TP/L). The coagulant doses recorded in this study were also lower than those used by Väänänen et al. (2016) in primary treatment (5–20 mg-Al³⁺/L or 10–30 mg-Fe³⁺/L), where <0.3 mg-TP/L was targeted in the effluent. In other words, the molar ratios of dosed coagulant to TP in the influent vary depending on the application, the target effluent TP concentration and the amount of phosphorus in soluble form.

In order to combine coagulation and flocculation with microscreen filtration, polymer doses of 1–5 mg-polymer/L can be required for primary and 0.5–1.5 mg-polymer/L for tertiary treatment (Väänänen 2017). The 1.9 mg/L emulsion dose added in this study would correspond to about 0.9 mg/L of active polymer dose (based on active matter fraction in the emulsion).

The average flow over 5 months was 61 m³/h, much lower than the design flow (280 m³/h). The online flow measurements indicate significant daily variations with flow peaks up to 125 m³/h (Figure 4). This suggests that, with the current chemical pre-treatment and filtration process, it is possible to take care of future increases in flow and loading, if required.

The variations in neither the MBBR influent turbidity (146–431 NTU) nor the flow affected the daily Discfilter effluent turbidity values, which remained between 0.6 and 10 NTU (Figure 4). The high influent variations could be explained by the lack of a primary clarifier at this WWTP. These effluent turbidity values would correspond to SS concentrations between 1.5 and 25 mg-SS/L, using the conversion factor reported in the Methods section. All turbidity values in the effluent (except for one day) were below the required effluent limit value of 35-mg SS/L (corresponding to 13.8 NTU).

The online measurements show that turbidity and flow peaks often occur simultaneously (Figure 4). SS loading to the Discfilter was between 20 and 300 g-SS/m²/h, leading to a backwash frequency of about 12%, on average, during the monitoring period. The loading and backwash frequency values indicate that the filter's capacity was not fully utilized, as the SS loading to the plant was much lower than the design value. The unit was sized to deal with the maximum instantaneous flow so that the plant could treat all wet weather flows during seasonal high loading events. The filtration process could also be used to target even lower TP effluent concentrations, with more aggressive coagulant and flocculant doses. According to Väänänen (2017) it is possible to achieve very low TP concentrations (<0.1 mg/L) in tertiary treatment effluents using high coagulant doses (up to 5 mg-Al³⁺/L).

The total solids (TS) content in the filter sludge varied between 4.0 and 6.8 g/L. The centrifuge dewatered the sludge to 19–25% dry solids. The centrifuge centrate, pumped back upstream of the MBBR, represented about 8% of total influent flow, when in operation. The SS and TP concentrations in the centrate were between 0.3 and 1.8 g-SS/L, and 4 and 13 mg-TP/L, respectively, contributing little to the MBBR load.

Further results from the sampling campaign in May (Table 2) show that, on average, reductions of 97% of COD and 99.8% of NH₄-N were achieved, fulfilling the effluent requirements. Average TN removal was 43% and nitrogen leaves the WWTP mainly as NO₃-N (the plant was not designed for denitrification). Average SS removal was 91%, which also met the effluent demands of the plant. This result is supported by turbidity removal results up to 99%.

The effluent TP requirement was achievable, giving a total removal efficiency of about 90%. As an alternative to chemical P removal, enhanced biological phosphorus removal (EBPR) is now a well-known and established process. In a long-term study, Tykesson et al. (2005), showed that it is possible to produce effluents containing less than 0.5 mg-TP/L using EBPR. However, they also point out that meeting a low effluent requirement of 0.3 mg-TP/L consistently is difficult, as even minor and short operating problems can lead to sudden instability in the biological process. Such variations can be handled instantly using chemical precipitation combined with microscreens, where rapid chemical dosing responses can help achieve more stable overall process performance (Väänänen 2017).

Van Haandel & van der Lubbe (2012) observed that MBBR biomass does not settle well and as phosphate removal is required in most countries, the cost of pre-treatment chemicals needs to be included in the operating costs of separation techniques. The estimated average annual cost of chemical pre-treatment in conjunction with the Discfilter installed at Taninges WWTP is about 14 kEUR (Table 3). The 0.027 EUR/m³ cost of treated wastewater is 4 times higher than for a comparable WWTP in Sweden, where coagulation-flocculation-Discfilter (10 micron) is used to polish activated sludge effluent (Kängsepp et al. 2016). This arises because of the higher SS and TP loads in the MBBR effluent at Taninges, which require higher coagulant and flocculant doses, and probable overdosing on site due to flow proportional control. If, for example, load-proportional feed-forward or feedback control of chemical dosing could be applied, annual chemical consumption could be reduced and the operating costs of the plant reduced even further.

The microscreen panels are cleaned chemically three times annually at Taninges WWTP. The cleaning chemicals cost about 0.6 EUR/a/m² filter area, about one third of the cost (about 1.9 EUR/a/m²) for the Discfilter installation in Sweden (Kängsepp et al. 2016), where chemical cleaning is carried out every six weeks. Chemical cleaning can be done less frequently at Taninges because high pressure cleaning (80 bar) is done monthly and the pores are 4 times larger (40 micron).

The total footprint of the MBBR process combined with coagulation-flocculation-filtration is about 160 m². Microscreens can be delivered with their own stainless-steel tank, which together with the reduced footprint, significantly reduces the construction costs of the plant.

This study has shown that a small-scale WWTP (5,700 PE) can be upgraded (to 12,000 PE) by combining MBBR with a coagulation-flocculation-Discfilter process to meet current and future treatment needs. It has also shown that this compact and robust treatment system can yield very good reductions in turbidity, NH₄-N and COD. The discharge limits of 2 mg-TP/L and 35 mg-SS/L have been achieved easily.

Nelson Llano (Veolia Water Technologies), Jerome Alloin and Quentin Zimmert (Veolia Eau) are acknowledged for assistance during sampling and analyses on the site.