Abstract

Continuously flushing moving bed sand filter was operated in pilot scale for phosphorus (P) and nitrogen removal with simultaneous particle removal. The wastewater tested was either final effluent from a municipal wastewater treatment plant (WWTP) with nitrogen removal in moving bed biofilm reactors (MBBRs) followed by coagulation and dissolved air flotation (DAF) for P and suspended solids (SS) removal, or different mixtures of this final effluent and effluent from the MBBR-stage. The study focused on the applicability to achieve low total phosphorus (TP) concentrations (below 0.1 mg/L) and suspended solids concentrations (below 10 mg SS/L), plus good denitrification (removal rate over 750 g NO3-N/m3-d), by treating wastewater having variable concentrations of TP (from 0.19 to 7.3 mg/L), SS (from 3 to 169 mg/L) and total nitrogen (from 8 to 27 mg/L). The target effluent TP limit was easily achieved when adding coagulant to WWTP effluent. With correct coagulant dose (Al/TP-molar ratio >4) and good particle removal the target effluent TP could also be reached when treating mixed WW with fairly high influent TP. Very high denitrification rates were achieved with adequate influent P concentration and external carbon source. Low denitrification rates were observed when limited by low concentrations of biodegradable carbon and phosphorus.

LIST OF ACRONYMS

     
  • Al

    Aluminium

  •  
  • BOD

    Biochemical oxygen demand

  •  
  • BSCOD

    Biodegradable soluble COD

  •  
  • COD

    Chemical oxygen demand

  •  
  • DAF

    Dissolved air flotation

  •  
  • DN

    Denitrification

  •  
  • DO

    Dissolved oxygen

  •  
  • FCOD

    Filtered COD

  •  
  • FTN

    Filtered total nitrogen

  •  
  • FTP

    Filtered total phosphorus

  •  
  • GRA

    Gardermoen WWTP

  •  
  • MBBR

    Moving bed biofilm reactor

  •  
  • MBSF

    Moving bed sand filter

  •  
  • N

    Nitrogen

  •  
  • NO₂-N

    Nitrite nitrogen

  •  
  • NO3-N

    Nitrate nitrogen

  •  
  • NO3-Neq

    Equivalent NO3-N concentration of NO3-N, NO₂-N and DO

  •  
  • OP

    Ortho phosphorus (PO4-P)

  •  
  • P

    Phosphorus

  •  
  • PAC

    Polyaluminium chloride

  •  
  • RFID

    Radio frequency identification device

  •  
  • SS

    Suspended solids

  •  
  • SV

    Sludge volume

  •  
  • TCOD

    Total COD

  •  
  • TN

    Total nitrogen

  •  
  • TP

    Total phosphorus

  •  
  • WW

    Wastewater

  •  
  • WWTP

    Wastewater treatment plant

INTRODUCTION

Norway has a long tradition of removing phosphorus (P) from municipal wastewater (WW) using chemical precipitation. The most common demand is for 90% annual average removal of total P (TP) and well-running wastewater treatment plants (WWTP) come down to about 0.3 mg TP/L in the final effluent. By implementing new regulations for environment conservation, the demand for effluent quality is becoming stricter in certain regions. With stringent demands on removal of P, nitrogen (N) and biochemical oxygen demand (BOD) from WW, many WWTPs are in need of upgrading their current treatment processes. This is the case for the Gardermoen WWTP (GRA), Jessheim, Norway, that will be required to reduce effluent TP to below 0.1 mg/L as an annual average. The requirement for nitrogen removal is unchanged at 70% as an annual average, but since the plant is currently operating at above design loads for both organic matter and nitrogen, GRA is looking for a process that can also do some denitrification to prolong the service life of the plant. This prompted GRA to explore the use of moving bed sand filters (MBSF) as a polishing step for the existing WWTP (Fylkesmann 2015)

MBSF have been in use in the water treatment industry since early 1980s with several advantages contra conventional filtration (Jonsson et al. 1997). MBSF is a type of ‘continuously flushing sand filter’ where the filtration is not being interrupted for back flushing and cleaning. Filter media cleaning is continuously taking place while the filter is in operation. Detailed description and operation of the MBSF can be found in earlier literature (Oesterholt & Bult 1993; Hultman et al. 1994; Loffill et al. 2010).

Sand movement is one of the most important operational parameters for MBSFs. Traditionally, the sand circulation rate is measured by counting the time taken for the movement of a long metal rod embedded with the sand. In this study we used an online monitoring tool called Sand-Cycle to monitor the sand circulation in the filter. This tool has been developed by BW Products BV, Netherlands, and is applied successfully in several MBSFs in the Netherlands. The system uses RFID (radio frequency identification device) tags to monitor the movement of sand grains in the MBSF. RFID tags are encapsulated in a specially designed cover which is resistant to harsh environmental conditions. A reading device mounted on the airlift pump identifies and counts the RFID tags passing it. The codes, dates and times of each reading passes to a host system application for data collection, logging, processing and transmission. The huge number of data generated 24/7 is used for statistical analysis of the operation conditions of the filter. Results of the analyses are visualized via Sand-Cycle's specially designed on-line dashboard (Wouters et al. 2016).

Several researchers have used MBSF for polishing of WWTP effluent to improve the water quality and it has been shown to be able to achieve low TP concentrations (Hultman et al. 1994; Newcombe et al. 2008). Jonsson et al. (1997) compared the MBSF with two-media downflow sand filter and concluded that both filters performed well for nutrient removal. Also, they reported less influence by nitrogen bubbles on the process of the MBSF compared to the two-media filter. Oesterholt & Bult (1993) compared the removal efficiency by MBSF and microfiltration. MBSF could achieve below 0.6 g P/m3 and microfiltration achieved lower than 0.2 g P/m3. Both systems had removal of 80–100% of suspended solids (SS).

The research presented here was focused on studying the applicability of MBSF for polishing treated WW to achieve very low TP and nitrogen concentrations in the effluent. Research targets were to obtain TP less than 0.1 mg/L and SS less than 10 mg/L in the treated effluent, and to remove 0.75 kg NO3-Neq per m3 sand per day, when the water temperature was above 10 °C and the COD/N ratio was less than 5 g COD/g NO3-Neq.

MATERIALS AND METHODS

Experiments were carried out at GRA. In addition to the people and industries in the area, GRA serves as the WWTP of the Oslo International Airport. The plant is running with a load of about 76,000 population equivalents and average flow of around 11,000 m3/day. The full-scale treatment process at GRA consists of mechanical pre-treatment, primary sedimentation, biological treatment by MBBR (moving bed biofilm reactor process with seven bioreactors in series), chemical coagulation with an aluminium-based coagulant (PAC, type ‘Ekoflock 90’) and flocculation with an anionic polymer (Zetag 4105), followed by dissolved air flotation (DAF) for particle separation (Rusten & Ødegaard 2007). GRA utilizes spent de-icing fluid (monopropylene glycol, MPG) runoff from the Oslo airport as an external carbon source for denitrification (DN) in the MBBR (Rusten et al. 1996; Hem et al. 2008).

The pilot MBSF was a cylindrical filter with the trade name ‘Dyna-Sand DST07’ provided by Nordic Water Products AB. The filter had a 0.7 m2 cross sectional area and 3.6 m3 maximum wet volume. It was filled with 1.8 m3 sand having 1.2–2 mm grain size. Filter bed height was 2.87 m. Figure 1 shows a schematic of the pilot plant setup. Either final effluent of the WWTP or different mixtures of this final effluent and effluent from the MBBR-stages prior to coagulation and solids separation was pumped to a 1 m3 IBC container and mixed well. WW from the container, the influent water to the pilot, was pumped to the MBSF. Filtered water was collected in a 1 m3 IBC tank and the wash-water was collected in a small container. Coagulant and carbon source were dosed directly in to the inlet pipe. A static mixture in the inlet pipe mixed the chemicals with influent WW. Flocculation occurred inside the influent pipe and the sand filter. Ethanol was consumed by the denitrification bacteria grown on sand surfaces inside the filter. The pilot plant was equipped with online measurements of influent flow (Siemens Sitrans FM Mag 5000), nitrate (Hach Lange N-ISE sc), dissolved oxygen (DO), temperature (Hach Lange-LDO®) and two pH sensors (Hach Lange).

Figure 1

Pilot plant set-up used at the Gardermoen WWTP.

Figure 1

Pilot plant set-up used at the Gardermoen WWTP.

Composition of the pilot influent was changed by mixing different amounts WWTP final effluent and effluent from the MBBR-stages during the testing period. Treated effluent from the WWTP (type 3) alone was used for further removal of P and N from treated WW (polishing). MBBR effluent from the reactor 7 (type 2) was mixed with type 3 water to add P and SS to the influent water. To provide adequate nitrate for the DN process, nitrified water from the MBBR reactor 5 (type 1) was mixed with type 3 water. In addition to that, the commercial product called Nutriox, having 113 g NO3-N/L was dosed as an external nitrate supply, to increase the nitrate loading when it was necessary. Aluminium based coagulant (PAC, type ‘Ekoflock 90’) was used for P removal studies. TP removal with different influent water qualities, coagulant doses and loading rates were tested.

Anoxic DN was tested with different influent water qualities, loading rates and COD/N ratios. Carbon source is the controlling factor of the structure and function of the denitrifying bacterial community (Xu et al. 2018). Municipal wastewater is rich with biodegradable soluble organic matter (BSCOD). However, Ekama et al. (1986) showed that the BSCOD concentration in WW reduces to approximately zero for wastewater biologically treated at 20 °C and with more than three days solids retention time. This is applicable to all the water types used for the present study, which have already undergone the full-scale MBBR treatment. Thus, an external carbon source, ethanol, was used to provide BSCOD for DN testing.

Influent, effluent and wash-water of the filter were sampled using three peristaltic pumps. Sampling pumps were run continuously from 11:00 h to 9:00 h of the next day. All the samples were analysed in a laboratory facility at the WWTP. Total COD (TCOD), filtered COD (FCOD), total P (TP), filtered TP (FTP), dissolved PO4-P (OP) and suspended solid (SS) were measured for all the samples. Additional measurements of total N (TN), filtered total N (FTN), NO3-N and NO2-N were performed on the samples of DN tests. Whatman GF/C glass fibre filters were used for filtration of samples and analysis of suspended solids (SS). SS was analysed according to Standard Methods SM 2540 D and E (Standard methods 2005). Phosphorus/nitrogen related parameters and COD were measured using Merck test cuvette kits, a Merck thermos-reactor Spectroquint TR 320 and a Merck Spectroquant Prove Spectrophotometer (Merck/Millipore-Sigma, USA and Canada). For sludge volume measurements, the Imhoff cone method was chosen over the graduated cylinder, due to the low SS concentrations in the wash-water. Sludge volume in the Imhoff cone was measured after 30 min and 24 h. Sand circulation was monitored by both manual measurements and online Sand-Cycle detections. Wash-water flow was measured using a calibrated container and a stop watch.

RESULTS AND DISCUSSION

Pilot-plant influent concentrations of the P removal and DN studies are given in Table 1. All the data are based on flow proportional samples collected during the test periods. Influent samples were taken downstream of where Nutriox was mixed into the wastewater and before adding the external carbon source. Nitrogen based parameters were not analyzed during the P removal studies.

Table 1

Summary of influent water quality of different TP removal and N removal studies

  SS mg/L TCOD mg/L FCOD mg/L TP mg/L FTP mg/L PO4-P mg/L TN mg/L FTN mg/L NO2-N mg/L NO3-N mg/L 
P removal WWTP Effluent (Type 3) Min. 10 0.19 0.019 <0.005     
Ave. 16 38 26 0.38 0.054 0.014     
Max. 31 66 40 0.64 0.177 0.056     
P removal Mixed WW (Type 3 + 2) Min. 12 35 16 0.23 0.035 <0.005     
Ave. 66 115 36 3.11 1.697 1.263     
Max. 169 306 56 7.29 4.230 4.081     
DN WWTP Effluent (Type 3) Min. 19 13 0.09 0.062 <0.005 0.03 5.3 
Ave. 27 22 0.27 0.084 0.055 12 11 0.05 8.3 
Max. 20 33 31 0.61 0.191 0.170 16 14 0.27 11.7 
DN Mixed WW (Type 3 + 1) Min. 14 15 0.2 0.050 <0.005 14 11 0.05 8.5 
Ave. 23 48 29 1.18 0.771 0.741 19 17 0.14 13.9 
Max. 57 148 105 3.12 2.380 2.340 27 24 0.27 20.3 
  SS mg/L TCOD mg/L FCOD mg/L TP mg/L FTP mg/L PO4-P mg/L TN mg/L FTN mg/L NO2-N mg/L NO3-N mg/L 
P removal WWTP Effluent (Type 3) Min. 10 0.19 0.019 <0.005     
Ave. 16 38 26 0.38 0.054 0.014     
Max. 31 66 40 0.64 0.177 0.056     
P removal Mixed WW (Type 3 + 2) Min. 12 35 16 0.23 0.035 <0.005     
Ave. 66 115 36 3.11 1.697 1.263     
Max. 169 306 56 7.29 4.230 4.081     
DN WWTP Effluent (Type 3) Min. 19 13 0.09 0.062 <0.005 0.03 5.3 
Ave. 27 22 0.27 0.084 0.055 12 11 0.05 8.3 
Max. 20 33 31 0.61 0.191 0.170 16 14 0.27 11.7 
DN Mixed WW (Type 3 + 1) Min. 14 15 0.2 0.050 <0.005 14 11 0.05 8.5 
Ave. 23 48 29 1.18 0.771 0.741 19 17 0.14 13.9 
Max. 57 148 105 3.12 2.380 2.340 27 24 0.27 20.3 

Used water type shown within brackets.

Water temperature varied from 11 °C to 22 °C during P removal tests and between 9 °C and 19 °C during the DN testing. MBSF was running with sand circulation rates between 6 and 10 mm per min during the studies. Wash-water volume was controlled between 8% and 10% of the influent flow. Hydraulic filtration rate ranged between 4.2 and 9.9 m/h.

Phosphorus removal

P removal was tested both with and without coagulation in the filter, with WWTP effluent alone or WWTP effluent water mixed with WW from either the MBBR reactor 7 or MBBR reactor 5 of the WWTP. WWTP effluent water had SS concentrations from 8 to 31 mg/L (average 16 mg/L). Water temperature varied between 11 °C and 17 °C (average 14 °C) during the testing period. Sand circulation rate varied between 6.2 and 9.8 mm/min (average 7.8 mm/min). During the first few days of testing we had poor control of the wash-water from the filter and average wash-water flow was 730 L/h (equal to 13% of the average influent water), with very large variations and a maximum of 1,496 L/h.

Figure 2(a) and 2(b) show the TP removal when testing WWTP effluents with and without coagulants. Graph (A) is TP removal versus influent TP concentrations. Graph (B) shows influent TP and effluent TP concentrations. The green line indicates the targeted effluent TP limit in mg/L. Influent TP concentrations varied between 0.19 mg/L and 0.64 mg/L, with an average of 0.38 mg/L. Except for one sample, the coagulant treated samples achieved effluent results well below the expected limit. The majority (87%) of the samples without coagulant also achieved below 0.1 mg TP/L with only filtration. We found that the average FTP and OP in influent samples were as low as 0.054 and 0.014 mg/L, respectively. It shows that almost all the dissolved P had already been converted to particulate form by the coagulant used in the DAF stage of the full-scale WWTP. The results indicate that, with good particle removal, very low levels of TP can be achieved even without further coagulant addition, and an average TP concentration well below the 0.1 mg/L limit can be reached.

Figure 2

TP removal by the MBSF when testing WWTP effluents with and without coagulant. All the samples were 22 h flow proportional samples. (a) TP removal versus influent concentrations in mg/L. The line indicates 100% removal. (b) Influent and effluent TP concentrations. Green horizontal line indicates the expected effluent TP limit. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wst.2019.266.

Figure 2

TP removal by the MBSF when testing WWTP effluents with and without coagulant. All the samples were 22 h flow proportional samples. (a) TP removal versus influent concentrations in mg/L. The line indicates 100% removal. (b) Influent and effluent TP concentrations. Green horizontal line indicates the expected effluent TP limit. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wst.2019.266.

By mixing water types, the P and SS concentrations in the influent increased. Water temperature varied between 13 °C and 22 °C (average 16 °C) during the testing period. Sand circulation rate varied between 7 and 10 mm/min (average 8.3 mm/min) and wash-water was 486 to 632 L/h with an average of 566 L/h (equal to 10% of the average influent water flow).

Figure 3(a) and 3(b) show TP removal when testing mixed WW types with and without coagulants. Graph 3(A) is TP removal versus influent TP concentrations during the tests and graph 3(B) shows influent TP and effluent TP concentrations. Influent TP concentrations varied between 0.23 and 7.29 mg/L (average 3.11 mg/L). Unlike WWTP effluent alone, FTP and OP concentrations were high and varied from 0.04 to 4.2 mg/L (average 1.7 mg/L), and from 0.01 to 4.1 mg/L (average 1.26 mg/L), respectively. Pilot effluent TP concentration for the tests without coagulation varied between 0.08 and 3.6 mg TP/L, where the removal efficiency varied from 10% to 69%, except for two samples with very low influent TP concentrations, having 91% and 95% removal. With coagulation, the pilot effluent TP concentration varied from 0.04 to 3.9 mg TP/L, which is equivalent to 36% to almost 99% TP removal.

Figure 3

TP removal by the MBSF when testing mixed WW with and without coagulant. All the samples were 22 h flow proportional samples. (a) TP removal versus influent concentrations in mg/L. (b) Influent and effluent TP concentrations. Numbers show the coagulant dose (mg Al/L). Green horizontal line indicates the expected effluent TP limit. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wst.2019.266.

Figure 3

TP removal by the MBSF when testing mixed WW with and without coagulant. All the samples were 22 h flow proportional samples. (a) TP removal versus influent concentrations in mg/L. (b) Influent and effluent TP concentrations. Numbers show the coagulant dose (mg Al/L). Green horizontal line indicates the expected effluent TP limit. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wst.2019.266.

Influent flow varied from 4 m3/h to 7 m3/h during both tests with and without coagulant. Different coagulant doses from Al/TP molar ratio 0.52 to 24.2 were tested during these tests. Al dose (mg Al/L) related to each sample is shown in the two graphs of Figure 3. It was clearly observed that inadequate Al dose with increasing influent TP concentrations resulted in poor effluent qualities. Effluent limits below 0.1 mg TP/L were achieved when the influent concentrations were low and sufficient amounts of coagulant were added. The low removal was obtained with high influent TP concentrations and low Al/TP molar ratios. An Al/TP molar ratio of at least 4 was necessary to achieve <0.1 mg TP/L.

Denitrification

Biological denitrification is a microbial process of reducing nitrate to gaseous forms of nitrogen. This process is facilitated in anoxic conditions and when adequate organic substrate (BSCOD) is available in the WW. Due to low cost, methanol is commonly used as external carbon source for DN. However, researchers have reported ethanol as the best carbon source over methanol due to the significantly shorter start-up time and higher denitrification rates (Rusten et al. 1996; Rusten & Ødegaard 2007). Therefore, an ethanol solution, having 1,681 g FCOD/L, was used as external carbon source for this study. BSCOD in the influent will be consumed for reducing not only NO3-N, but also NO₂-N and DO inside the reactor. Therefore, we used the conversion described by Koopman et al. (1990) to calculate an equivalent concentration of NO3-N (NO3-Neq) by converting influent NO2-N and DO to equivalent amounts of NO3-N. For this conversion, 1.0 mg NO2-N is considered as 0.6 mg NO3-Neq and 1.0 mg DO is considered as 0.35 mg NO3-Neq.

The ratio between BSCOD and available nitrogen is one of the key parameters for biological nitrogen removal. The minimum theoretical organic matter requirement for anoxic DN is 2.86 g COD/g NO3-Neq. But in practice the COD/NO3-Neq requirement is always higher than the theoretical value due to the amount of COD used for biomass growth (Rusten et al. 1996; Fu et al. 2009). The amount of FCOD added as external carbon source and the amount of NO3-Neq were used for COD/N calculations in this study. Using COD/N ratios expressed as g COD/g NO3-Neq is especially useful when comparing substrate consumption ratios in systems run with different influent DO concentrations.

Figure 4 shows (A) denitrification rates versus loading rates, and (B) COD/N ratios versus DN rates and influent PO4-P concentrations during the tests with both WWTP effluent and mixed WW. Indicating the influent NO3-N limitations, it is clearly observable that the removal rate is increasing with the increasing loading rates. MBSF with high loading rates achieved very good DN rates with a maximum rate up to 1.4 kg NO3-Neq per m3 sand per day (kg NO3-Neq/m3-d) with a COD/N ratio of 4.0 g CODadded/g NO3-Neq. This rate was far higher than the research goal, which was 0.75 kg NO3-Neq/m3-d at temperatures over 10 °C and COD/N ratio less than 5 g CODadded/g NO3-Neq. From the COD/N ratio of 4.0 g CODadded/g NO3-Neq and 4.3 g CODconsumed/g NO3-Neq removed, a sludge yield for the biological DN process of 0.33 g CODbiomass/g COD removed can be calculated, using the procedure in Rusten et al. (1996).

Figure 4

DN rates versus NO3-Neq loading, C/N ratios and influent PO4-P concentrations for DN tests with both WWTP effluent and mixed WW. All the samples were 22 h flow proportional samples. Numbers indicate the COD/N ratios of the influent (after ethanol addition).

Figure 4

DN rates versus NO3-Neq loading, C/N ratios and influent PO4-P concentrations for DN tests with both WWTP effluent and mixed WW. All the samples were 22 h flow proportional samples. Numbers indicate the COD/N ratios of the influent (after ethanol addition).

There are many factors like water temperature, loading rates, nutrient and carbon source availability and operational conditions, that can influence the DN rate. DO concentrations in tested WWTP effluent water varied between 5.2 and 8.6 mg O2/L and in tested mixed WW was between 5.8 and 8.6 mg O2/L. DO concentration in the filter effluent samples of DN tests was always measured near zero. We have compensated the influence of DO on DN rates by converting the DO to NO3-Neq. Water temperature varied from 10.7–16.1 °C and 8.7–18.7 °C for tests with WWTP effluent and mixed WW, respectively. Hydraulic filtration rate during testing ranged between 4.2 and 7.1 m/h. Sand circulation was maintained between 7.6 and 10.9 mm/min during all the DN tests. Though all these factors may influence the DN rates the influence was comparatively small and we could not see a clear relationship.

Some researchers have reported reduced DN performance with COD/N ratios below about 3.8 or 4 g COD/g NO3-Neq (Rusten et al. 1996; Rusten & Ødegaard 2007; Rathnaweera et al. 2018). Other than that, the microbial activities can also be limited due to the limitation of bio available P in the influent WW. Hultman et al. (1994) observed limited DN rate when the concentration was below 0.1 mg/L soluble P in the effluent from the filter.

In Figure 4(b), NO3-Neq removal rates and influent PO4-P concentrations are plotted against the COD/N ratio. Low removal rates in the left part, where the COD/N ratio is less than 4 and the maximum OP is 0.17 mg PO4-P/L (inside box with a continuous line), can be explained by low COD/N ratio and low OP in the influent water. Low DN in the area marked with a box with broken line, except the samples marked with green circle, is OP limitation. The limiting factor of the samples with green circles and in the other part of the graph could be either low NO3-N load or poor biological activities, but for most cases low effluent NO3-N concentrations limited the DN rate. It was clearly observed that the DN rate was limited when the influent water had less than 0.4 mg PO4-P/L. Looking at the results for mixed WW, and eliminating all tests with effluent concentrations below 1.0 mg NO3-N/L (considered NO3-N limited), average DN rates were only 0.51 kg NO3-Neq/m3-d for tests with influent OP <0.4 mg PO4-P/L (average influent 0.1 mg PO4-P/L) and as high as 1.30 kg NO3-Neq/m3-d for tests with influent OP >1.2 mg PO4-P/L (average influent 1.8 mg PO4-P/L).

Particulate P was on average 77% of TP in WWTP effluent and 45% of TP in mixed WW. A portion of dissolved P is consumed by the bacteria for their functional uses and converted to particulate form. Most of the particulate P is removed by the filtration process. Figure 5 shows TP removal versus influent TP concentrations for DN tests. Effluent TP for tests with WWTP effluent varied from 0.06 mg/L to 0.31 mg/L (average 0.15 mg/L) and for tests with mixed WW varied from 0.09 mg/L to 2.21 mg/L (average 0.74 mg/L). It is clear that the effluent TP concentration increased with increasing influent concentrations.

Figure 5

TP removal versus influent TP concentrations for DN tests. All the samples were 22 h flow proportional samples.

Figure 5

TP removal versus influent TP concentrations for DN tests. All the samples were 22 h flow proportional samples.

Suspended solids

Continuous sand flushing system of the filter functioned well. Figure 6 shows the particle removal during P removal and DN tests. Figure 6(a) shows good particle removal during the P removal tests with both WWTP effluent and mixed WW and with and without coagulants. Average removal of 65% (without coagulation) and 66% (with coagulation) were achieved with WWTP effluent. With mixed WW 79% (without coagulation) and 87% (with coagulation) were achieved. Pilot effluent SS concentrations for the tests with WWTP effluent varied between 2 and 11 mg/L, and effluents from mixed WW tests varied from 1 to 19 mg SS/L. Figure 6(b) shows that all the effluent samples, except one sample, of the tests with WWTP effluent achieved the targeted limit of 10 mg SS/L. The majority of the samples from the mixed WW tests also were below the targeted limit, including all samples with influent concentrations less than 70 mg SS/L.

Figure 6

SS removal with increasing inlet concentrations during P removal (a) and (b) and DN (c) and (d) testing.

Figure 6

SS removal with increasing inlet concentrations during P removal (a) and (b) and DN (c) and (d) testing.

Figure 6(c) shows the SS removal during DN tests. SS removal varied between −1.5 mg SS/L and 48 mg/L (average 15 mg SS/L). The scattered distribution including negative removal was obtained due to the rapid biofilm growth during DN with external carbon source. Influent SS varied from 1 to 58 mg SS/L and effluent was between 1 and 17 mg SS/L, as seen in Figure 6(d). However, indicating good functioning of the sand wash system, maximum removal was 87.5% (average 44%) with WWTP effluent and 96% (average 67%) with mixed WW.

Sludge volume (SV) of the wash-water was measured during the tests. SV indicates the tendency of the sludge solids to settle and thicken during the sedimentation/thickening process. Table 2 shows specific sludge volumes after 30 min settling and 24 h thickening in Imhoff cones. Values are shown for tests with coagulation (including both WWTP effluent and mixed wastewater) and tests with DN (including both WWTP effluent and mixed wastewater), respectively. Average and median wash-water SV after 30 min of settling was slightly larger for coagulation tests than for DN tests. However, after 24 h thickening the average and median SV was significantly higher for the coagulation tests than for the DN tests. The scatter was also considerably larger for the coagulation tests, shown by the high standard deviation. The median sludge concentration after 24 h was 17 g SS/L for the coagulation tests and 31 g SS/L for the DN tests, indicating significantly better gravity thickening potential for the DN tests sludge. However, in a shallow Imhoff cone with no polymer addition and no slow mixing both these results are quite good and indicate that both sludges may thicken fairly well in a full-scale plant.

Table 2

Sludge volume measurements and concentrations during the tests with coagulant and the tests with DN

 Specific sludge volume (mL/g SS)
 
Wash-water sludge concentration after 24 h thickening (g SS/L) 
After 30 min
 
After 24 h
 
Avg. ± St. D. Median Avg. ± St. D. Median Median 
With coagulation 117 ± 37 122 59 ± 30 60 17 
With DN 102 ± 33 101 32 ± 9 32 31 
 Specific sludge volume (mL/g SS)
 
Wash-water sludge concentration after 24 h thickening (g SS/L) 
After 30 min
 
After 24 h
 
Avg. ± St. D. Median Avg. ± St. D. Median Median 
With coagulation 117 ± 37 122 59 ± 30 60 17 
With DN 102 ± 33 101 32 ± 9 32 31 

Sand-Cycle as a tool for MBSF operation

Online monitoring of sand circulation was a very convenient tool for operation of the MBSF. Figure 7 shows the operator interface (dashboard) of the Sand-Cycle system.

Figure 7

Sand-Cycle dashboard on 01.03.2018. Blue line is the average sand circulation measurement. Two upper and lower lines indicate the standard deviation of the sand circulation measurements. Double arrow on left shows the period with good filter run and the arrow right is a period with trouble in the filter. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wst.2019.266.

Figure 7

Sand-Cycle dashboard on 01.03.2018. Blue line is the average sand circulation measurement. Two upper and lower lines indicate the standard deviation of the sand circulation measurements. Double arrow on left shows the period with good filter run and the arrow right is a period with trouble in the filter. Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wst.2019.266.

The Sand Circulation is determined by the tag measurements and the turnaround time of a particular tag. The circulation speed is presented as 4-h moving averages in millimetres per minute. Sand-Cycle online measurements were compared with the manual sand movement measurements of the pilot plant. The comparison showed fairly good agreement between Sand-Cycle and manual measurements.

Active Bed Volume is a calculation related to the number of tags that have been identified during the last 24 h and it indicates the percentage of the filter bed actively involved in the treatment process. During the time of Figure 7, only 26 tags out of a total of 31 tags were identified and it was calculated as 83% of active bed volume. The date and time shown is the time and date where the last tag was identified. Filter homogeneity is the fraction of tags having similar time intervals compared to the average interval. RFID tags with deviating intervals results in the homogeneity being lower than 100%. Homogeneity shown in Figure 7 is 72% and this indicate that the sand movement inside the filter is unequal at the time the dashboard picture was taken.

Figure 7 shows the filter status during 48 h from 27th February to 1st March 2018. The dark blue line shows the real-time sand circulation and the grey area covered by light grey lines indicates the asymmetrical standard deviation profile of sand circulation measurements. The first part of the graph, indicated with a green arrow, was the time where the filter ran with normal sand circulation. During undisturbed run, sand circulation did not fluctuate, and the standard deviation was small. During the second half of the graph, a part of the sand bed started to clog and stopped moving. This influenced the tag measurements and it started to show uneven sand circulation with expanded standard deviations in the graph. At the same time, it indicated reduced homogeneity. This was a robust tool for early identification of filter problems and it gave a possibility to take remedial actions before so much of the sand bed clogged that the filtration had to stop. Sampling and chemical analyses were not carried out during the brief periods with poor sand movement.

CONCLUSIONS

The MBSF was successfully evaluated for the ability to achieve very low TP concentrations, low SS concentrations and good denitrification when treating municipal wastewater.

When treating WWTP effluent (<0.7 mg TP/L) the goal of <0.1 mg TP/L in the effluent was easily achieved with coagulation prior to the MBSF. Even without coagulation (only filtration) the majority (87%) of the samples achieved below 0.1 mg TP/L, and the average was as low as 0.075 mg TP/L.

Treating mixed WW (<8 mg TP/L) very good TP removal was possible with adequate coagulant dose and good filtration. An Al/TP molar ratio of at least 4 was needed to achieve 0.1 mg TP/L in the MBSF effluent. Without coagulant, only particulate P was removed, and TP removal efficiency was low.

Testing the MBSF for DN and using ethanol as external carbon source, a critical COD/N ratio of 4.0 g CODadded/g NO3-Neq was found for full denitrification. The sludge yield calculated from the consumption of external carbon source was 0.33 g CODbiomass/g COD consumed. MBSF with high loading rates achieved very good DN rates with a maximum rate of 1.4 kg NO3-Neq/m3-d with a COD/N ratio of 4.0 g COD/g NO3-Neq. This removal rate was far higher than the research goal, which was 0.75 kg NO3-Neq/m3-d at temperatures above 10 °C and COD/N ratio less than 5 g COD/g NO3-Neq.

Many factors like water temperature, loading rates, nutrient and carbon source availability and operational conditions will influence the DN rate. Influence of influent DO was compensated by converting it to NO3-Neq. Many runs were rate limited by low effluent NO3-N concentrations, so it was not possible to see any temperature effects. However, it was clearly observed that the DN rate was limited when the influent to the MBSF had less than 0.4 mg PO4-P/L. Looking at the results for mixed WW and eliminating all tests with effluent concentrations below 1.0 mg NO3-N/L (considered NO3-N limited), average DN rates were only 0.51 kg NO3-Neq/m3-d for tests with influent OP <0.4 mg PO4-P/L (average influent 0.1 mg PO4-P/L) and as high as 1.30 kg NO3-Neq/m3-d for tests with influent OP >1.2 mg PO4-P/L (average influent 1.8 mg PO4-P/L).

The majority of the samples from the P removal tests with both WWTP effluent and mixed WW, achieved the targeted limit of below 10 mg SS/L, with an average of 5.9 mg SS/L in the MBSF effluent. Though the biofilm growth increased the biosolids concentration in the MBSF during DN testing, SS removal was good with an average effluent concentration of 6.4 mg SS/L. Sludge volume measurements of the wash-water showed a better thickening potential for the sludge from DN testing compared to the sludge from the tests with coagulant addition.

Online monitoring of sand circulation was a very convenient tool for operation of the MBSF.

ACKNOWLEDGEMENT

The authors acknowledge the valuable technical assistance from Mr Morten Kjeverud, Mr Egil Sønsterud, Mr Arne Petter Toverud, Ms Mathilde Francis and all the operators at the Gardermoen WWTP, valuable help from Aquateam COWI AS and Nordic Water Products AB. This project was financed by grant no. 245655 from the Regional Research Council of the Capital of Norway, Gardermoen WWTP and Nordic Water Products AB.

REFERENCES

REFERENCES
APHA/AWWA/WEF
2005
Standard Methods for the Examination of Water and Wastewater
,
21st edn
.
American Public Health Association (APHA)/American Water Works Association/Water Environment Federation
,
Washington, DC
,
USA
.
Fylkesmann
2015
Ny utslippstillatelse på avløpssektoren for Ullensaker kommune (Discharge permit from the local pollution control authority in Norwegian). https://www.norskeutslipp.no/Templates/NorskeUtslipp/Pages/company.aspx?id=61&CompanyID=17703&epslanguage=no&SectorID=100
.
Hem
L. J.
,
Rusten
B.
&
Skjefstad
J.
2008
Biological degradation of spent de-icing fluids in a municipal wastewater treatment plant – Experiences and challenges. Paper No. 2007-01-3349. SAE 2007 Transactions Journal of Aerospace. March 2008
.
Hultman
B.
,
Jonsson
K.
&
Plaza
E.
1994
Combined nitrogen and phosphorus removal in a full-scale continuous up-flow sand filter
.
Water Science and Technology
29
(
10–11
),
127
134
.
Jonsson
L.
,
Plaza
E.
&
Hultman
B.
1997
Experiences of nitrogen and phosphorus removal in deep-bed filters in the Stockholm area
.
Water Science and Technology
36
(
1
),
183
190
.
Koopman
B.
,
Stevens
C. M.
&
Wonderlick
C. A.
1990
Denitrification in a moving bed upflow sand filter
.
Research Journal of the Water Pollution Control Federation
62
(
3
),
239
245
.
Loffill
E.
,
Alkhaddar
R. M.
,
Phipps
D. A.
&
Faram
M. G.
2010
Development of residence time distribution measurement techniques to improve reliability and accuracy, NOVATECH 2010, GRAIE, Lyon, France
.
Rathnaweera
S. S.
,
Rusten
B.
,
Korczyk
K.
,
Helland
B.
&
Rismyhr
R.
2018
Novel biofilm reactor for denitrification of municipal wastewater
.
Water Science and Technology
78
(
7
),
1566
1575
.
Rusten
B.
,
Wien
A.
&
Skjefstad
J.
1996
Spent Aircraft Deicing Fluid as External Carbon Source for Denitrification of Municipal Wastewater – From Waste Problem to Beneficial Use
. In:
51st Purdue Industrial Waste Conference Proceedings. Ann Arbor Press, Inc.
,
May 6–8, 1996
, pp.
359
369
.
Wouters
J. W.
,
Heslegrave
C.
,
Sutman
P.
&
Van
O. E.
2016
Making the most of continuously moving bed polishing (bio)filters for N and P removal, using sand-cycle remote control
. In:
Proceedings, 10th European Waste Water Management Conference
,
11–12 October 2016
,
Manchester, UK
, pp.
36
43
.