Potential of off-gas analyses

Abstract

This chapter demonstrates how and why a larger variety of variables can be monitored with a single off-gas sampler and analyzer on sequentially operated compared to continuously fed and aerated reactors. Measurements on a full-scale sequentially operated aerobic granular sludge reactor treating municipal wastewater were used to illustrate this potential and new insights were obtained on this specific process by doing so. The oxygen (O2), carbon dioxide (CO2), methane (CH4) and nitrous oxide (N2O) liquid-gas transfer rates were calculated and their cyclic profiles were explained by the processes in the liquid phase.

Aeration characteristics were calculated and their gradual improvement within cycles was explained by surfactants degradation. Liquid phase concentrations (e.g. of N2O) could be estimated with a novel calculation procedure (based on equations derived in Chapter V). The O2 consumption rate could be estimated because no convective transport of O2 occurred via the liquid phase during aeration, while this would form a large fraction of the mass balance in continuously operated reactors. Greenhouse gas emissions were calculated for the whole monitoring period. The CO2 emission and O2 absorption per cycle could be used to obtain information on the TOC/COD ratio and COD/N ratio of the influent that entered each batch.

Finally, the sludge production could be estimated from the O2, nitrogen, nitrate and COD balances. Not only could some of these variables only be calculated due to the sequential operation, but they also were more representative for the whole reactor, since the absence of feeding during aeration creates a more homogeneous off-gas composition.

Introduction

In sequentially operated reactors with unaerated feeding phases, no convective transport occurs via the liquid phase during the other phases. Therefore, changes in the concentration of ammonium, nitrate or phosphate directly reflect their conversions (Pronk et al., 2015). Liquid phase measurements are thus useful for monitoring, but they can also be used for control. For example, a liquid phase oxygen sensor is used to control the aeration rate to obtain a fixed oxygen set-point or predefined cyclic profile that is known to support good nitrogen removal (Lochmatter et al., 2013). Ammonium measurements can also be used to adjust this oxygen set-point automatically to keep the nitrogen removal optimal when the granule size or influent composition changes (Isanta et al., 2013, Kagawa et al., 2015, Pronk et al., 2015). Finally, the length of the aeration phase can be automatically adjusted based on liquid phase oxygen sensors in combination with frequent non-aerated periods, during which the drop in the oxygen concentration reflects the consumption rate. This can be used to detect when ammonium is nearly completely oxidized (Dobbeleers et al., 2017a, Stes et al., 2018).

Liquid phase measurements are thus typically used for monitoring and control, similar to other wastewater treatment processes, but there is an untapped potential for off-gas measurements, especially in case of sequentially operated reactors, with or without granular sludge.

Off-gas analyses can provide insight into the processes occurring inside a reactor.

The difference between the composition of the aeration gas (usually air) and off-gas is indicative for the liquid-gas transfer rate, which is in turn affected by the consumption and production of volatile substances. For example, oxygen (O2) is absorbed by the liquid phase and subsequently consumed by biological oxidation processes, whereas carbon dioxide (CO2) is generally emitted, because it is produced during heterotrophic growth, while consumption by nitrifiers is minimal (Hellinga et al., 1996). Nitrous oxide (N2O) and nitric oxide (NO) can also be emitted, as they are possible by-products of nitrification and denitrification (Kampschreur et al., 2009). Finally, methane (CH4) can be emitted since it enters the reactor from the sewers or the sludge handling liquors (Daelman et al., 2012). Overall, off-gas cannot only be used to monitor the aeration performance (Gillot et al., 2005b, Redmon et al., 1983), but it can also reveal information on the wastewater composition, conversions occurring inside the reactor (Hellinga et al., 1996, Leu et al., 2010, Weissenbacher et al., 2007) and allows quantification of pollutant emissions (Bollon et al., 2016, Caniani et al., 2019, Daelman et al., 2012, Daelman et al., 2013b, Marques et al., 2016, Prata et al., 2018).

Off-gas measurements have three advantages over liquid phase measurements. A gas sample is well mixed and it has been in contact with a large volume of water. In contrast, liquid phase sensors and analyzers measure the concentration at a specific location in the reactor (Mears et al., 2017). Gas phase analyses also typically do not require chemical reagents, e.g. oxygen analysis can rely on its paramagnetic properties and carbon dioxide,

nitrous oxide and methane analysis on their infrared absorption. Automated ammonium and phosphate measurements on the other hand often require reagents (Vanrolleghem and Lee, 2003). Finally, maintenance is minimal because the gas phase contains less fouling and corrosive substances compared to the wastewater itself (Hellinga et al., 1996). For these reasons, on-line gas phase analysis is used in the fermentation industry for monitoring and control (Mears et al., 2017, Sonnleitner, 2013).

Despite their potential, off-gas analysers are not part of the typical instrumentation applied for monitoring and control of conventional treatment plants with continuous feeding and aeration (Cornelissen et al., 2018a, Olsson et al., 2005). The liquid flow through a reactor typically produces spatial concentration gradients, causing different gas phase concentrations for different sampling locations (Amerlinck et al., 2016, Caniani et al., 2019, Leu et al., 2010, Nogita et al., 1982, Rosso et al., 2011). Therefore, a single sampler is not sufficient to capture the behaviour of the whole aerated surface. Moreover, simultaneous feeding and aeration makes it hard to extract reaction rates Ṙ^Li (g.min^-1; i refers to the reacting substance, for example oxygen) from mass balances of volatile substances over a reactor (Eq. 6.1).

Simultaneous feeding and aeration: ^dCi

LV^L

dt =(ṁ_in,i^L -ṁ_out,i^L )-ṁ_i^L-G(t) -Ṙ_i^L(t) Eq. 6.1 Not only the liquid-gas transfer rates ṁ^L-Gi (g.min^-1) and liquid phase concentrations C^Li

(g.m^-3), but also the liquid phase mass flow rates, ṁ^Lin,i and ṁ^Lout,i (g.min^-1) through the liquid volume V^L (m³) need to be known or measured. In contrast, sequentially operated reactors with unaerated feeding, such as aerobic granular sludge reactors, have most of the

reactions occurring during a separate reaction phase. Aeration without simultaneous feeding creates a well-mixed liquid phase and therefore also the off-gas composition is the same for different places above the reactor. The absence of convective flow also simplifies the mass balances (Eq. 6.2), making the estimation of reaction rates of volatile substances such as oxygen easier and more accurate (Figure 6.1).

Separated feeding and aeration: ^dCi

LV^L

dt =-ṁ_i^L-G(t) -Ṙ_i^L(t) Eq. 6.2 Given the advantages of off-gas analyses in general and the advantages for sequentially operated reactors in particular, this contribution presents on-line measurements of oxygen, carbon dioxide, nitrous oxide and methane in the off-gas of a full-scale aerobic granular sludge reactor and discusses potential applications. First, the liquid-gas transfer rates are analysed in detail and variations are explained by processes occurring in the liquid phase.

Then, the on-line estimation of other variables is demonstrated, i.e. aeration characteristics, liquid phase concentrations and reaction rates. Next, it is demonstrated that integration of the liquid-gas transfer rates over longer periods can give information on the greenhouse gas emissions, wastewater composition and sludge production. Finally, practical applications for everyday monitoring and control are formulated based on the results.

Figure 6.1. Conceptual difference between off-gas analysis with a single sampler applied to a continuously fed and aerated reactor (left) and a sequentially operated reactor during the aeration phase (right).

Methodology

3.1. Sampling location

A four and a half day monitoring campaign was performed from 31 July to 4 August 2017 at one of the three 1250 m³ Nereda® aerobic granular sludge reactors at the municipal wastewater treatment plant in Dinxperlo (NL) of the water board Rijn en IJssel. On average, the plant treats 3370 m³.d⁻¹ with respective concentrations of organics, total nitrogen and total phosphorus of 538 g COD.m^-3, 48 g N.m^-3 and 7 g P.m^-3 in 2017. The total suspended solids in the mixed liquor constituted 11 g TSS.L^-1 at the time of the measurements, of which 84%

had a diameter above 0.2 mm. The reactor bottom was covered with fine-bubble diffusers.

A typical dry weather cycle of the reactor takes about 5.5 hours (Figure 6.3). First, simultaneous upward feeding and discharge takes place without aeration for one hour. Next, the water level was decreased 15 cm to ±7.35 m during a twenty minutes period to compensate for the gas hold-up during the subsequent 3.5 hour reaction phase. The reaction phase was subdivided into a period with strong and continuous aeration to favour nitrification in particular, followed by a period with weaker, intermittent aeration to stimulate denitrification of the accumulated nitrate while keeping reactor contents mixed. The intermittent aeration was automatically started once the ammonium concentration reached its set-point. The cycle ended with a ±45 minute period during which settling and sludge withdrawal took place and to match the cycles of the three parallel reactors. The cycle length and influent volume per batch were automatically adapted to the hydraulic and pollutant load conditions. Phosphorus was mainly removed biologically during the reaction period, but iron chloride was also

automatically dosed inside the reactor or in the sand bed afterwards to complement the biological phosphorus removal when necessary.

3.2. Measurements

A floating hood was placed upon the reactor. The hood had an area of 0.55 m² to collect off-gas from the 170 m² water surface. The dead volume inside the hood was reduced with polyurethane foam, until about 0.1 m³ would remain once placed upon the water surface (Figure 11.1 and Figure 11.2 in the Appendix of Chapter VI). Off-gas was sampled from the hood and sent through a cooler to dry before entering an on-line analyzer to measure the mole fractions of oxygenand carbon dioxide (NGA 2000 MLT1 by Rosemount, Emerson) and methane and nitrous oxide (Xentra Continuous Emission Analyzer 4900 by Servomex). The on-site atmosphere was analyzed for 5 minutes every hour to account for changes in the composition of the aeration air, which can be significant for carbon dioxide, methane and nitrous oxide (Figure 11.3 in the Appendix of Chapter VI) due to local production from other unit processes on the wastewater treatment plant or nearby roads, industry, farmland etc. The atmospheric temperature, pressure and relative humidity were also monitored (Bosch BME280).

Standard on-line monitoring data were used to supplement the off-gas analysis. The liquid phase temperature, dissolved oxygen (LDO, Hach), ammonium (Amtax, Hach), nitrate plus nitrite (Nitratax, Hach) and phosphate concentration (Phosphax, Hach) were measured inside the reactor. The rotational speed of the three positive displacement blowers (Aerzen Blower Delta Hybrid/D12S), the position of the valves between the blowers and the air diffusers and the influent flow rate were logged as well.

3.3. Calculations Gas flow rates

The volumetric air flow rate (m³.min^-1) at the intake of each of the three blowers was calculated every half minute based on their monitored frequency of rotation, using specifications provided by the supplier. The total intake of air for the three aerobic granular sludge reactors was calculated as the sum of the three blowers’ intakes. The part of this total flow that reached the reactor under consideration Q^Gin (m³.min^-1) was obtained assuming a perfect separation in two equal parts at every junction and considering no flow through valves that were closed at that moment (Figure 6.2). Finally, the off-gas flow rate Q^Gout (m³.min ^-1) was calculated by correcting the air flow rate towards the reactor with Charles’ law (Eq. 6.3), neglecting any changes in the molar flow rate due to nitrogen gas stripping (Redmon et al., 1983).

Q_out^G (t) = Q_in^G(t)T_reactor^L (t)

T_atm^G (t) Eq. 6.3

The off-gas was assumed to have the same temperature as the reactor T^Lreactor (K) by the time it reached the surface, which is reasonable due to the thorough heat transfer with fine bubble aeration and high water column.

Figure 6.2. Simplified scheme of the gas flows at the plant.

Conversion of mole fractions to concentrations

Gas phase mole fractions x^Gi (mole.mole^-1) were converted to concentrations C^Gi obtain the atmospheric concentrations C^Gin,i (g.m^-3), while the measured reactor temperature T^Lreactor was used for the off-gas concentrations C^Gout,i (g.m^-3).

The mole fractions x^Gi in this expression (Eq. 6.4) were derived from the measured values x^Gmeasured,i after several corrections (Eq. 6.5).

x_i^G(t)=x_measured,i^G (t+τ)p_atm,cal^G

p_atm^G (t)(1-RH^G(t)133.322∙10^{8.0727 -}

1732.32 T^G(t)-39.466

p_atm^G (t) ) Eq. 6.5

First, a correction for a pure time delay τ (min) was applied, which originates mainly from the residence time of gas in the reactor and in the tubing. This delay was fixed at 3 min for the off-gas since it took this long on average before the measured oxygen fraction started dropping when the aeration phases started. No time delay was used for the atmospheric sampling because of the short tubing. Secondly, the pressure ratio was introduced in Eq. 6.5 to undo the fluctuations of the signal (Figure 11.4 in in the Appendix of Chapter VI) by compression or

Blower

decompression inside the analyzer when the atmospheric pressure p^Gatm lowered or increased compared to its value during calibration p^Gatm,cal (Pa). The third factor corrects for the water vapour that condensed in the cooler before analysis, taking into account the relative humidity in the sampled gas RH^G (-) and the equilibrium water content calculated with the empirical Antoine equation (Shriver and Drezdzon, 1986). The reactor temperature and 100% relative humidity were used for the off-gas sampling, as suggested by Redmon et al. (1983), whereas the measured atmospheric temperature and relative humidity were used for the atmospheric sampling. Finally, the missing mole fractions in the off-gas during the atmospheric sampling were filled up by linear interpolation and vice versa.

Liquid-gas transfer rates

There were two main assumptions for the calculation of the liquid-gas transfer rates:

the measured off-gas concentrations were representative for the complete reactor surface and the gas phase was in pseudo steady-state. The former was motivated by the good mixing during the aeration phases and the latter was based on the short residence time of the gas in the reactor relative to the interphase mass transfer and biological conversions rates (Chapter V). In this case, the liquid-gas transfer rate of component i, ṁ^L-Gi (g O2, CO2, CH4 or N2O.min

-1), could be expressed as the difference between the mass flow out of and into the reactor via the gas phase (Eq. 6.6).

ṁ_i^L-G(t) = Q_out^G (t)C_out,i^G (t)−Q_in^G(t)C_in,i^G (t) Eq. 6.6 A positive liquid-gas transfer rate is a stripping/emission rate. If it is negative, the sign can be reversed to obtain a gas-liquid transfer or absorption rate ṁ^G-Li.

Aeration characteristics

The overall volumetric liquid-gas transfer coefficient KLai (d^-1) of a substance i was estimated from the calculated liquid-gas transfer rate ṁ^L-Gi (Eq. 6.6) and measured liquid phase concentration C^Li and mean gas phase mole fraction x^Gmean,i via Eq. 6.7 (Chapter V).

K_La_i= ṁ_i^L−G (t)

V^L∙(C_i^L(t) − h_i(t)x_mean,i^G (t)p_mean^G (t)M_i

RT_reactor^L (t)) Eq. 6.7

where, hi denotes the Henry coefficient (g.m^-3 in the liquid phase per g.m^-3 in the gas phase) corrected for the measured reactor temperature (Table 5.4), x^Gmean,i the mean of the mole fraction in the atmosphere and in the off-gas (Eq. 6.8) and p^Gmean the mean of the hydrostatic pressure at the bottom and top of the reactor (Eq. 6.9).

x_mean,i^G (t) =x_out,i^G (t)+x_in,i^G(t) 2

Eq. 6.8 p_mean^G (t) = p_atm^G (t) + ρgH

2 Eq. 6.9

with ρ (kg.m^-3) the density of the wastewater, g (m.s^-2) the gravitational acceleration and H (m) the height of the water column.

The liquid-gas transfer coefficient for oxygen KLaO2 could be calculated directly from the measurements (Eq. 6.7) because its liquid phase concentration C^LO2 was measured on-line. However, no liquid phase measurements of CH4 or N2O were performed, so their KLai

was derived indirectly from the relationship with KLaO2 shown in Eq. 6.10 (De heyder et al., 1997), using diffusion coefficients Di (cm².s^-1) from literature (Table 5.4).

K_La_i(t)=K_La_O2(t)√ D_i

D_O2 Eq. 6.10

Besides KLaO2, two additional aeration characteristics were calculated (Henze et al., 2008), namely the oxygen transfer efficiency and average aeration efficiency. The oxygen transfer efficiency OTE (%) was calculated as the fraction of the injected oxygen mass flow that was absorbed by the liquid phase (Eq. 6.11).

OTE(t) = ṁ_O2^G-L(t)

Q_in^G(t)C_in,O2^G (t) Eq. 6.11

To compare the OTE over time and between different reactors, it was corrected for the dissolved oxygen concentration, reactor height and expressed at the standard temperature T^Lstd of 20°C and standard pressure p^Gstd of 1013 hPa (Gillot et al., 2005a, Gillot et al., 2005b), which resulted in the specific standard oxygen transfer efficiency SSOTE (Eq. 6.12).

SSOTE(t) = OTE(t)

The average aeration efficiency AE (kg O2.kWh^-1) relates the amount of oxygen absorbed to the required aeration energy and was calculated according to Eq. 6.13.

AE = m_O2^G-L

W Eq. 6.13

m^G-LO2 (kg O2) represents the total amount of oxygen absorbed during the monitoring campaign, obtained by integration of the liquid-gas transfer rate of oxygen ṁ^L-GO2 over time.

The total work required for aeration W (kWh^-1) was calculated via integration of the power (kW) needed at the rotor shaft to obtain the monitored frequency of rotation, using specifications provided by the supplier.

Liquid phase concentrations

The liquid phase concentration of a substance i was estimated from its liquid-gas transfer rate ṁ^L-Gi (Eq. 6.6) and coefficient KLai (Eq. 6.10) by rearranging the liquid-gas transfer model in Eq. 6.7 to Eq. 6.14.

C_i^L(t)= ṁ_i^L-G (t)

K_La_i(t)V^L+h_i(t)x_mean,i^G (t) p_mean^G M_i

RT_reactor^L (t) Eq. 6.14

Reaction rates

The overall consumption rate of O2 in the liquid phase Ṙ^LO2 (g.min^-1) was estimated from its absorption rate ṁ^G-LO2 (g.min^-1) via a mass balance (Eq. 6.1), in which the accumulation term was calculated through a central difference approximation, applying a Δt of 0.5 min (Eq. 6.15).

Ṙ_O2^L (t) =ṁ_O2^G-L(t) - C_O2^L (t+Δt)-C_O2^L (t-Δt)

2Δt V^L Eq. 6.15

Mass balances

Several unmeasured variables were derived from mass balances over the complete monitoring period. The oxygen balance (Eq. 6.16, in g O2) expresses that the total absorbed amount of oxygen m^G-LO2 was used for aerobic COD conversion (R^Laer,COD) and nitrification (R^Lnit,O2).

m_O2^G-L=R_aer,COD^L +R_nit,O2^L Eq. 6.16

The COD balance (Eq. 6.17, in g COD) states that the amount of COD that entered via the influent m^Lin,COD either left via the effluent (m^Lout,COD), was incorporated into the sludge (R^sludgeCOD), was aerobically converted (R^Laer,COD) or was converted during denitrification (R^Lden,COD).

m_in,COD^L =m_out,COD^L +R_COD^sludge+R_aer,COD^L +R_den,COD^L Eq. 6.17

The total nitrogen balance (Eq. 6.18, in g N) represents that all nitrogen that entered via the influent m^Lin,N either left via the effluent (m^Lout,N), was incorporated into the sludge (R^sludgeN) or denitrified (R^Lden,N).

m_in,N^L =m_out,N^L +R_N^sludge+R_den,N^L Eq. 6.18

Finally, the nitrate balance (Eq. 6.19, in g N) expresses that all nitrate produced via nitrification R^Lnit,N either left via the effluent m^Lout,N or was denitrified R^Lden,N.

R_nit,N^L =m_out,NO3^L +R_den,N^L Eq. 6.19

Assuming no nitrification-denitrification via nitrite, substitutions could be made in the four mass balances based on the stoichiometry of the nitrification (Eq. 6.20) and denitrification reaction (Eq. 6.21) and a typical nitrogen content of sludge of 0.07 g N.g COD^-1 (Eq. 6.22) (Henze et al., 2000).

R_nit,O2^L =4.57 R_nit,N^L Eq. 6.20

R_den,COD^L =2.86 R^L_den,N Eq. 6.21

R_N^sludge=0.07 R^sludge_COD Eq. 6.22 The resulting set of four equations was solved (Appendix of Chapter VI) for four unmeasured variables: the sludge production R^sludgeCOD, the amount of aerobically converted COD R^LCOD, nitrified nitrogen R^Lnit,N and denitrified nitrogen R^Lnit,N.

The net ammonium and nitrate removal were also calculated, based on the on-line ammonium and nitrite/nitrate measurements (Figure 6.3). The mass of ammonium removed during every cycle j, R^LNH4,j (g N), was estimated from the decrease of the monitored ammonium concentration during the reaction phase from C^LNH4,start reaction j to C^LNH4,end reaction j

(Eq. 6.23).

R_NH4,j^L =[CNH4,start reaction jL -CNH4,end reaction jL ]V^L Eq. 6.23 The nitrate removal during cycle j, R^LNO3,j (g N), was estimated separately for the reaction phase and the residual denitrification during the other phases (van Dijk et al., 2018). During the reaction phase, the nitrogen removal R^LNO3,reaction j was calculated as the decrease of the total nitrogen mass, i.e. the sum of the ammonium and nitrate (Eq. 6.24).

RNO3,reaction jL =[(C_NH4^L +C_NO3^L )start reaction j-(C_NH4^L +C_NO3^L )end reaction j]V^L Eq. 6.24 During the remaining phases, the removal R^LNO3,rem j was calculated as the decrease of the nitrate concentration corrected for the output of nitrate via the effluent (Eq. 6.25)

R_{NO3,rem j}^L =[CNO3,end reaction j-1L -CNO3,start reaction jL ]V_in^L- ∑ C_NO3^L (t)Q_in^L(t)∆t

start reaction j

end reaction j-1

Eq. 6.25 The respective values of R^LNH4,j and R^LNO3,rem j for every cycle j during the whole monitoring period were then added for comparison with the nitrified and denitrified nitrogen calculated via the mass balances.

Apart from the mass balance calculations over the whole monitoring period, a COD balance was also set up over every cycle j to estimate the amount of catabolised organics (aerobic and anoxic). This short-term balance states that the catabolised organics R^LCOD,j

equal the absorbed oxygen m^G-LO2,j minus the theoretical oxygen consumption for the removal of ammonium R^LNH4,j, plus the COD catabolised via denitrification (Eq. 6.26).

R_COD,j^L =m_O2,j^G-L-4.57 R_NH4,j^L +2.86 R_NO3,j^L Eq. 6.26

Figure 6.3. Schematic representation of the cyclic ammonium and nitrate nitrogen concentration profiles measured at the top of the reactor, indicating the values that were used to calculate the ammonium and nitrate removal. The phase durations are for a typical dry weather cycle.

Results and discussion

4.1. Dynamics within cycles Liquid-gas transfer rates

In this section, the on-line measured liquid-gas transfer rates at the wastewater treatment plant in Dinxperlo (Eq. 6.6; Figure 6.4A) are presented for two cycles and related to

In this section, the on-line measured liquid-gas transfer rates at the wastewater treatment plant in Dinxperlo (Eq. 6.6; Figure 6.4A) are presented for two cycles and related to

Dans le document Wastewater treatment with aerobic granular sludge : challenges and opportunities for modelling and off-gas analyses (Page 126-149)