international journal of greenhouse gas...

Fc

Ra

b

a

ARRA

KOFCNM

1

lAtasbOcuWc

rr

h1

International Journal of Greenhouse Gas Control 42 (2015) 485–493

Contents lists available at ScienceDirect

International Journal of Greenhouse Gas Control

j ourna l h o mepage: www.elsev ier .com/ locate / i jggc

ield measurements of NOx and mercury from oxy-fuel compressionondensates at the Callide Oxyfuel Project

ohan Stangera,∗, Timothy Tinga, Lawrence Beloa, Chris Sperob, Terry Walla

Chemical Engineering Department, University of Newcastle, AustraliaCallide Oxyfuel Project, Australia

r t i c l e i n f o

rticle history:eceived 3 July 2015eceived in revised form 26 August 2015ccepted 26 August 2015

eywords:xyfuellue gasompressionOx

ercury

a b s t r a c t

The Callide Oxyfuel Project (COP) is the world’s largest operating oxy-fuel plant and has successfullydemonstrated the retrofitting capability for oxy-fuel to capture CO2. The compression of the raw flue gasat the COP has also demonstrated that NOx and Hg can also be removed as part of the liquid conden-sate. This work presents field tests conducted on the condensates to determine the stability of capturedNOx and Hg species when depressurised. These tests involved sampling liquid condensate directly intoa customised aeration vessel and measuring the evolved gases over an 8–12 h period. The low-pressurecondensate (∼4 bar) showed that 3–18% of captured NOx species and 0.5–1.2% of the Hg were volatile,while the high-pressure condensate (24 bar) re-emitted 2–68% of captured NOx and 0.05–12.5% of thecaptured Hg. These tests showed that volatile Hg was related to volatile NOx and that this volatility ofcondensates changed with time as the compression plant operated from start-up. Equilibrium calcula-
tions of HNO2 in the gas and liquid phases supported the volatility measurements, suggesting that therate of oxidation of HNO2 to HNO3 in the condensed phase is slow. Overall, the conditions which favouredNOx stability in the condensates, namely longer residence and higher pressure also favoured Hg stability.This work has shown that emissions from an oxy-fuel compression plant must include those emanatingfrom depressurised condensates and suggest that the re-emitted species may not the same as in typicalcombustion flue gas, but the result of higher-pressure conversion.
. Introduction

Climate change and population growth have invariably beeninked with increasing levels of CO2 in the atmosphere (IPCC, 2014).s the world’s energy demand increases, there is an urgent need

o decouple anthropogenic CO2 emissions from electricity gener-tion. Carbon capture and storage (CCS) from fossil fuelled powertations is seen as part of a suite of technologies that can responsi-ly mitigate CO2 emissions while maintaining economic growth.xy-fuel combustion is one CCS technology that is approachingommercial-scale demonstration. Its progress has been well doc-mented in terms of combustion (Wall et al., 2009, 2011, 2013a;all and Stanger, 2011; Wall, 2007; Buhre et al., 2005) but the

ompression of oxy-fuel flue gas still has a number of uncertainties.Oxyfuel combustion of coal takes place in a mixture of O2 and

ecycled flue gas. The product is a flue gas rich in CO2, with N2emoved in the oxygen separation. This obviates the need to have a

∗ Corresponding author.E-mail address: [email protected] (R. Stanger).

ttp://dx.doi.org/10.1016/j.ijggc.2015.08.021750-5836/© 2015 Elsevier Ltd. All rights reserved.

© 2015 Elsevier Ltd. All rights reserved.

large scale capture device at the back-end of the boiler and allowsthe flue gas to be directly compressed. The higher partial pressuresof CO2 enable liquefaction processes to be applied, condensing theCO2 as a liquid to remove the gaseous inert impurities (O2, N2,Ar). The liquefaction of oxy-fuel derived CO2 has been proven atpilot scale at Vattenfall’s Schwarze Pumpe plant (Burchhardt andGriebe, 2013; Anheden et al., 2011; Strömberg et al., 2009) and atthe Callide Oxyfuel Project (COP) (Spero, 2014; Komaki et al., 2014;Spero et al., 2013a, 2013b; Court et al., 2011). The removal of otherimpurities such as SOx, NOx and Hg are less certain particularlybecause (a) the current array of environmental control technologiesare expensive to apply and (b) the compression process providesan opportunity to improve the footprint and capture efficiencies ofcontrol units.

A deeper understanding of the potential/actual removal of NOx

and Hg during flue gas compression has been obtained in previ-
ous laboratory testing (Ting et al., 2013, 2014; Stanger et al., 2014)and on slipstreams of real flue gas (Wall et al., 2013b). The higherpressure and longer residence times act to enhance oxidation ofinsoluble NO to soluble NO2, allowing for capture in compressioncondensates. Furthermore, the NO2 has previously been shown to

486 R. Stanger et al. / International Journal of Greenhouse Gas Control 42 (2015) 485–493

F sampl( d afteg ints as( our in

rftluvwtppae

tsNocpslp

2

atgmdt

ig. 1. COP compression circuit and sample points for the second test campaign. Gas2) after 2nd stage compression and intercooling; (3) after final stage compression anas outlet; (6) recycled flue gas from De-NOx liquefaction column. Liquid sample po9) high-pressure scrubber liquid outlet. (For interpretation of the references to col

eact with insoluble elemental mercury in the flue gas to remove itrom the gas stream. The removal of elemental mercury has impor-ant process implications because it can react with the downstreamiquefaction materials (brazed aluminium heat exchangers). These of other industrial control options for mercury, such as acti-ated carbon beds in the natural gas industry, are typically usedithout the presence of O2 or other oxidising species (e.g. NO2) and

heir operation in this environment is also uncertain. Other highressure removal concepts such as Air Product’s Sour Gas Com-ression (White et al., 2010, 2013; Torrente-Murciano et al., 2009)nd Linde’s LINCONOX (Winkler et al., 2011) processes have yet tostablish the impacts on mercury removal.

This field study at the Callide Oxyfuel Project has been under-aken to characterise the compression condensates in terms oftability after pressure release – in particular the re-emission ofOx and Hg after the liquid condensates are depressurised. Previ-us laboratory work has shown that dissolved gaseous impuritiesan be brought out of aqueous solution when discharged from high-ressure conditions (Stanger et al., 2014). This work has applied theame testing concepts to the compression system of the world’sargest oxy-fuel project to determine the stability of condensatesroduced at industrial scale.

. Experimental

The University of Newcastle has performed two test campaignst the COP. The first test campaign was based on the use of a bench
op compression unit with slipstream gases derived from COP withas analyses of products. The second campaign was based on directeasurements of NOx and Hg emissions associated with liquid con-
ensate removal from the compression circuit. This paper detailshe second test campaign results from the CPU liquids.

e points as numbered were (1) CPU feed after low pressure scrubber and dust filters;r cooling; (4) high-pressure scrubber feed after chilling; (5) high-pressure scrubber

numbered were: (7) low-pressure condensate; (8) high-pressure condensate; and the text, the reader is referred to the web version of this article.)

2.1. Sampling programme

The second test campaign on compression condensates was per-formed in June/July 2014. Table 1 shows the actual programme ofwork performed across the two weeks of testing. The sampling pro-gramme was focussed on testing the liquid condensates derivedfrom compression of flue gases for NOx and Hg emissions whendepressurised. From previous work in the laboratory (Stanger et al.,2014), it was expected that a portion of captured NOx gases remainsvolatile (as a dissolved gas) in the liquid and may be re-emitted intothe gas phase as an emission when the liquid is released to atmo-sphere. These volatility tests require several hours to desorb theNOx gases. While these liquid tests were being performed, a num-ber of gas samples were taken from different areas of the CPU todescribe the gaseous environment in which the condensates wereexposed. Fig. 1 shows the sample points on the COP CPU, with redpoints noted as gas samples and blue point noted as liquid samples.

2.2. Test equipment

The analytical equipment for this field campaign consisted oftwo Thermoscientific 42Hi NO-NO2 chemiluminscent analysers, anOhio Lumex 915+ elemental Hg analyser with in-house thermalconvertor (for determining Hgtotal/Hg2+), a Vaisala NDIR GMP343CO2 analyser (20,000 ppm range) and a Testo 350XL flue gas anal-yser. The two NOx analysers were used as dedicated units; oneunit for analysing gaseous streams and the other for measuring
emissions from condensates. The Hg analyser and CO2 analyserwere both dedicated to condensate emissions. As stated in previ-ous work, the chemiluminscent NOx analysers were used becausethey represent the industry standard for NOx measurement. How-ever, the NO2 values are reported with the knowledge that they

R. Stanger et al. / International Journal of Greenhouse Gas Control 42 (2015) 485–493 487

Table 1Summary of CPU Liquids Sampling Campaign, with bracketed numbers referring to locations in Fig. 1.

Date Samples taken Analysis NOTES

27/06/14 Gas phase• (1) CPU Feed (after LP Scrubber)• (2) Interstage (after cooling,∼4 bar)• (3) Compressor Final Outlet (after cooling, ∼24 bar)

Gas analysisNO, NO2

Sampled at ∼5 to 10 L/min

Condensates• (7) Intercooler, 4 bar• (8) Aftercooler, 24 bar

Volatility testsNO, NO2, Hg, CO2

Sampled ∼500 mL liquid in ∼5 L/mincompressed air for ∼8 h

28/06/14 Gas phase• (1) CPU Feed (after LP Scrubber)• (2) Interstage (after cooling, ∼4 bar)• (3) Compressor Final Outlet (after cooling, ∼24 bar)• (6) DeNOx Recycle to interstage• (4) HP Scrubber Inlet (after chiller)• (5) HP Scrubber Outlet (before Mol Sieve)

Gas analysisNO, NO2


Condensates• (7) Intercooler, 4 bar• (8) Aftercooler, 24 bar


Sampled ∼500 mL liquid in ∼5 L/mincompressed air for ∼8 h

29/06/14 Gas phase• (3) Compressor Final Outlet (after cooling, ∼24 bar)

Gas analysisNO, NO2


Condensates• (8) Aftercooler, 24 bar• (9) HP Scrubber, 24 bar


Sampled aftercooler at ∼750 mL liquid in∼5 L/min compressed air for ∼8 h

asopFfno

2

ocdmfiaadlttt3wccops3acawr

st

air was directed through a dry gas metre to the aeration vessel,with the combined air and impurities swept through the samplelines to the mobile test facility and the gas analysers. Over 30 mof ½ in. PFA tubing was used to convey these NOx/Hg emissionsinto the analysers giving approximately 5 min residence time for

re a ‘differential’ measurement between NO and total NOx. Asuch it is accepted that the NO2 measurement represents NO2 plusther species converted (in particular HNO2 and HNO3 in the gashase). Other work has been undertaken in the laboratory usingTIR to confirm the presence of higher order NOx species, howeverrom both an industrial and practical perspective, these tests couldot have been performed quantitatively with an FTIR due to theverriding (and variable) presence of H2O and CO2.

.3. Volatility testing of condensates

The testing of NOx and Hg gases released on depressurisationf liquid condensates is an uncertain area of oxy-fuel compressionircuits. The University of Newcastle first identified this as an issueuring bench scale compression testing in pursuit of closing theass balance for NOx and Hg. This field campaign appears to be the

rst attempt at determining the extent of volatile components from large-scale oxy-fuel flue gas compression circuit. A preliminaryttempt at sampling these liquids (at either 4 or 24 bar) was madeuring a prior test campaign in the May/June 2014, in which the

iquids were sampled under pressure, then transported to a mobileest facility for depressurisation. Fig. 2 shows the apparatus, withhe top sample cylinder (304SS, 300 mL) used to sample and conveyhe liquids to the analysers and the bottom cylinder (also 304SS,00 mL) used as the aerated depressurisation vessel. This systemas modelled off the work performed on the bench scale laboratory

ompression system. Preliminary results are given in Fig. 3 for aompressed air flow rate of 2 L/min and a sampled liquid volumef 160–190 mL with two condensate samples taken from the low-ressure circuit. The first sample was degassed immediately afterampling into the flowing stream of air and sampled using a Testo50 Flue Gas Analyser. Peak readings of 7.4 ppm and 20 ppm for NOnd NO2 (respectively) were measured, followed by a slow decay inoncentration over the course of 2 h. A second sample was degassedpproximately 6.5 h after sampling. In this case the peak readings
ere 3.2 ppm and 5.9 ppm of NO and NO2, indicating a significant
eduction in volatility.These early results indicated that liquid residence time between

ampling and depressurisation was a factor and the liquid aera-or was re-designed for the second week of testing in June/July

Sampled HP Scrubber at ∼500 mL liquid in∼5 L/min compressed air for ∼8 h

to allow for greater volume of sampled liquids and higher flow ofcompressed air to carry the volatile gases to the analysers. The newaeration cell was located next to the sample point and compressed

Fig. 2. First volatility test apparatus using liquid sampled at pressure then laterdepressurised. The top cylinder contains the pressurised liquid sample and bottomcylinder is the aeration vessel.


Fic

ttuftvrbbTftm

TttTuoaeapis

Fw

ig. 3. Preliminary results from first volatility tests performed on 31/05/14. Showsmpact of residence time on volatility of NOx gases released from the depressurisedondensate.

he gas to reach the mobile laboratory. A pulse counter recordedhe flow rate of compressed air through the dry gas metre contin-ously, though in practice the flow rate did not vary significantlyrom 5 L/min. The new aeration vessel is shown in Fig. 4, with aotal volume of 1.25 L, connected to the HP Scrubber. This aerationessel is constructed for use in water filtration and was pressureated for 8 bar. The polycarbonate construction material was usedecause its transparency allowed the sampled condensate flow toe controlled and provided some protection against overpressure.hese process and safety issues were considered over the potentialor adsorption of measured gas species. Given the long residenceimes of this experiment, the effect of adsorption was considered

inimal.The liquids were dropped out using flexible high-pressure

eflon-lined braided tubing with two needle valves for flow con-rol. The main condensate exhaust line was switched off for 1 mino build up sufficient liquids and ensure a seal from process gases.hen the liquids were sampled slowly over ∼5 min until the liq-id level reached ∼500 mL. Fig. 5 shows the liquid during samplingf the high-pressure condensate, with clearly visible bubbles andfter degassing for ∼1 h with minor bubbles visible. The degassing
xperiments were typically started at 9:30 am with the condenserllowed to degas throughout the 12 h shift, with a second testerformed prior towards the end of shift to allow an all-night exper-
ment to be achieved. In some cases, the aeration cell was carefullyhaken to allow greater removal of gases from the liquid surface.

ig. 5. Photograph of high-pressure condensate during sampling (left) initially with signifiith minor bubbles (right).

Fig. 4. New aeration vessel used in June/July test campaign for volatility analysis.

Care was taken to ensure no liquid entrainment in the gas linesgoing to the analysers.

During liquid volatility analysis, the second NOx analyser wasused to sample the gas points throughout the CPU circuit. These aredetailed in Table 1 and Fig. 1. Over 30 m of ¼ in. PFA lines were usedto direct gas flow from a needle valve installed at each sample point
to the test van. Multiple lines were used to allow faster switchingbetween sample points. Each PFA line had a pressure rating of 28 barto ensure safe sampling from pressurised points (i.e. up to 24 bar).
cant bubbles formed from release of dissolved volatile gases; and after 1 h degassing


Fig. 6. Volatile NOx measured from low-pressure liquid condensate at COP CPU.Interstage cooler/condenser at 4 bar, liquid sample approximately 500 mL. Sampletaken on 28/06/14 at 4:38 pm and analysed until 8:30 am the following morning.Inset: Same sample focussed on first 30 min.

Fig. 7. Volatile NOx measured from high-pressure liquid condensate at COP CPU.AtI

3

3

nwttacfptoptatlNN

Fig. 8. Relationship between volatile NOx and volatile Hg in CPU liquid condensates.

fterstage cooler/condenser at 24 bar, liquid sample approximately 500 mL. Sampleaken on 27/06/14 at 4:50 pm and analysed until 8:30 am the following morning.nset: Same sample focussed on first 30 min.

. Results from the test campaign at COP

.1. Volatility testing of condensates

The volatility tests were performed over three days and twoights and occurred over the course of many hours. These testsere highly experimental and dependent on the conditions within

he compression circuit. However it must first be stated that theseests were highly successful in capturing transient trends in NOx

nd Hg desorption and in providing a quantitative yield of volatileomponents. Figs. 6 and 7 show the overnight desorption profileor the volatile NOx for the low-pressure condensate and high-ressure condensate, respectively. Both figures show a generalrend of a sharp initial peak in both NO and NO2, lasting in therder of minutes (see inset in both figures), followed by a secondaryeak in NO2 occurring over the course of 4–5 h. The 15 min ini-ial delay in each case is due to the time at which sampling begannd includes both sampling period (∼10 min) and the residence
ime of the gas travelling to the mobile test facility (∼5 min). Theow pressure condensate in Fig. 6 shows lower concentrations ofOx with peak measurements at approximately 60 and 30 ppm forO2 and NO, respectively, with the secondary peak in NO2 reaching
Fig. 9. Relationship between volatile NOx and nitrate concentrations in CPU liquidcondensates.

∼10 ppm. The high-pressure condensate in Fig. 7 shows peak con-centrations of 410–414 ppm for both NO and NO2, followed by asecondary peak in NO2 of 80 ppm. The NO2 in Fig. 7 also shows atertiary peak occurring after 10 h, though this behaviour is not fullyunderstood.

The volatile components in both liquid condensates wasexpected to consist of dissolved gases of NO, NO2 and poten-tially HNO2 (which only weakly ionises in solution). As such thesecomponents are expected to be removed rapidly after depres-surisation. Table 2 summarises the results for the liquid volatilitytesting of the CPU condensates. The highest levels of volatility wereobserved to coincide for both NOx and Hg whilst also correspond-ing to the lowest nitrate (NO3

−) concentrations in the liquid. Suchtrends occurred for both the low and high-pressure condensate.Figs. 8 and 9 show the observed trends in volatility with respect tovolatile Hg and stable NO3

− concentrations, respectively. VolatileHg shows a linear relationship with volatile NOx, increasing mea-sured NOx emission out of the liquid. However, volatile NOx appearsto be reduced exponentially with increasing levels of NO3

−, sug-
gesting that conditions which favour higher NOx capture also actto reduce NOx (and Hg) volatility in the liquid.


Table 2Condensate volatility results.

Date Sample Liquid volume (mL) Remaining NOx/Hg concentration after volatile release Volatile gases

NO2− (mg/L) NO3

− (mg/L) Hg (�g/L) NOx (%) Hg (%)

27-June Aftercooler 418.2 0.89 718 1.5 67.9 12.527-June Intercooler 504.5 3.89 2360 14.2 3.25 0.5628-June Intercooler 480.4 0.88 676 1.8 18.0 1.2628-June Aftercooler 500.5 3.36 15,600 38.6 4.80 0.0729-June Aftercooler 753.6 4.8 45,400 115 1.99 0.0529-June High Pressure Scrubber 524.3 5.86 20,900 133 0.68 ND

Fig. 10. Effect of time on volatility of NOx in the liquid condensates.

Fc

3

semlaIivtsFr

ig. 11. The variation in NOx concentrations for gas sampling points above liquidondensates. Note: HP scrubber concentrations are taken from previous days testing.

.2. Variation in NOx gas & liquids with testing date

As discussed above, the volatility testing at COP CPU did showome expected general trends in behaviour. However, it wasxpected that the low-pressure condensate may have containedore volatile NOx species due to the lower residence time (allowing

ess HNO2 to convert to NO3−) and also the higher amounts of NO

nd NO2 (providing the highest equilibrium conditions for HNO2).n practise this was not observed, with higher volatility measuredn the high-pressure condensate on Day 1, while measuring lower
olatility on Day 2. This temporal variation may have been due tohe CPU start-up occurring at 6:30 am on Day 1 of testing. Fig. 10hows that volatile NOx measurements at the time of sampling andig. 11 shows the measured NOx concentrations taken at the cor-esponding gas sampling points. In the case of the HP scrubber, the
Fig. 12. Measured NOx and CO2 concentrations from the boiler side ContinuousEmissions Monitoring System (CEMS), provided by COP.

NOx values were taken from the previous day. Fig. 10 shows thatthe volatility of liquids discharged from the high-pressure stagebecomes increasingly lower over the course of the testing period.Conversely, the low-pressure condensate appears to rise from Day1 to Day 2 from 3.2% to 17%. This rise coincides with changing NOx

gas concentrations (Fig. 11) but also suggests that the CPU maynot have reached steady state in regards to liquid concentrations.Fig. 12 shows the NOx and CO2 concentrations taken from the COPContinuous Emissions Monitoring System (CEMS) over the courseof the testing period. Oxyfuel firing reached an approximate steadystate at 6:30 am on 27/06/14 and UoN gas and liquid sampling com-menced at ∼2:30 pm on the same day. From the NOx measurementsit is clear that gas steady state exiting the boiler had been reached,however these conditions may not be true for liquid condensates,which may have considerable hold-up throughout the process.

In an effort to incorporate the changing NOx levels over thecourse of the testing period, the equilibrium concentration of dis-solved gases NO, NO2 and HNO2 were estimated for each liquidsample based on the measured NO and NO2 concentrations in thegas phase. An estimation of HNO2 in the gas phase was calculatedaccording to the equation given by Schwartz and White (1981) asfollows:

NO + NO2 + H2O ↔ 2HNO2

Kp = (PHNO2 )2

PNOPNO2 RH= 2.7 × 10−2

where RH is the fractional relative humidity, assumed to be 1.The dissolved gas concentration in the liquid was estimated

using Henrys constants for each gas and the system pressure (i.e.
4 bar for low pressure condensate and 24 bar for high pressure con-densate and HP scrubber) and the gas concentrations (measuredfor NO and NO2 and estimated for HNO2). Values for Henry’s con-stant for NO, NO2 and HNO2 were also taken from Schwartz andWhite (1981) and are given as 1.9 × 10−3, 1.2 × 10−2 and 49 M/atm,


Table 3Estimated dissolved gas concentrations in the depressurised liquids of the CPU using measured NO and NO2 values.

Date of sampling Gas phase Dissolved in liquid Desorbed in air

NO (ppm) NO2 (ppm) HNO2

(estimatedppm)

HNO2

(mol/L × 10−3)NO(mol/L × 10−6)

NO2

(mol/L × 10−5)Estimatevolatility(%mol N)

NO2/NOratio (–)

Interstage27/06/2014 402.7 410.0 93.6 18.0 3.06 1.97 10.9 6.428/06/2014 532.1 312.2 93.9 18.0 4.04 1.50 38.2 3.7Compressor outlet27/06/2014 43.1 194.3 21.1 25.0 1.97 5.60 48.5 28.428/06/2014 74.4 221.6 29.6 35.0 3.39 6.38 3.1 19.929/06/2014 187.2 897.1 94.4 111.0 8.54 25.8 3.4 30.2HP Scrubber28/06/2014 43.7 249.7 24.1 28.0 2.11 6.22 1.9 29.5

Fig. 13. Comparison of equilibrium and dissolved gas estimations with measuredvoa

rsbaiapfwbtAcpsrNh

(traipdeN

gently agitating the liquid during aeration to increase the interfa-

olatility in depressurised liquid condensates. Values are expressed as a percentagef the total nitrogen moles in the liquid including NO3

−/NO2− and the integrated

mount of dissolved NOx gases released as measured in the volatility tests.

espectively. These estimates are given in Table 3 and were con-idered to represent the total amount of dissolved gas that mighte expelled when depressurised. It should be noted that HNO2 islso known to be formed from the absorption of NO2 and N2O4 (ass HNO3), however for these calculations it has been assumed thatny HNO2 present in the liquid must be in equilibrium with the gashase. A comparison of estimated volatility and measured volatilityor NOx is given in Fig. 13. It can be seen that the estimates compareell with measured values given the significant variation in liquid

ehaviour. Low-pressure condensates appear lower than estima-ion while high-pressure condensates appear higher than parity.

potential reason for such discrepancy may be that low-pressureondensates have higher mass transfer limitations given the lowerartial pressures of the gas constituents. High-pressure conden-ates may appear above the estimates due to other equilibriumeactions not accounted for (i.e. the significant secondary peak inOx emissions from high pressure liquids, occurring over severalours).

The calculated partial pressures of HNO3 above the condensatesusing NO3

− concentrations) showed that the expected HNO3 inhe gas phase was not significant (0.04–0.14 ppm) and that equilib-ium decomposition of HNO3 back to NO2 and O2 could potentiallyccount for up to 0.017 ppm NO2. This suggests that the NOx chem-stry is complex, involving both gas and liquid phase reactions (and
otentially mixed phase) and that pressure plays a major role inetermining the ultimate stability of the liquid condensates. As anxample of such complexity, the ratio of expected dissolved NO2 toO in the liquids are also given in Table 3. These values suggest that
Fig. 14. Volatility test for HP Scrubber, with multiple agitations of the test cell andassociated Hg release behaviour identified as peaks. Shows Hg emission from liquidis present in oxidised form (Hg2+). Sample taken on 29/06/14.

the amount of NO2 desorbing out of the liquid should be a factorof 3–6 times that of the desorbed NO for the low-pressure con-densates and a factor of 20–30 times higher in the high-pressurecondensates. For the volatility tests shown in Figs. 11 and 12, theprimary peaks showed a NO2 to NO ratio of 2 and 1 respectivelyfor the low and high-pressure condensates. The elongated desorp-tion profiles of both Figures approach a common NO2 to NO ratioof 2 after 8 h (not including the secondary peak in NO2). The higherthan expected levels of NO desorbing out of the liquids is a poten-tial indicator of HNO2 oxidation (to HNO3 and thus NO3

−) throughthe pathways (Schwartz and White, 1981):

3HNO2 ↔ HNO3 + 2NO + H2O (a)

3NO2 ↔ N2O5 + NO (b)

NO2 + HNO2 ↔ HNO3 + NO (c)

In this experimental case, there is the potential for these reac-tions to occur either within the liquid itself during depressurisation(and continued aeration) or throughout the transport line.

3.3. Observed mercury phenomena in volatility testing

Given the prolonged nature of the volatility tests, a number ofsystem disturbances were undertaken in an attempt to shorten therequired time for gas measurements to occur. These consisted of

cial surface area and allow higher mass transfer. The HP scrubberliquid was one such test that had a larger sample size (∼750 mL)coupled with a large amount of gas release. Fig. 14 shows thevolatility results for this test. The agitation appears as a series of

492 R. Stanger et al. / International Journal of Gree

Fig. 15. Zoomed time scale of volatility test for HP Scrubber, with multiple agitationsoHTf

ptHcoerpmmopatippprocHHvttvoapae

4

4

ecp3rss

the total NO captured in the condensate. It was also shown that

f the test cell and associated Hg behaviour. Sample taken on 29/06/14. Emission ofg from aeration cell interacts with NOx emission, first positively then negatively.he Hg baseline check after 11:30 am shows all back-end Hg was present in oxidisedorm.

eaks in NOx concentrations, in some cases reaching higher thanhe initial depressurisation peak. During this particular test, theg levels were measurably high enough to detect similar peakhanges that occurred with the NOx. However, it was furtherbserved that the Hg levels continued to increase after the NOx lev-ls began to subside. This curious observation led the operators toun the Hg analyser in Hg0 mode, bypassing the furnace and therebyreventing the oxidised mercury (Hg2+) from being converted oreasured. Fig. 14 shows this bypass in step (3) illustrating that theeasured Hg levels consist entirely of oxidised Hg. Further analysis

f the coincident Hg peaks (shown in Fig. 15) revealed that theseeaks initially rise with the (first 4) NOx peaks, then become neg-tive in the subsequent (4) NOx peaks. This phenomenon suggestshat a portion of the Hg emissions from the liquid condensates isn oxidised form. Certainly the final peak in Hg in Fig. 14 is com-letely oxidised, however the same cannot be said of the first 4ositive Hg peaks in Fig. 15. A likely scenario is that the initial Hgeaks are elemental (i.e. Hg0) and that the subsequent agitationeleases HNO2 and NO2 which then reacts with Hg0 in the gas phasef the aeration cell and throughout the transport line. Equilibriumalculations for HNO2 (e.g. in Table 3) suggest that 90% of releasedNO2 would decompose to NO and NO2. The relative volatility ofg species oxidised with NOx gases has been demonstrated in pre-ious laboratory experiments (Ting et al., 2014) and it is also likelyhat more than one Hg product may be formed from such a reac-ion. These results provide clear evidence that interaction betweenolatile species derived from liquid condensate depressurisation isccurring and that the products are themselves partially volatilend in oxidised form. It is suggested here that these oxidised Hgroducts are potentially soluble and that their subsequent capturefter depressurisation is not only possible but imperative givenxpected toxicity of soluble oxidised Hg species.

. Conclusions

.1. Capture of NOx emissions during compression

This study and our previous work has shown that the NOx

missions exiting the COP boiler in oxy-firing may be successfullyaptured by up to 60% during compression to the COP maximumressure of 24 bar. Laboratory scale work to higher pressures of
0 bar increased this potential capture to 85–90% removal. Theeaction mechanism relies on the initial gas phase kinetic conver-ion of NO to NO2, which is enhanced at higher pressure, and theubsequent absorption of NO2/N2O4 in the water condensates to
nhouse Gas Control 42 (2015) 485–493

HNO2 and HNO3. However, the slower liquid phase oxidation ofHNO2 to HNO3 limits the stability of NOx on depressurisation asdoes the desorption of product NO out of the liquid. At this stage, itremains unclear which of these slower mechanisms is rate limiting.

4.2. Capture of Hg emission during compression

The removal of Hg from the flue gas exiting an oxy-fuel boilerand ash collection system occurs predominantly through fly-ashinteractions at the fabric filter. Oxy-fuel combustion has beenshown to produce higher levels of flue gas borne Hg in oxidisedform compared to air firing (Nelson et al., 2013; Spörl et al., 2014),and these species are considered readily removed in the initialquenching of flue gas to remove water (being highly soluble).The remaining gaseous Hg (present as Hg0) is relatively insolu-ble and must be removed prior to contact with cryogenic heatexchange surfaces in the cold box. From the current investigations,this Hg0 can be readily removed from the flue gas during com-pression through a gas phase, kinetically controlled, reaction withNO2/N2O4. The product of this reaction is unknown but relativelystable in the liquid condensates. At this stage, the effect of longerexposure to NO2 (such as in the case of NO2 recycle to the com-pressor of the COP CPU) is unknown but this is expected to increaseremoval.

4.3. Volatility of depressurised condensates

The release of NOx species from liquids upon depressurisationfrom the CPU has been measured at the COP and found to be widelyvariable. The variation was shown to occur as the COP CPU plantcontinued to operate and it was speculated that this change wasthe result of the liquid phase reaching a steady state within the cir-cuit. Measurements at the COP CPU indicated that the low pressure(4 bar) condensate showed a release of 3–18% captured NOx and0.5–1.2% of captured Hg. The high-pressure condensate (24 bar)released 2–68% of captured NOx and 0.05–12.5% of captured Hg.The liquid produced from the High Pressure Scrubber released thelowest amount of NOx (0.7%). Equilibrium calculations of HNO2 ingas and liquid phase were based on plant measurements of NO andNO2. These values suggested that HNO2 contributed up to ∼10%of the gas phase NOx but was the dominant dissolved gas (>99%)in the condensates due to its high Henry’s constant and weak ion-isation. These equilibrium estimates of dissolved gases correctlypredicted the trend in volatility, however the high pressure (24 bar)liquids were over-predicted and the low pressure (4 bar) conden-sates were under-predicted in absolute values. As stated above, therate of oxidation of HNO2 to HNO3 is unknown under these con-ditions and thus the required residence time to ensure stability asNO3

− remains uncertain. The COP measurements have shown thatthe greater the concentration of NO3

−, the lower the proportionof volatile NOx species. Methods for enhancing this process wouldreduce the risk of NOx re-emission.

4.4. Re-emission of Hg species from depressurised liquidcondensates

The COP measurements have shown that the volatility of Hgspecies in the compression condensates is directly related to theamount of volatile NOx. However the maximum amount of volatileHg was measured at ∼12% of the total mercury captured in the con-densate, while the maximum amount of volatile NOx was ∼65% of

x

the volatile NOx species readily reacted with volatile Hg during re-emission in the gas phase and that this product was of oxidisedform. Overall, the conditions which enhance NOx stability in the liq-uids (longer residence time, higher pressure) clearly act to enhance

f Gree

Hei

4

odgNo(tkosedtaimso

A

vatAafvu

R

A

B

B

C

I

R. Stanger et al. / International Journal o

g stability and reduce volatility. Methods for reducing NOx re-mission would be beneficial to ensuring Hg species are stabilisedn the liquid.

.5. Industrial implications

This work has demonstrated that the relatively simple conceptf NOx gases being captured as HNO3 in the aqueous condensatesuring oxyfuel flue gas compression is complicated by dissolvedases (mainly HNO2). The compression system actively convertsOx and Hg into soluble products and while this has been previ-usly shown to be highly effective in our previous work and at COPStanger et al., 2014; Nelson et al., 2013), there remain environmen-al implications for their re-emission. In particular, HNO2 is wellnown to play a role in atmospheric smog chemistry, while solublexidised Hg species are considered to removed from the atmo-phere at a regional level through wet and dry deposition (Nelsont al., 2009). As such, future oxy-fuel plants will require permittingiffering from conventional systems. This research has highlightedhe need for potential mitigation options for such emissions suchs extended residence time at pressure, off-gas recycle, chem-cal/photochemical oxidation. Establishing the conditions which

aximise their capture and stability will not only reduce the emis-ions of future oxy-fuel plants, but will offer a distinct advantagever other CCS options.

cknowledgements

The authors also wish to acknowledge financial assistance pro-ided through Australian National Low Emissions Coal Researchnd Development (ANLEC R&D). ANLEC R&D is supported by Aus-ralian Coal Association Low Emissions Technology Limited and theustralian Government through the Clean Energy Initiative. Theuthors would expressly like to thank the staff of the Callide Oxy-uel Services Pty Ltd and the CS Energy Operations team for theiraluable contribution to this work. Their tireless efforts provideds with important help prior, during and after this field campaign.

eferences

nheden, M., Burchhardt, U., Ecke, H., Faber, R., Jidinger, O., Giering, R., et al., 2011.Overview of operational experience and results from test activities inVattenfall’s 30 MWth oxyfuel pilot plant in Schwarze Pumpe. Energy Proc. 4,941–950.

uhre, B.J.P., Elliott, L.K., Sheng, C.D., Gupta, R.P., Wall, T.F., 2005. Oxy-fuelcombustion technology for coal-fired power generation. Prog. EnergyCombust. Sci. 31, 283–307.

urchhardt, U., Griebe, S., 2013. Tests and results of Vattenfalls oxyfuel pilot plant.In: 3rd IEA Oxyfuel Combustion Conference, Ponferrada, Spain.

ourt, P., Briglia, A., Tranier, J.-P., Perrin, N., 2011. Callide CO2 capture pilot plant
design. In: 2nd IEA GHG Oxyfuel Combustion Conference, Yeppoon, Australia,September 13th.
PCC, 2014. Climate change 2014: synthesis report. In: Core Writing Team,Pachauri, R.K., Meyers, L.A. (Eds.), Contribution of Working Groups I, II, and IIIto the Fifth Assessment Report of the Intergovernmental Panel on ClimateChange. PICC, Geneva, Switzerland, 151 pp.

nhouse Gas Control 42 (2015) 485–493 493

Komaki, A., Gotou, T., Uchida, T., Yamada, T., Kiga, T., Spero, C., 2014. Operationexperiences of oxyfuel power plant in Callide Oxyfuel Project. Energy Proc. 63,490–496.

Nelson, P., Nguyen, H., Morrison, A., Malfoy, H., 2009. Mercury Sources,Transportation and Fate in Australia, Final Report to the Department ofEnvironment, Water, Heritage and the Arts, RFT 100/0607.

Nelson, P., Morrison, A., Bray, P.S., Malfoy, H., Stanger, R., Spero, C., 2013.Measuring trace elements during the December 2012 Callide Oxy-fuel trial. In:3rd Oxyfuel Combustion Conference, Ponferrada, Spain.

Schwartz, S.E., White, W.H., 1981. Solubility equilibria of the nitrogen oxides andoxyacids in dilute aqueous solution. In: Pfafflin, J.R., Ziegler, E.N. (Eds.),Advances in Environmental Science and Engineering. Gordon and BreachScience Publishers.

Spero, C., (2014). Callide Oxyfuel Project – Lessons Learned, Report to the GlobalCCS Institute, Available at: http://www.globalccsinstitute.com/publications/callide-oxyfuel-project-lessons-learned (accessed 14.09.15).

Spero, C., Yamada, T., Nelson, P., Morrison, T., Bourhy-Weber, C., 2013a. CallideOxyfuel Project: combustion and environmental performance. In: 3rd OxyfuelCombustion Conference, 10–12th September, Ponferrada, Spain.

Spero, C., Yamada, T., Rees, G., Court, P., 2013b. Callide Oxyfuel Project: overview ofcommissioning experience. In: 3rd Oxyfuel Combustion Conference, 10–12thSeptember, Ponferrada, Spain.

Spörl, R., Belo, L., Shah, K., Stanger, R., Giniyatullin, R., Maier, J., et al., 2014.Mercury emissions and removal by ash in coal-fired oxy-fuel combustion.Energy Fuels 28, 123–135.

Stanger, R., Ting, T., Wall, T., 2014. High pressure conversion of NOx and Hg andtheir capture as aqueous condensates in a laboratory piston-compressorsimulating oxy-fuel CO2 compression. Int. J. Greenh. Gas Control 29,209–220.

Strömberg, L., Lindgren, G., Jacoby, J., Giering, R., Anheden, M., Burchhardt, U., et al.,2009. Update on Vattenfall’s 30 MWth oxyfuel pilot plant in Schwarze Pumpe.Energy Proc. 1, 581–589.

Ting, T., Stanger, R., Wall, T., 2013. Laboratory investigation of high pressure NOoxidation to NO2 and capture with liquid and gaseous water under oxy-fuelCO2 compression conditions. Int. J. Greenh. Gas Control 18, 15–22.

Ting, T., Stanger, R., Wall, T., 2014. Oxyfuel CO2 compression: the gas phasereaction of elemental mercury and NOx at high pressure and absorption intonitric acid. Int. J. Greenh. Gas Control 29, 125–134.

Torrente-Murciano, L., White, V., Chadwick, D., 2009. Removal of SOx and NOx fromoxyfuel-derived CO2. In: IEA 1st Oxyfuel Combustion Conference, RadissonHotel, Cottebus, Germany.

Wall, T.F., 2007. Combustion processes for carbon capture. Proc. Combust. Inst. 31,31–47.

Wall, T., Stanger, R., 2011. Industrial scale oxy-fuel technology demonstration.In: Zheng, L. (Ed.), Oxy-Fuel Combustion for Power Generation andCarbon Dioxide (CO2) Capture. Woodhead Publishing, Cambridge, UK, pp.54–76.

Wall, T., Liu, Y., Spero, C., Elliott, L., Khare, S., Rathnam, R., et al., 2009. An overviewon oxyfuel coal combustion – state of the art research and technologydevelopment. Chem. Eng. Res. Des. 87, 1003–1016.

Wall, T., Stanger, R., Santos, S., 2011. Demonstrations of coal-fired oxy-fueltechnology for carbon capture and storage and issues with commercialdeployment. Int. J. Greenh. Gas Control, 5.

Wall, T., Stanger, R., Liu, Y., 2013a. Gas cleaning challenges for coal-fired oxy-fueltechnology with carbon capture and storage. Fuel, 85–90.

Wall, T.F., Stanger, R., Ting, T., Liu, D., 2013b. Gas Quality Impacts, Assessment andControl in Oxyfuel Technology for CCS, Final Report to ANLECR&D, project6-0710-0061, Available at: http://www.anlecrd.com.au/ literature 162108/Gas quality impacts, assessment and control Final Report - UoN (accessed14.09.15).

White, V., Torrente-Murciano, L., Sturgeon, D., Chadwick, D., 2010. Purification ofoxyfuel-derived CO2. Int. J. Greenh. Gas Control 4, 137–142.

White, V., Wright, A., Tappe, S., Yan, J., 2013. The air products-VattenfallOxyfuel CO2 compression and purification pilot plant at Schwarze Pumpe. In:3rd Oxyfuel Combustion Conference, 9–13 September, Ponferrada, Spain.

Winkler, F., Schoedel, N., Zander Jr., H., Ritter, R., 2011. Cold DeNOx developmentfor oxyfuel power plants. Int. J. Greenh. Gas Control 5 (Suppl. 1),S231–S237.

international journal of greenhouse gas...

Documents