international journal of greenhouse gas...

Oms

Ra

b

a

ARA

KOMNCFC

1

ctpCtwrco

U

h1

International Journal of Greenhouse Gas Control 41 (2015) 50–59

Contents lists available at ScienceDirect

International Journal of Greenhouse Gas Control

journa l homepage: www.e lsev ier .com/ locate / i jggc

xyfuel derived CO2 compression experiments with NOx, SOx andercury removal—Experiments involving compression of

lip-streams from the Callide Oxyfuel Project (COP)

ohan Stanger a,∗, Tim Ting a, Chris Spero b, Terry Wall a

Chemical Engineering, the University of Newcastle, NSW, AustraliaCallide Oxyfuel Services Pty Ltd., QLD, Australia

r t i c l e i n f o

rticle history:eceived 23 April 2015ccepted 23 June 2015

eywords:xyfuelercuryOxompressionlue gasO2 capture

a b s t r a c t

Oxyfuel combustion is a CO2 capture technology which is approaching commercial demonstration. Ofpractical interest is the use of the compression circuit to allow low-cost cleaning options for various fluegas impurities. This work has focussed on three species – NOx SOx and Hg – and their removal duringcompression of “real” oxyfuel flue gas sampled as a slip stream from the demonstration Callide OxyfuelProject. The flue gas slip stream was compressed using a bench-scale piston compressor developed toallow measurements of impurity concentrations after each compression stage using adjustable pressures.Several operating configurations were investigated including variable pressures from 5 to 30 bar, inter-stage temperature changes and flow rate. Slip streams taken before and after SOx removal allowed theimpact of mixed NOx/SOx gases to also be investigated. The results from the “real” oxyfuel flue gas experi-ments for the three species were similar to those performed in the laboratory using synthetic flue gas andreported previously. The capture of SO2 was found at be greater at low pressures than NOx capture, with90% removal of SO2 by a pressure of 10 bar, with NOx capture extending to higher pressures. The effectof residence time during compression had the greatest influence at higher pressures (>10 bar) where the
kinetic rate of NO oxidation to NO2 increases less with pressure increase. Capture of NOx was increasedfrom 55% to 75% by doubling the residence time in the compressor and could be further extended to 83%by increasing back end pressure from 24 bar to 30 bar. Lowering the temperature during compressionproduced the greatest NOx and Hg capture. Overall, the results indicate that capture of mercury duringcompression occurred as a consequence of high pressure, longer residence time and concentration of NO2.
. Introduction

The increase in anthropogenic CO2 emissions has been linked tolimate change and one of the largest sources of CO2 is the genera-ion of electricity (IPCC, 2007). Currently coal fired power stationsroduce 42% of the world’s electricity (World Coal Associationoal Statistics, 2012) and coal combustion accounts for 43% ofhe world’s CO2 emissions (IEA Report, 2012). However, with theorld’s population predicted to increase to 8.2 billion by 2030, the

ising demand for electricity is expected to increase the need foroal fired power generation (IEA, 2008). Oxy-fuel combustion isne CO2 capture technology which reduces the emission from coal

∗ Corresponding author at: Department of Chemical Engineering, Building EB,niversity of Newcastle Callaghan, 2308, Australia.

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

ttp://dx.doi.org/10.1016/j.ijggc.2015.06.022750-5836/© 2015 Elsevier Ltd. All rights reserved.

© 2015 Elsevier Ltd. All rights reserved.

fired power utilities and allows safe storage of the CO2 in deepunderground aquifers.

Oxy-fuel technology is based on the combustion of coal in a mix-ture of O2 and recycled flue gas, producing a flue gas rich in CO2which is then compressed, liquefied and injected underground. Thetechnology has been significantly reviewed as it moves towardsdemonstration (Buhre et al., 2005; Toftegaard et al., 2010; Wallet al., 2009; Wall, 2007; Wall et al., 2011). With the recycle of fluegas, oxyfuel systems are known to produce higher impurity levelsand these gas cleaning challenges can be met with a number ofplant configurations (Wall et al., 2013a). Traditional gas cleaningunits (such as FGD’s for SOx removal) have been shown to oper-ate to similar performance capabilities as in air fired conditions
(Thybault et al., 2009; Wall et al., 2013b). However, a particularbenefit of oxyfuel combustion (and of considerable interest) is thepassive removal of impurities during CO2 compression.

R. Stanger et al. / International Journal of Greenhouse Gas Control 41 (2015) 50–59 51

Table 1Industrial mercury limits.

Mercury limit Concentration �g/Nm3 Note

Expected Hg in flue gas from coal 10–50 Without captureOccupational safe working limit 10 Based on TWA (8 h time weighted average)Australian gas quality standard 1 AEMO natural gas rulesa

Brazed aluminium HEX standard 0.1 ALPEMA recommendationb

Typical design specification for carbon bed 0.01 For natural gas processingc,d

Analytical limit 0.0001–0.002 Instrument dependent

ints, 2007 (Australian Energy Market Operator (AEMO), 2007).dition, 2010 (ALPEMA, 2010).itigation” (Francis, 2012).stion” (Santos, 2010).

abc(HdflfpumphmTa2pga(iwoa

cebheohssucoc

2

2g

aago

Table 2Summary of compressor details and modifications.

Initial system Modifications

Description RIX Industries Model SA-3 oilfree, three stage, axial pistonair compressor, designed forSCUBA tank filling

Functional 3HP electric motor 2HP max after fitting withvariable speed drive

Maximum pressure 248 bar Pressure gauges + controlto 40 bar, relief valve to50 bar

3CFM flow rate at a speed of2300RPM

2CFM

Compressor Piston Size:1st stage ∼60 mm diameter2nd stage ∼25 mm diameter3rd stage ∼12 mm diameterStroke ∼24 mmOil freeCooling: forced air fan

Air cooling coils removed,gas sample point added to1st stageCondensers and interstagepressure controls + addedgas sampling point to 2nd& 3rd stageWater cooled

Gas analysis Thermoscientific 42i-HLchemiluminscent NOx analyser(0–1000 ppm range)Ohiolumex RA-915 + Hg0

Analyser(0–20,000 ng/m3 range)M&H dilution system (1:10critical orifice + venturisampling)Testo flue gas analyser 350XL(for FEED)(O2, CO, NO, NO2, SO2)Bosch wideband O2 sensorAccutherm dry gas meter(0–60 L/min, pulse outputevery 10 L)

Furnace System added forthermal conversion ofgaseous Hg2+ at 700 ◦CConnected to interstagesample points with ON/OFFand needle valves for flowregulation

Liquid analysis Horiba pH probe (calibratedrange pH 1–7)Horiba nitrate probe(calibrated to 2000 mg/L NO3

−)Dionex ion chromatograph(calibrated to 500 mg/L NO3

−)Hg content in liquid

Dilution requiredUse of KMnO4/H2O2 fortotal N & nitrite, dilutionrequiredDilution requiredUsed SnCl2/HCL injectioninto liquid combined with

a Australian Energy Market Operator; Gas Quality Standard – System Injection Pob Aluminium Plate-Fin Heat Exchanger Manufacturer Association Standards 3rd Ec Francis, “Brazed Aluminium Heat Echangers – BAHX – Surveillance, Analysis, Md Santos, “Challenges in Understanding the fate of mercury during oxyfuel combu

Air Products have demonstrated the potential removal of SOxnd NOx during compression using the well-known lead cham-er process (White et al., 2008, 2013). Other systems, such LINDE’sompression plant at Vattenfall’s Oxyfuel Pilot at Schwarze PumpeThybault et al., 2009; Yan et al., 2009) and the University ofamburg’s Oxyfuel pilot plant (Koepke et al., 2009) have alreadyemonstrated that acidic condensates as low as pH of 0.8 can be

ormed by compressing oxyfuel flue gas containing NOx. However,ittle work has been performed on mercury removal in an oxy-uel CPU. Mercury (as gaseous Hg0), must be substantially removedrior to the CO2 liquefaction, as brazed aluminium heat exchangerssed in the process are well known to be corroded in the presence ofercury (Santos, 2010). From the natural gas industry, the current

ermissible limit on Hg concentration entering a brazed aluminiumeat exchanger is 0.01 �g/m3, corresponding to the current limit ofeasurement. A summary of typical Hg concentrations is given on

able 1. Given that Hg concentrations exiting an oxy-fuel boilerre between 2.7 and 45 �g/m3 (Nelson et al., 2013; Spörl et al.,014), this suggests that a significant level of cleaning is requiredrior to CO2 liquefaction. It has been previously theorised that flueas bourn mercury (present as Hg0) will be absorbed in the nitriccid formed from compressing the raw CO2 from an oxyfuel systemAllam et al., 2007). However, a recent study on mercury behaviourn pressurised systems found that the dominant removal pathway

as by reaction with NO2 in the gas phase (Ting et al., 2014). With-ut NO2 as a gaseous oxidiser, Hg0 was both absorbed and desorbeds a dissolved gas.

This ties the removal of mercury to the absorption of NOx inompression condensate, and (if successful) could replace the morexpensive traditional options such as activated carbon beds. A num-er of fundamental studies on the compression of oxyfuel flue gasave been reported by the University of Newcastle (UoN) (Tingt al., 2014; Stanger et al., 2014; Ting et al., 2013) with a focusn NOx and Hg behaviour at higher pressure. The latest studyas involved the development and laboratory testing of a benchcale piston compressor to study behaviour under real compres-ion conditions (Stanger et al., 2014). This fundamental work wasndertaken in support of the Callide Oxyfuel Project (COP). Thisurrent study details the use the bench scale compressor with realxyfuel flue gas sampled from slip streams during a two week fieldampaign at the COP.

. Experimental

.1. Laboratory piston compression apparatus (continuousas/continuous liquid)

The apparatus used for evaluating Hg behaviour in oxyfuel was laboratory scale compression system built from a three stageir compressor (RIX Industries Model SA-3) and modified to allowas sampling between compression stages and interstage variationf pressure. This apparatus is described elsewhere (Stanger et al.,

“Volatile” NOx + Hg in liquidtested using gas analysersystem (see Fig. 2)

M&H Dilution System+Ohio Lumex RA–915 + Hg0

gas analyser above

2014). Fig. 1 shows a diagram of the apparatus using the same gasand Hg feeding system with the addition of a H2O bubbling satura-tor prior to entering the compressor. Table 2 provides a summary ofthe compressor details and its modifications. The compressor wasfitted with a variable speed drive to change the gas in-take (i.e. ineffect changing the residence of the gas). After stages 2 and 3 a back
pressure regulator, pressure relief valve and pressure gauge werefitted to allow the pressure to be altered within the compressioncircuit. Sampling of the gas occurred using a dilution system whichpulled 9 L/min of compressed air through a critical orifice and ven-

5 of Gre

taa(ta

umoatcdpucmrtiwaairt

2

csaao2oym

F

2 R. Stanger et al. / International Journal

uri providing a 0.5 bar vacuum for sampling 1 L/min of gas through second orifice. This combination diluted the sample gas by 1:10nd allowed the NOx analyser to handle high NO2 concentrationstypically limited to ∼100 ppm NO2). A dry gas meter measuredhe gas flowing out of the compressor after passing through anctivated carbon scrubber.

Liquid samples were taken after compression stages 2 and 3sing the moisture separators provided with the compressor. Theodifications involved removing the copper cooling coils in favour

f flexible high pressure teflon tubing braided with stainless steelnd a 300 mL 304SS sample cylinder. The liquid sampling modifica-ions are shown in Fig. 2. These additions allowed for indirect waterooling of the compressed gas leaving the piston stage and for con-ensation of water. This step was considered important from theerspective of simulating mixing conditions between gas and liq-id prevalent in compression (i.e. high surface area, wetted walls,ondensing droplets). The gas/liquid mix was then directed to theoisture separators, where the liquid was allowed to build-up until

eaching a suitable sample volume. The gas exited from the top ofhe separator and passed through the back pressure regulator andnto the next stage (i.e. either 3rd stage piston or exit). The liquid

as sampled in a highly controlled method using a fine needle valvend could be sampled directly into a container or directed towardsn air-aspirated sample cylinder (300 mL, 304SS). For the field test-ng at Callide only total liquid samples were taken due to timeestraints. The volatile analysis of the liquid condensates typicallyakes several hours and uses both the NOx and Hg analysers.

.2. Comment on NOx gas analysis

The analytical method used for NOx gas analysis was based onhemiluminescent detection. These units are the commonly usedystems in commercial application. This method detects a total NOxnd uses a convertor to provide the difference between total NOxnd NO2. In this case, the NO2 difference signal may be comprised
f other N species, of which N2O4 is expected to be measured asxNO2 (due to the equilibrium between NO2/N2O4) and amountsf N2O3, N2O, HNO3, HNO2 (all of which may be detected to varyinget unquantifiable degrees). It should also be noted that the gaseasurements are made at atmospheric pressure after exiting the
Gas Sample Point

MFC

MFC

MFC

O2

1% NO/N2

N2/CO2

bypass

Hg0 permeation tube

+

water bath

H2O

Saturator

Variable Speed

Motor

vent

ig. 1. Laboratory compression system with gas feeding system (feeding O2/N2/NO, Hg0,

enhouse Gas Control 41 (2015) 50–59

pressure vessels. It is expected that this may also have an influenceon species such as N2O4, N2O3 and HNO2. As such the term NO2 isused here with the understanding that other species may be presentand future work should be focussed on clarifying the presence andquantification of these gases.

2.3. Callide Oxy-fuel Project field trial – compression of real fluegas

The field trial of the laboratory compressor was undertaken overa two week period in October 2013 with the following aims:

1) Determine the effect of SO2 on NOx conversion during compres-sion.

2) Determine the influence of residence time under pressure.3) Determine the influence of temperature on NOx conversion.4) Compare laboratory compression results to actual gas measure-

ments on the Callide CPU.

The effect of SO2 during compression was expected to be associ-ated with reactions with the NO2 via the lead chamber mechanism,rapidly producing H2SO4 and NO (i.e. NO2 acting as an oxidiser)with the overall impact being to reduce Hg conversion in compres-sion. The concentration of SO2 was altered by selectively samplingfrom before and after the low pressure caustic scrubbers. Fig. 3shows the Callide CPU diagram with slip stream sample point indi-cated. The high SOx slip stream was taken at the inlet of the quenchcolumn, where flue gas enters the column directly from the boilerblock at ∼150 ◦C, containing high moisture (∼22% H2O) and highparticle concentration (∼150 mg/Nm3). The slip stream was sam-pled by flexible ∼12 mm PFA tubing, which has a high thermalresistance and low Hg absorption. Sampling of the high SOx streamoccurred without heated lines and thus was cooled and partiallycondensed out into a large impinger and dust filter prior to entryinto the compressor. The residence time of the gas passing through
the laboratory compression system was controlled by changing thespeed of the drive motor which in turn altered the compressorintake. Overall, the low flow rate tests were performed at ∼10 L/min(STP) and the high flow rates were ∼20 L/min (STP). The effect oftemperature was examined by cooling down the two condensing
Pressure

Control

Pressure

Control

Liquid Sampling

exhaust

dry gas

meter

AC

Scrubber

Gas Sampling

Dilution Air

To gas

analysers

H2O), three stage axial piston compression and interstage gas and liquid sampling.


F on, gaa

utpioo

icfrvgsp

Ft

ig. 2. Detailed diagram of interstage compression line showing water condensatind air aspiration.

nits placed after the 2nd and 3rd stage pistons. These condensersypically sat in a bucket of water allowed to stay at ambient tem-erature (∼34 ◦C), however, a full testing day was performed with

ced water, with the level of ice maintained in the bucket through-ut the testing period allowing a constant measured temperaturef ∼0 ◦C.

The laboratory compression system and associated analyticalnstruments were located down on the CPU bench between theaustic scrubbers and the Air Separation Unit. A photograph takenrom a nearby observation point shows the UoN set-up locationelative to the Callide Oxyfuel Plant is shown in Fig. 4. A high topan was used to house the analytical equipment (e.g. computers,as analysers) and a gazebo was erected to house the compressor
ystem and associated sampling equipment. A summary of sam-ling activities is given in Table 3.
ig. 3. Diagram of Callide CPU with gas slipstream sample points (red) for compression

hat the 2 compression symbols represent 2 stages of compression each. Condensates are

s-liquid separation and liquid sampling system using controlled de-pressurisation

2.4. Note on actual temperature and pressure (ATP)

Under these testing conditions the flow rates were measuredfrom the compressor exhaust using a dry gas meter. This gas couldbe considered to have water content under saturated conditionsat 30 bar pressure and ambient temperature (34 ◦C) of ∼1860 ppm,which have been assumed to be negligible. Rather than correct fora lower temperature, these flow rates were set at ambient condi-tions and reflect the true volumetric flow rate (and hence, residencetime) at the time of sampling. As such the term ATP here refers to34 ◦C and 101.3 kPa for this study.

apparatus and other samples including liquids (blue) taken during field trial. Note removed after compressor stages 2 and 4.

54 R. Stanger et al. / International Journal of Greenhouse Gas Control 41 (2015) 50–59

Fig. 4. Photograph of the UoN se-up with the parked white van with analyticalequipment relative to the CPU and ASU of the Callide A Power Station, showing theoxy-fuel retrofit.

Table 3Description of sampling points and test operations.

Sampling Pointand approximategas conditions(Spero et al., 2013)

Aim Samples taken oncompressor

Scrubber 1 inlet(CPU feed)Stack conditions∼150 ◦C19–22% H2O150–250 mg/Nm3

dust

High SOxLow flow 10 L/min(Preliminary)

Full gas suite0/24/8/3 bar

High SOxLow flow 10 L/min

Full gas suiteFeed/3rd/2nd/1st stage0/24/8/3 bar2nd stage suite(5–10bar)

Scrubber 1 outlet(compressor feed)Clean-coolconditions35 ◦C5–7% H2O<0.02 mg/Nm3 dust

Low SOxHigh Flow 20 L/min(Preliminary run)

Full gas suiteFeed/3rd/2nd/1st stage0/30/8/3 bar

Low SOxHigh flow 20 L/min

Full gas suiteFeed/3rd/2nd/1st stage0/24/8/3 bar2nd stage suite5–18 bar

Low SOxLow flow 10 L/min


3

3

oifoigflt

of the data, giving a lower NO concentration throughout the com-

Low SOxLow flow 10 L/minICED condensers


. Results

.1. Laboratory compression unit field testing

The compression system was run for a total of 5 days, typicallybtaining a full suite of analysis over ∼6 h. A typical result is shown

n Fig. 5, whereby the compression stages are sampled in reverserom 3rd stage to feed in order to avoid the downstream impactsf sampling (i.e. a reduced gas flow lengthening the residence time

n the compressor). Also of note is the asymmetric cycles in theas concentrations. These fluctuations are considered typical of realue gas and ultimately result in longer required sampling periodso obtain sufficient variation. In practice, at least three cycles were

Fig. 5. Typical gas results for a full suite of compressor operation with samplingfrom high to low pressure.

considered a minimum before switching the sample. Fig. 5 showsthat these cyclic variation begin with NO and Hg, in particular theHg represents variation of ∼30%. As the pressure rises from 1stthrough to 2nd stage the cycles can be observed to pass into theNO2 as greater amounts of NO was converted, however, by the 3rdstage the residence time inside the final stage acts as a blendingvessel, smoothing out the cycles. Fig. 5 also shows a break in datacollection, where the concentrations drop to zero. These periodsrepresent a manual baseline check of the Hg analyser using zeroair passed through an activated carbon bed, which was followedby a switch between the Hg sample flowing through the furnace(set at 700 ◦C) to obtain a Hgtotal or bypassed to obtain the Hg0. Forevery compressor test performed there was no significant differ-ence between Hgtotal and Hg0 observed, suggesting that no oxidisedmercury species are present in the sampled gas streams.

Full suite analysis was performed for conditions of high and lowflow in the compressor with the low SOx slip stream, as well as a lowflow run from the high SOx slip stream. These runs used comparablepressures in the three stages of the compressor; namely 1–3 bar inthe 1st stage, 8–10 bar in the 2nd stage and 24 bar in the 3rd stage.The 3rd stage pressure of 24 bar was set in order to match the pres-sure of the Callide CPU compression block. Two other full suite runswere also performed on the low SOx slip stream; the first was ahigh flow condition using 30 bar in the 3rd stage and the other wasa low flow condition using the ice in the condenser. A second set ofexperiments were also performed on the 2nd stage only, typicallyvarying the pressure from 5 to 18 bar. These fundamental runs wereobtained at the high flow condition on the low SOx slip stream andat the low flow condition on the high SOx slip stream. The resultsfor these experiments, both full suite analysis and fundamental 2ndstage pressure tests are summarised in Figs. 12–14 in terms of NO,NO2 and Hg. The NO concentration (Fig. 6) decreased exponentiallywith higher pressure, with the rate of conversion becoming sloweras concentration is reduced. The majority of the reaction occurs atlower pressure with approximately 68% of the NO converted withinthe 10 bar pressure range. The apparent scatter in the data wascaused by the cyclic variation in the plant and compared to thisscatter, the overall impact of the flow rate and SO2 concentrationdo not appear to be significant. However, the low SOx condition didresult in a lower NO concentration in the 2nd stage. The iced con-densers do appear to show an impact greater than the variability

pression range, despite having a higher feed NO. This is expectedgiven that the conversion of NO to NO2 favours lower temperature.

The formation of NO2 in Fig. 7 shows the opposite in the lowpressure region, increasing in concentration up to 10 bar, however,

of Gre

ttsd(ttsleatsa

iitict1iswsatHilist

The field trial results were compared against the previous lab-oratory results. The previous laboratory results were based on

R. Stanger et al. / International Journal

he concentrations appear to reach a plateau at higher pressures ashe rate of capture becomes increasingly significant. It is in the 3rdtage compression that the NO2 concentrations show a significantifference between experimental conditions, with the high flowand thus, lower residence time) having the highest NO2 concen-ration and therefore the lowest capture rate. The extended 30 barest in the high flow condition suggests that greater capture is pos-ible at higher pressure. The low flow conditions with both high andow SO2 concentration do not appear to provide significant differ-nce in NO2 concentration, however, both low flow conditions give

lower NO2 than the high flow condition. It should also be notedhat the low SOx condition also resulted in a higher NO2 in the 2ndtage. The iced condensers gave the lowest final NO2 concentration,pproximately half the high flow condition.

Of the three gas species, Hg showed the most variation both ints cycles and between testing days. During the course of the test-ng program, the Callide plant operators and engineers revealedhat several return loops from the back of the CPU can be mixednto the incoming feed gas. Largest among these was the use oflean CO2 product gas to pulse the primary dust filter located afterhe caustic scrubbers. These pulses were set at approximately every5 min, which is similar to most of the cyclic fluctuations measured

n the NOx concentrations, however the actual pulse rate could beet by operators and this value varied depending on which operatoras asked. Fig. 8 gives the effect of higher pressure on Hg in the gas

tream. As with the NO/NO2, these numbers are represented as anverage for each sample period and therefore appear scattered dueo the effect of the cycle trend. As stated earlier, no difference ingtotal and Hg0 was observed, proving that all of the Hg is present

n the gas stream as elemental Hg. This was consistent with the
aboratory results. The effect of pressure is pronounced on the Hgn the gas stream, being systematically reduced with higher pres-ure. There does not appear to be any clear effect of high flow, buthe iced condensers show lower Hg concentration exiting the 2nd
Fig. 6. Removal of NO with pressure.

Fig. 7. Formation of NO2 with pressure.

enhouse Gas Control 41 (2015) 50–59 55

stage. This value is the same exiting the 3rd stage, indicating thatno capture of Hg occurs in the 3rd stage under the iced condensercondition. It must be stated that not all conditions are representedacross the full pressure range, with the low flow/low SOx miss-ing the 3rd stage Hg measurement and the high SOx/low flow notincluded. The incomplete data set for the low flow low SOx was dueto a power failure, which resulted in a loss of data on the Hg anal-yser. The high SOx low flow data set was not included because of thenegative influence of the SO2 on the Hg baseline – a consequenceof the high response analyser used.

As a summary of the removal potential during compression,Fig. 9 shows the total NOx and SOx exiting each pressure settingof the compressor. Within the variability of the data, there appearsno clear trend in NOx between testing conditions in the 2nd stage,however, the 3rd stage high pressure region reflects the same trendof the NO2, with the higher flow producing less capture, followedby the low flow condition and finally the iced condenser produc-ing the greatest NOx capture. The SO2 shows a much larger effectfrom increased pressure, with 80% of the SO2 captured at 10 barin the 2nd stage and 96% captured by the 3rd stage. Overall, theimpact of SO2 on NOx capture does not appear to be significantunder these conditions. Furthermore, it was expected that no NOxcapture would occur until the SO2 was completely removed, butthis was not observed, with both SO2 and NOx being capturedsimultaneously.

3.2. Comparison of laboratory and field results

a full suite analysis using controlled NOx injection between 500and 1500 ppm NO and controlled Hg injection of 1000 ng/m3.Figs. 10–12 show these comparisons in terms of the amount of NO

Fig. 8. Removal of Hg with pressure.

Fig. 9. Impact of compression pressure on NOx/SOx removal.

56 R. Stanger et al. / International Journal of Greenhouse Gas Control 41 (2015) 50–59

Fl

cccssflrpclTr1

smf1tt

nid(iHw

Fla

ig. 10. Comparison of NO converted during compression between field tests andaboratory results at three different NO feed rates.

onverted, the amount total NOx captured and the amount of Hgaptured. The NO converted is based on the difference between NOoncentrations from the feed NO to each pressure setting. Fig. 10hows that the two high flow cases in the field trials lie on theame conversion trend, approximately 10–15% lower than the lowow case. The 2nd stage low flow condition in the field results cor-esponds well with the laboratory work performed at the sameressure and flow rate (10 L/min), while the 3rd stage low flowondition in the field results corresponds well with the trend inaboratory data despite being at different final pressure settings.he previous 3rd stage laboratory results were performed at 30 bar,ather than 24 bar, however, these results are also very close to00% NO conversion.

By comparison, the amount of total NOx removed from theystem through compression gives a measure of the capture perfor-ance under different conditions. Fig. 11 shows the NOx captured

or the two data sets. Once again, all results compare well up to0 bar, however, in the higher pressure range the high flow condi-ions show a lower overall capture rate due to the shorter residenceime.

The capture of Hg during compression involves a significantumber of variables, in particular the concentration of NO2, which

n turn is affected by residence time, temperature and water con-
ensation. A comparison of Hg capture across the two data setsFig. 12) shows a relative increase with higher pressure lead-ng to a significant amount (95–100%) of Hg removed at 30 bar.owever, the data sets diverge during the 2nd stage compressionith much higher Hg capture demonstrated in the laboratory test-
ig. 11. Comparison of NOx captured during compression between field tests, andaboratory results at three different NO feed rates. Laboratory tests were performedt the low flow rate of 10 L/min. The high flow tests were performed at 20 L/min.

Fig. 12. Comparison of Hg captured during compression between field tests andlaboratory results. Laboratory tests were performed at the low flow rate of 10 L/min.The high flow tests were performed at 20 L/min.

ing at high NOx levels (1000–1500 ppm) and much lower resultsfor the low NOx laboratory tests at 500 ppm. A potential reasonfor this difference is that these laboratory tests were performedin Newcastle during winter when laboratory temperatures weretypically below 20 ◦C leading to lower inlet water vapour con-centrations and a lower condenser temperature. By comparisonthe field testing occurred at Callide during spring with ambientair and condenser temperatures of ∼34 ◦C. The effect of lowertemperature in the field tests can be clearly observed in the icedcondenser test which showed a significantly higher capture rate inthe 2nd stage compression. In the 3rd stage at 24 bar, the iced con-denser result shows higher capture than the high flow conditionat the same pressure settings gaining approximately 10% highercapture.

4. Discussion

The field trial at the Callide Oxyfuel Project examined threemain variables during compression, namely: the effect of SO2, res-idence time and temperature. The presence of SO2 was found toimpart a minor reduction on the conversion of NO into NO2 inthe 2nd stage of the compressor but did not affect the amountof NOx captured. No effect of SO2 was observed in the 3rd stagemost likely because the SO2 was removed in the 2nd stage. How-ever, SO2 was found to be readily captured during compression,suggesting that it remains a viable and passive cleaning option.Residence time was shown to have the greatest impact in the 3rdcompression stage, with a doubling of flowrate lowering the con-version of NO to NO2 by 10% and reducing NOx capture by 20%.Extending the pressure range from 24 bar to 30 bar had a minorimpact (+5%) on NOx captured but higher impact (+12%) on Hgcaptured.

Recent work presented on the Schwarze Pumpe (Yan et al., 2013)oxyfuel pilot-plant compared different trials in NOx and SOx cap-ture including traditional FGD systems and CPU options. Linde, theoriginal suppliers of the Schwarze Pumpe CPU, have undertakenpilot work on their LICONOX process which uses a high pressurescrubber with ammonia water run at approximately pH 7. Passingthrough the compressor units, the NOx removal was ∼50%, withan additional 25–30% removed in the trial scrubber. The potentialbenefit of this process was an ability to regenerate the solution toproduce N2. An added benefit was the stabilisation of the solution
chemistry enabling better handling of by-products. Air Productshave also conducted a pilot trial of their sour gas compressioncircuit. Under this process, both SOx and NOx are included in com-pression, with SO2 being preferentially removed in the first column(15 bar) and the NOx being removed in the second column (30 bar).


Fos

TtHotolmiobhuaf

aaiHaulS

4

caoafaNp(bdtcibt

ig. 13. Modelled NO conversion and equilibrium N2O4 at 10 bar for temperaturesf 35 ◦C and 0 ◦C. Greyed dotted line gives approximate residence time in the 2ndtage of the laboratory compressor in the experiments.

he SO2 oxidation and removal was previously considered due tohe catalytic oxidation from NO2 (forming NO) (Allam et al., 2007).owever, recent modelling work has suggested that this processccurs in the liquid phase (Normann et al., 2013), either throughhe lead chamber process (resulting in NO desorbed from the liquid)r the rashig mechanism (resulting in the formation of N2O). This

iquid phase concept is supported by the field compressor experi-ents which showed only an influence on NO conversion and no

nfluence on NO2 or the amount captured from the gas phase. More-ver, in their pilot trials Air Products showed that SOx capture coulde affected by the feed SOx/NOx ratio with lower ratios favouringigher capture rates. The SO2 concentrations during the Air Prod-cts testing ranged as high as 2000 ppm with NOx between 180nd 400 ppm. In the Air Products trials, these SOx/NOx ratio rangedrom 3 to 11 producing SO2 capture rates from 100% down to 20%.

By comparison, the SOx/NOx ratio in the Callide field tests waspproximately 1, suggesting that sufficient NOx would be avail-ble for accelerated SOx capture. The complex chemistry involvedn this process is not fully understood, nor is the effect of pressure.owever, in both the lead chamber process and the rashig mech-nism, a significant amount of H2SO4 is required and it remainsp to future work to determine whether the use of water or alka-

ine solutions can effectively buffer the formation of N2O in mixedOx/NOx conditions under pressure.

.1. Temperature effects

The effect of lowering temperature from 35 ◦C to 0 ◦C using ICEDondensers produced a significant increase in the amount of NOxnd Hg captured. As previously predicted (Wall et al., 2013c), thexidation of NO to NO2 is partially driven by an equilibrium withn “activated” NO species which favours lower temperature. Theollowing dimerization of NO2 to N2O4 also favours lower temper-ture ultimately producing a greater conversion of the feed NO into2O4. Fig. 13 gives the modelled NO conversion to NO2 and theredicted equilibrium N2O4 for both temperatures in the 2nd stage10 bar). The model shows that the difference in NO conversionetween temperatures is relatively small compared to the pre-icted difference in N2O4 concentration. Fig. 14 shows the influence
hat temperature and NOx concentration has on the gas phase Hg0
oncentration. In both figures, the grey dotted line indicates the res-dence time inside the second stage of the laboratory compressor. Inoth temperature cases, the Hg0 is consumed well before the end ofhe compression stage. In comparison, the lower temperature also

Fig. 14. Modelled Hg0 conversion at 10 bar for temperatures of 35 ◦C and 0 ◦C, basedon the reaction with N2O4. Greyed dotted line gives approximate residence time inthe 2nd stage of the laboratory compressor.

changes the amount of water condensed in each compressor stage– increasing the amount condensed in the second stage from 84% to98% of the feed H2O and decreasing the amount in the third stagefrom 9% down to 1% of the feed H2O. Despite this reduction, theamount of NOx captured in the 3rd stage was increased, suggestingthat the amount of liquid water formed is not a driving influencein capturing NOx in the condensate.

The combination of higher NO conversion and higher NO2/N2O4concentrations had the greatest impact on the Hg captured in the2nd stage. While the literature is scarce on the mechanism of thereaction between Hg and NOx, the modelled effects of temperatureon the NOx kinetics and equilibrium suggest that N2O4 is in fact thelikely reactive species as previously suggested (Hall et al., 1995).In the previous report (Wall et al., 2013c) the laboratory resultsindicated that the presence of liquid water was not required for Hgcapture, but the capture of NOx from the gas phase into the liquidretarded the Hg capture. The iced condenser condition in the fieldtests changed the amount of liquid water present and did not resultin a significant increase in extra NOx captured in the 2nd stage. Itdid, however, have a large influence on Hg captured, supporting theprevious laboratory observations.

4.2. Practical implications

Slipstream measurements using actual flue gas taken from thebench-scale compressor agree with synthetic flue gas measure-ments taken from the bench-scale compressor in the laboratory.The field measurements taken from a slipstream from a real oxy-fuel flue gas showed similar conversion with higher pressure andin terms of actual concentration. The natural cyclic variability ofthe real flue gas is a feature which could not be replicated in thelaboratory.

The impact of SO2 on NOx conversion in compression was minor,producing a measurable difference in NO conversion but no differ-ence in NOx captured. The expectation that SO2 would react withNO2 and thus reduce its capture is derived from the work of AirProducts, however the ratio of SOx to NOx is apparently sufficient
to provide the accelerated SOx capture effect without significantinfluence. This suggests that passive capture of both SOx and NOxduring compression is viable providing the operational envelopefor N2O formation is determined (N2O was not measured in thecurrent campaign).

5 of Gre

cH

efi

4

tpauocucts

Nwnofsbocbda

rhuacpwHatpIcsc

Tibdossawsgast

8 R. Stanger et al. / International Journal

The amount of Hg that was removed from both the experimentalompressor sampling a slip-stream and the Callide CPU satisfy theg limit for natural gas compression limit of 10 ng/m3.

The practical implications suggested in previous work (Wallt al., 2013c) on laboratory measurements of synthetic gas are con-rmed by the present results on real oxy-fuel gas at the COP.

.3. Implications confirmed from previous laboratory work

Gaseous Hg0 in the flue gas can be effectively removed fromhe gas stream using the compression circuit during back-end CO2rocessing. This reduces the risk of mercury corrosion of cryogenicluminium heat exchangers and suggests that dedicated removalnits, such as an activated carbon bed are not necessary in anxy-fuel circuit. These findings represent a paradigm shift in mer-ury control for oxy-fuel technology and reduces the emphasis onpstream speciation and capture. Furthermore, this passive gasleaning option in oxy-fuel is a low cost advantage over other cap-ure technologies such a post-combustion capture which requireignificant cleaning prior to CO2 removal.

The main pathway for oxidation occurs in the gas phase withO2/N2O4, rather than in the acidic condensate. The conditionshich enhance NO oxidation are the same as that for Hg oxidation,

amely higher pressure and residence time. However, the presencef liquid water in an oxy-fuel compression circuit is both necessaryor removing NOx and a hindrance for removing Hg. This work hashown that Hg0 can be dissolved in acidic liquid at high pressure,ut does not form a stable salt without the presence of a gaseousxidiser. Therefore the premature capture of NOx into the liquidondensates will retard the removal rate of Hg0. Future design ofack-end oxy-fuel CO2 processing plants could enhance the Hg oxi-ation rate by prolonging the residence time prior to intercoolingnd condensate removal.

The final oxidised Hg product has yet to be identified. Thisesearch has shown that Hg0 is removed from the gas phase butas yet to identify the product of oxidation. Likely candidate prod-cts are a Hg-NOx salt or HgO, however, the valence of the Hg atomffects solubility, and may well impact the potential removal in theondensate. The results of this work have shown that a significantortion of the gaseous Hg was not removed in the condensate andas not present in the gas as Hg2+. It is therefore possible that theg is being deposited in a section of the laboratory compressionpparatus which is not being regularly contacted with water or thathe compound is not soluble. On a larger scale, this may ultimatelyresent a risk of blockage or a safety hazard during shut-down.

dentification of the final oxidised form of Hg under compressiononditions will offer insights into its behaviour during compres-ion and ultimately result in better circuit designs to optimise itsapture.

Highly acidic condensates represent a corrosion and safety risk.he analysis of the condensates indicated a significantly high acid-

ty which consisted predominantly of HNO3. Given that the massalance closure was relatively poor, there is high possibility thateposition of HNO3 is occurring in non-wetted sections. Given thatbserved blockages were found in areas surrounding the back pres-ure regulators and piston head it is also possible that other NOxpecies such as N2O5 (a white solid at room temperature) maylso be formed. In this work, the blockage was easily removedith water and future design options may take advantage of such

olubility to remove potential build-ups. The emission of NOxases produced from depressurising the liquid condensates present

potential safety hazard. Liquid condensates from compressionhould be depressurised in a controlled way to allow off-gassingo be recycled back into the process.

enhouse Gas Control 41 (2015) 50–59

5. Conclusions

The removal of NOx, SOx and Hg during compression of “real”oxyfuel flue gas was investigated using a novel (piston) laboratorycompression unit. The unit was previously used under laboratoryconditions using a synthetic flue gas and was used on a sampledslipstream of “real” oxy-fuel flue gas derived from the Callide Oxy-fuel Project. The field campaign at Callide had four main aims:

1) To determine the impact of passively cleaning SO2 and NOx inthe compression circuit.

2) To vary the residence time within the compression circuit.3) To determine the impact of temperature on NOx conversion.4) To compare the (experimental) compressor results to measured

NOx and Hg concentrations within the Callide CO2 ProcessingUnit.

It was found that SO2 had only minor impact on NO conversionand no impact on the capture of NOx during compression. It was alsoobserved that the SO2 was removed at a faster rate than the NOx,obtaining >80% removal by 10 bar. This was found to be consistentwith other international pilot trials.

The effect of shorter residence time has confirmed the impactof gas kinetics and was shown to decrease the conversion of NO toNO2 though this effect was mainly observed in the 3rd compressionstage after 60% conversion of NO. The overall capture of NOx duringthe compression circuit could be improved from 55% to 75% witha doubling of residence time and further extended to 83% with anincrease in pressure from 24 to 30 bar.

A reduction in temperature during high pressure intercooling(using an ICED condenser) had the greatest effect on NOx capture athigher pressure, while Hg capture was removed at lower pressure.Modelling suggested that the lower temperature produced a higherequilibrium concentration of N2O4 which is considered the likelyreactant species involved in Hg removal.

Overall, the field measurements taken from the laboratory com-pressor using real flue gas were found to agree with the previouslaboratory measurements. This provides confidence in laboratoryexperiments to study NOx/SOx/Hg mechanisms prevalent in largescale oxyfuel compression. The removal of these flue gas impu-rities during compression has the potential to reduce the overallcost of oxyfuel and represents a significant advantage that oxy-fuelhas over other CCS technologies, such as post-combustion capture,which require removal of impurities prior to entering the CO2 com-pressor unit.

Acknowledgements

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

References

Allam, R.J., Miller, E.J., White, V., 2007. Purification of Carbon Dioxide. Air Productsand Chemicals, Inc., EP.

ALPEMA, 2010. The standards of the Brazed Aluminium Plate-fin Heat ExchangerManufacturers Association -3rd. edition.

Australian Energy Market Operator (AEMO), 2007. “Gas Quality Standard-SystemInjection Points” 2007.

of Gre

B

F

III

K

N

N

S

S

S

S

T

T

Yan, J., et al., 2013. Characteristics of SOx and NOx in CO2 Processing and for

R. Stanger et al. / International Journal

uhre, B.J.P., et al., 2005. Oxy-fuel combustion technology for coal-fired powergeneration. Prog. Energy Combust. Sci. 31 (4), 283–307.

rancis, A., 2012. Brazed Aluminium Heat Exchangers – (BAHX) Surveillance –Analysis – Mitigation. In: Benyahia, A.A. (Ed.), In: Proceedings of the 3rd GasProcessing Symposium. Elsevier: Oxford, pp. 170–182.

EA, 2008. World Energy Outlook.EA Report, 2012. CO2 emissions from fuel combustion – Highlights.PCC, 2007. IPCC Fourth Assessment Report: Climate Change 2007 Synthesis Report.

[September 27, 2007]. Available from: <http://www.ipcc.ch/publications anddata/publications ipcc fourth assessment report synthesis report.htm/>.

oepke, Daniel, et al., 2009. Liquefaction of oxyfuel flue gas – experimental resultsand comparison with phase equilibrium calculations. In: IEA 1st OxyfuelCombustion Conference, Radisson Hotel, Cottbus, Germany.

elson, P., et al., 2013. Measuring trace elements during the December 2012Callide Oxy-fuel trial. In: 3rd Oxyfuel Combustion Conference, Ponferrada,SPAIN.

ormann, F., et al., 2013. Nitrogen and sulphur chemistry in pressurised flue gassystems: a comparison of modelling and experiments. Int. J. Greenh. GasControl 12 (0), 26–34.

antos, 2010. Challenges in Understanding the Fate of Mercury during OxyfuelCombustion, [26/11/2010]. Available from: <http://www.ieaghg.org/docs/General Docs/IEAGHG Presentations/SSantos - Hg Presentation MEC7 - June18.pdf/>.

pörl, R., et al., 2014. Mercury emissions and removal by ash in coal-fired oxy-fuelcombustion. Energy Fuels 28 (1), 123–135.

pero, C., et al., 2013. Callide Oxyfuel Project: Combustion and EnvironmentalPerformance. In: 3rd Oxyfuel Combustion Conference, in 10–12th September,Ponferrada, SPAIN.

tanger, 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 (0),209–220.

hybault, Cyril, et al., 2009. Behaviour of NOx and SOx in CO2

compression/purification processes – experience at 30MWth oxy-coalcombustion CO2 capture pilot plant. In: IEA 1st Oxyfuel CombustionConference Redisson Hotel, Cottbus, Germany.

ing, 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 (0), 15–22.

enhouse Gas Control 41 (2015) 50–59 59

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 (October), 125–134.

Toftegaard, M.B., et al., 2010. Oxy-fuel combustion of solid fuels. Prog. EnergyCombust. Sci. 36 (5), 581–625.

Wall, T., et al., 2009. An overview on oxyfuel coal combustion – state of the artresearch and technology development. Chem. Eng. Res. Des. 87 (8), 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 (S1).

Wall, T., Stanger, R., Liu, Y., 2013a. Gas cleaning challenges for coal-fired oxy-fueltechnology with carbon capture and storage. Fuel 108 (June), 85–90.

Wall, T., Stanger, R., McDonald, D., 2013b. Coal-fired oxy-fuel technology forcarbon capture and storage. In: Baukal Jr, Charles E. (Ed.), In Oxygen EnhancedCombustion. , 2nd edition. CRC Press.

Wall, T., Stanger, R., Ting, T., 2013c. ANLECR&D Report: Oxyfuel derived CO2

compression experiments with NOx and mercury removal – construction of acompressor and laboratory experiments prior to experiments involvingslip-streams from the Callide Oxyfuel Project (COP).

Wall, T.F., 2007. Combustion processes for carbon capture. Proc. Combust. Inst. 31(1), 31–47.

White, V., et al., 2008. Purification of Oxyfuel-Derived CO2. In: In 9th InternationalGreenhouse Gas Control Technologies. Elsevier, Washington DC.

White, V., et al., 2013. The Air Products-Vattenfall Oxyfuel CO2Compression andPurification Pilot Plant at Schwarze Pumpe. In: 3rd Oxyfuel CombustionConference 9–13 September, Ponferrada, SPAIN.

World Coal Association Coal Statistics, 2012. Available from: <http://www.worldcoal.org/resources/coal-statistics//>.

Yan, J., et al., 2009. Flue gas cleaning processes for CO2 capture from oxyfuelcombustion – experience of FGD and FGC at 30MWth oxyfuel combustion pilotplant. In: IEA 1st Oxyfuel Combustion Conference Radisson Hotel, Cottbus,Germany.

Oxyfuel Combustion CO2 Capture. In: in 3rd Oxyfuel Combustion Conference9-13 September, Ponferrada, SPAIN.

Hall, B., Schager, P., Ljungstrom, E., 1995. An experimental study on the rate ofreaction between mercury vapour and gaseous nitrogen dioxide. Water Air SoilPollut. 81 (1–2), 121–134.

international journal of greenhouse gas...

Documents