Available online at www.sciencedirect.

com
Proceedings
of the
Combustion
Institute
Proceedings of the Combustion Institute 33 (2011) 3383–3390
www.elsevier.com/locate/proci

Experimental analysis of the combustion behaviour

of oxyfuel flames in a gas turbine model combustor
Peter Kutne *, Bhavin K. Kapadia, Wolfgang Meier, Manfred Aigner
Deutsches Zentrum für Luft- und Raumfahrt (DLR), Institut für Verbrennungstechnik,
Pfaffenwaldring 38, D-70569 Stuttgart, Germany

Available online 1 September 2010

Abstract

Power generation by using oxyfuel combustion in a gas turbine cycle is a promising option to reduce
carbon dioxide ðCO2 Þ emission, while using fossil fuels. In order to use this process some significant
changes to the gas turbine are required, regarding which open questions still exist. An important question
is whether reliable operation with oxyfuel combustion under gas turbine conditions is possible. The paper
describes experiments on partially premixed swirl stabilized oxyfuel flames carried out in a gas turbine
model combustor at atmospheric pressure. To characterize the behaviour of the oxyfuel flames a systematic
parameter study for oxidisers consisting of 20–40% oxygen ðO2 Þ, equivalence ratios from 0.5 to 1, and pow-
ers of 10–30 kW was carried out. OH -chemiluminescence imaging was used to visualize the flame structure
and stability. The results show a strong influence of the O2 concentration on the combustion behaviour in
contrast to the equivalence ratio which has only a very small effect. To obtain quantitative results, laser
Raman scattering has been used on selected flames to measure simultaneously the major species concen-
trations, mixture fraction and temperature. The results reveal differences in the flame stabilisation mecha-
nism, compared to methane ðCH4 Þ/air flames in the same burner.
Ó 2010 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Oxyfuel combustion; Gas turbine model combustor; Laser Raman scattering

1. Introduction remain the largest source for energy conversion

in the next decades. Thus there is an increased
The reduction of CO2 emissions has become a interest in power generation industry towards tech-
major topic of research in the last decade due to nologies, which help to reduce CO2 emissions from
the negative environmental impact of CO2 and fossil fuels. These technologies can encompass
other green house gases on world climate. In either upgradation of existing power plants or use
2007 the European Commission has declared to of new systems which can provide a significant
reduce the emission of greenhouse gases by 20% advantage. This interest of industry is documented
until 2020 and increase the use of renewable by a large number of EU projects with industrial
energy. Though there are big efforts in the area of involvement, the likes of ENCAP, DECARBit,
renewable energy sources, the fossil fuels will BIGCCS, CAESAR, CACHET, DYNAMIS,
CAPRICE, COACH, ECCO, NANOGLOWA,
CLC Gas Power. The importance of CO2 capture
*
Corresponding author. Fax: +49 711 6862578. technologies for the reduction of CO2 emissions
E-mail address: [email protected] (P. Kutne). is also approved by a study of McKinsey (‘Carbon

1540-7489/$ - see front matter Ó 2010 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.proci.2010.07.008
3384 P. Kutne et al. / Proceedings of the Combustion Institute 33 (2011) 3383–3390

Capture and Storage: Assessing the economics’) Reynolds numbers were varied from 8362 to
[1], presented at the EC in Bruessel in 2008. 16300 for five flames of which two were buoyancy
The oxyfuel combustion of coal in a steam tur- controlled. The flame lengths were calculated from
bine process is regarded as a possible way to use OH -chemiluminescence and indicated flames to
the oxyfuel process for CO2 reduction. Research be 2–3 times shorter than air flames with adiabatic
on this field is very active with the outcome that flame temperatures in the region of 3050 K. Dita-
the first demonstration plants are in operation, ranto and Hals [8] discussed the effect of stoichiom-
and the power generation industry is willing to etric operation and high O2 content in oxidiser on
invest in this technology. Another way of particu- thermo-acoustic oscillations in sudden expansion
lar interest is the use of oxyfuel combustion in gas jet configuration. They reported occurrence of
turbines. This process offers the possibility to use thermo-acoustic instabilities as O2 content in the
the same post combustion techniques as for the oxidiser was increased, characterizing different
oxyfuel coal process, in combination with an effi- instability modes dependant on flow velocity and
cient combined cycle process. The Graz cycle flame speed variations. Open questions still exist
introduced in 1995 by Jericha et al. [2] is such a regarding flame characteristics when oxyfuel mix-
cycle that continues to be improved [3–5]. Despite tures substitute conventional air operation in a
the general interest in the process, the oxyfuel com- gas turbine burner. Aspects pertaining to the kind
bustion for gas turbines has never been taken of flame suitable, i.e., diffusion/premixed/partially
beyond thermodynamic calculations. Initially the premixed, the optimum CO2 dilution in oxidiser
large effort due to the air separation unit (ASU) for stable operation, consequent exhaust gas tem-
is seen as one reason, as for this form of combus- peratures, and occurrence of instabilities in the
tion part of the turbine energy is used by the combustor need to be addressed. The work pre-
ASU to remove the N2 present in air. In the mean- sented here seeks to answer some of these questions
while there has been a considerable development by investigating atmospheric pressure flames on a
of ASUs, in particular as parts of a power plant. technical swirl combustor. The combustor is
There are also several oxyfuel processes promising fuelled with CH4 and a mixture of CO2 and O2 as
high efficiency that offset the energy loss due to the oxidiser. Two sets of experiments were conducted
air separation making this process competitive to with the first focusing on determining flame stabil-
other CO2 avoiding techniques. The crux of the ity for equivalence ratios 0.5 to 1, and for different
matter seems to be the changes in the turbo O2 concentrations in oxidiser. OH -chemilumines-
machinery that must accompany changes in the cence imaging is used to visualize the size, shape
burner. Since this process replaces nitrogen ðN2 Þ and position of heat release zone. The second set
by CO2 and hence uses a mixture of steam and of experiments involve laser Raman scattering
CO2 as a working fluid, the turbo machinery has measurements to characterize the flames in detail
to be adapted to the different physical properties. by obtaining the main species concentration, mix-
At high flame temperatures CO2 has a heat capac- ture fraction and temperature, which would also
ity more than 1.6 times higher as that of N2 . serve as a database for model validation. For the
Because CO2 is an infrared active molecule the purpose of comparison with earlier measurements
radiant heat transfer has to be taken into account. of CH4 =air flames, lean operation conditions were
Beside the physical differences the high amount of chosen for the Raman measurements, although
CO2 has also an effect on the chemical reactions, stoichiometric conditions would be more relevant
because it can act as a reaction partner and its high for technical application of oxyfuel combustion.
chaperon efficiency enhances third body reactions. Investigation of technical operation condition,
In addition the cycle calculations indicate, that a with pressurized flames, preheated oxidiser and
gas turbine for oxyfuel combustion has to work stoichiometric conditions will be the focus of mea-
at higher pressures than existing stationary gas tur- surements planned in 2010.
bines. Assuming that it is feasible to build such a
turbine, another question arises: is the oxyfuel
combustion under gas turbine conditions viable? 2. Experimental setup
Williams et al. [6] investigated syngas and meth-
ane flames for premixed swirl stabilised conditions 2.1. Burner Geometry
for two different oxidisers of air and O2 =CO2 =N2 .
Simple flame images for different conditions have The burner used was a gas turbine model com-
been presented along with exhaust gas emissions. bustor (DLR dual swirl burner), as shown in
They report lower nitrogen oxides concentrations Fig. 1. The burner design originates from a gas
ðNOx Þ for the quasi-oxyfuel flames and higher car- turbine combustor with an air blast nozzle for
bon monoxide concentrations (CO), suggesting liquid fuels, which is modified for gaseous fuel
stoichiometric operation at 20–24% O2 as ideal operation. This burner configuration was used in
for low emissions. Sautet et al. [7] studied the length several projects before for CH4 =air combustion
of natural gas/oxygen diffusion flames in a jet bur- under atmospheric and high pressure (detailed
ner for free and confined configurations. Fuel jet description of the burner[9]). So a detailed data
 P. Kutne et al. / Proceedings of the Combustion Institute 33 (2011) 3383–3390 3385

release zone. The OH chemiluminescence detec-

tion system consisted of an image-intensified
CCD camera (Roper Scientific, 512  512 pixels)
with achromatic UV lens (Halle, f = 100 mm).
OH chemiluminescence near 310 nm was detected
through an interference filter in the wavelength
region 295–340 nm. The exposure time for the
image intensifier was set to 400 ls to capture the
integrated (along the line-of-sight) spontaneous
emission from OH in the electronically excited
state ðOH Þ. For each measurement point 200
images were recorded.
OH -chemiluminescence signals are line-of-
sight integrated and thus the spatial resolution is
compromised. Moreover, the exact effects of strain
rate and the gas composition on the OH concen-
trations are not known in detail, especially for such
high CO2 concentrations. For that reason, it is not
possible to interpret the chemiluminescence images
quantitatively. Nonetheless, OH chemilumines-
cence signals are a qualitative marker for the
regions of high heat release and therefore a good
indicator of the position and shape of the flame
zone.
Fig. 1. Nozzle and combustion chamber of the DLR
burner. 2.3. Laser Raman scattering

base exists to compare the oxyfuel flames to nor- Laser Raman scattering was applied to measure
mal air combustion [9]. the major species concentration, mixture fraction,
The oxidiser mixture was supplied through a and temperature simultaneously. The setup con-
central nozzle (diameter 15 mm) and an annular sists of three double-pulse Nd:YAG laser systems,
nozzle (i.d. 17 mm, o.d. 25 mm contoured to an a pulse stretcher, a focussing lens, the detection
o.d. of 40 mm) as a co-swirling flow to the flame. optics, a spectrograph, and an ICCD camera. The
Between the two oxidiser flows the CH4 was fed three Nd:YAG lasers (Spectra-Physics PIV 400)
through a ring of 72 channels (0.5  l0.5 mm). provide six pulses at k ¼ 532 nm of 7 ns duration
The exit plane of the fuel inlet and the inner oxi- each at a repetition rate of 10 Hz which are com-
diser nozzle is located 4.5 mm below the exit plane bined to one beam. To reduce the intensity at the
of the outer air nozzle. The overall flow field of the windows and to avoid optical breakdown in the
flames is characterized by a conically shaped focal region the pulses were elongated by a pulse
inflow of fresh gas, an inner recirculation zone stretcher, and furthermore the laser beams were
(IRZ) and outer recirculation zone (ORZ) as focused by a combination of two perpendicularly
sketched in Fig. 1. In the shear layer formed arranged cylindrical lenses (f = 300 mm) to gener-
between the inflow and the IRZ, the mixing of ate a blurred focus of an average diameter
hot combustion products with fresh gas leads to 6 0:5 mm. In this way, a total pulse energy of
a continuous ignition and stabilization of the 0.9 J could be irradiated into the combustor. The
flame. The burner was mounted in an optically pulse stretcher consists of two beam splitters and
accessible combustion chamber. The chamber six mirrors, which divide each pulse into a number
measures 85  85 mm (cross-section) and is of individual pulses. After travelling along different
120 mm height. It consists of four planar quartz path ways, the individual pulses are re-combined
windows supported at the corners by steel posts yielding a pulse with duration of about 350 ns
(diameter 10 mm). The flow exits the chamber [11]. The Raman scattered light was collected at
through a rectangular to conical exhaust section. 90° by an achromatic lens system with an aperture
All gas flows are controlled via mass flow control- of 150 mm and relayed onto the entrance slit of the
lers, which where calibrated in house. The O2 and spectrograph (SpectraPro 300i, Acton Research,
CO2 are premixed upstream of the combustor ple- f = 300 mm, 490 lines/mm, f# = 4, dispersion
num by a mixing unit (Sulzer, type SMI-W). 6 nm/mm). For rejection of elastically scattered
light a holographic notch filter was placed in front
2.2. OH -chemiluminescence imaging of the entrance slit. A slit width of 1 mm, a slit
height of 14 mm and a magnification of the detec-
OH (electronically excited) chemiluminescence tion optics of 2 resulted in an imaged probe volume
was used to study the size and shape of the heat of 7 mm in direction of the laser beam and 0.5 mm
3386 P. Kutne et al. / Proceedings of the Combustion Institute 33 (2011) 3383–3390

in diameter. After spectral separation, the Raman 3.9% for CO2 , 2% for H2 O, 48% for CO, 7.37%
bands from the different molecular species ( CO2 ; for O2 and 6.8% for temperature. In order to take
O2 ; CO;N2 ; CH4 ; H2 O, and H2 ) were detected by into account daily variations in signal intensity
an intensified CCD camera (1340  1300 pixels) due to change in window transmittance air mea-
along the line of 7 mm. The images were binned surements were performed before and after actual
on chip to super pixels (268 points on the spectral flame measurement. A scaling factor was calcu-
axis and 28 on the spatial axis). This yielded 28 lated by taking ratio of the respective day reference
measurement volumes of 0.25 mm length each. measurement to laser intensity normalised calibra-
The sampling rate of the ICCD is 10 Hz and the tion air reference measurement.
gate time used was 1 ls. The energy of the total
beam recorded by a power meter before the mea-
surement volume was used for normalization of 3. Results and discussion
the measured Raman signals. Due to the presence
of carbon atoms in the oxidiser and in the fuel, it The stability of swirl stabilized oxyfuel=CH4
was not possible to define a mixture fraction by flames was studied for O2 mole fractions of 20–
using carbon-containing species. Therefore the 40%, equivalence ratios of / ¼ 0:5–1 and thermal
mixture fraction was derived from hydrogen-based powers of 10–30 kW. However, attempts of oper-
species. To determine the mixture fraction the ating the burner with 6 22%O2 were unsuccessful
concentration of all hydrogen containing species even with conditions of / ¼ 1:0 at 20 and 30 kW
(methane, water and hydrogen) were used to calcu- resulting in unstable operation and blow out.
late the element mole fraction of hydrogen Z H2 with Figure 2 shows the range of stable flame opera-
formula 1. The coefficient lH;Y denotes the molar tion for different oxidiser compositions and ther-
proportion of hydrogen in species Y. The coeffi- mal powers versus the Reynolds number. The
cient X Y is the mole fraction of species Y. The Reynolds number is calculated using the minimum
mixture fraction f was derived by formula 2, where effective flow cross-section of the burner outer co-
Z H2 ;flame is the value calculated at each measurement swirl nozzle (25 mm) with flow rate calculated for
position and Z H2 ;fuel is the value calculated from the standard temperature and pressure. The viscosity
original fuel composition. of the mixture is calculated using Wilke formula-
tion [12]. The plot shows two sets of iso-power
Z H2 ¼ lH;CH4  X CH4 þ lH;H2 O  X H2 O lines, one from 12; 000 < Re < 19; 000 represent-
þ lH;H2  X H2 ð1Þ ing 10–20 kW and the second from 19; 000 <
Re < 28; 000 for 15–30 kW. Each of the curves in
Z H2 ;flame 12; 000 < Re < 19; 000 range starting from the left
f ¼ ð2Þ represents equivalence ratios of 0.49, 0.54, 0.71 and
Z H2 ;fuel
1. Similarly for 19; 000 < Re < 28; 000 the equiva-
A calibration was used to calculate the species lence ratios are 0.51, 0.71 and 1. It can be seen that
concentration from the measured Raman signals. as the fraction of O2 in the oxidiser is reduced by
The temperature was calculated using the ideal increasing the amount of CO2 , the Reynolds num-
gas law. The calibration accounted for the temper- ber increases considerably for the same power.
ature and species concentration dependant Raman This is mainly due to the increased CO2 content
scattering signal intensity, as well as for the spectral for the same equivalence ratio. The plot indicates
signal overlap between species. The determination that as the O2 content in the oxidiser is increased
of this calibration was performed by measuring
Raman signal from reference gases and flames.
The reference gas measurements used were air,
CO2 ; CO;CH4 and H2 . The reference gases were
preheated to varying temperatures up to 800 K.
Hence providing temperature dependence up to
800 K. Methane/air flames varying in equivalence
ratio between 0.8 and 1.4 with temperatures from
1500 to 2100 K provide the higher temperature
range coefficients. These coefficients were used to
fit temperature dependant curves. To verify the
accuracy of the calibration curves, re-calculation
of calibration flame temperatures and species mole
fractions were carried out to compare with equilib-
rium composition and CARS temperature mea-
surements. The accuracy of the calibration was
estimated by calculating the RMS of the percent-
age difference to actual values for methane/air Fig. 2. Stable flame operation for different mixtures and
flames. The accuracy for individual species was power levels.
 P. Kutne et al. / Proceedings of the Combustion Institute 33 (2011) 3383–3390 3387

the flame can be operated stably for much leaner short exposures with background correction and
conditions. This can be attributed to two effects: normalization applied, so that intensities are com-
the flame speed and/or the Reynolds number. In parable between the images. Because of the rota-
oxyfuel combustion the O2 fraction of the oxidiser tional symmetry of the flame, the 2D images of
can be seen as an additional variable influencing the integrated signal could be deconvolved by a
the combustion behaviour. In fact, the O2 concen- three point-Abel inversion to derive a section of
tration has a large effect on the flame speed. Mea- the symmetry plane through the centre of the flame
surements of the laminar flame speed, which were [13]. None of the flames investigated burned
carried out as part of this project show, that the directly at the fuel nozzle; they were thus lifted
flame speed at 6 bar rises by a factor of 2.4 when and partially premixed before ignition. Starting
the O2 fraction is increased from 30–40%. Only at the left, the OH -chemiluminescence intensity
for O2 concentrations around 40% the laminar for the 26% O2 flame is relatively low. The flame
flame speed is comparable with the flame speed zone is distributed almost over the whole combus-
of a laminar CH4 =air flame (50 cm/s at 6 bar tion chamber, except along the inflow of fresh gas.
and 473 K preheat temperature). This behaviour There is combustion present even in the outer
is similar also for measurements at 3 and 10 bar. recirculation zone. While increasing the oxygen
Unfortunately no experimental results exist for fraction the flame zone moves upstream towards
atmospheric pressure, but it can be expected that the burner plate. At 30% O2 , the flame shape
the effect at atmospheric pressure is similar to that resembles those from stably burning CH4 =air
for pressurized cases. A change of the O2 concen- flames in the same combustor. The flames with
tration is thus expected to have a major effect on 34% and 38% O2 exhibited periodic thermo-acous-
the flame stability. The oxidiser composition also tic pulsations. While non-pulsating flames typi-
influences the flow field. For example, compared cally have a conical shape, the flames became
to 38% O2 oxidiser the 26% O2 oxidiser leads to a flatter with the onset of pulsations. This behaviour
40% increase in Reynolds number for the same was already observed in CH4 =air flames investi-
equivalence ratio at a power of 20 kW. This effect gated previously in this burner [9]. For CH4 =air
is described in detail below. The higher flame speed as well as for CH4 =oxyfuel, the flames were subject
and lower Reynolds number hence assist flame sta- to thermo-acoustic instabilities over a substantial
bilisation and lean operation at higher O2 levels. range of operating conditions. The adiabatic flame
The stabilisation behaviour does not change signif- temperatures with increasing O2 content of the
icantly with power, the flames at higher Reynolds flames are 1671, 1856, 2004 and 2140 K respec-
number have similar stability limits. tively (flame temperatures calculated with CAN-
The influence of variations of the oxygen frac- TERA). Higher flame temperatures increase, of
tion on the flame is further shown in Fig. 3. Aver- course, the reactivity leading to higher laminar
aged OH -chemiluminescence images for flames at flame speeds and shorter ignition delay times. This
21.4 kW thermal power and / ¼ 0:71 for 26%, change in chemical reaction rates along with the
30%, 34% and 38% oxygen (left to right) are reduced Reynolds numbers are expected to be
shown. The displayed images cover a field of view main causes for the change of the flame structure.
of 42:5 mm 6 r 6 42:5 mm and 0 mm 6 h 6 The results from the laser Raman scattering
110 mm, where r is the radial position relating to experiments allow for a more quantitative view
the burner axis, and h is the height above the bur- into the flame. In the following, results from a
ner. The images are the ensemble average of 200 flame with Pth ¼ 22:7 kW, / ¼ 0:76, 30% O2 are

Fig. 3. Abel-deconvoluted OH chemiluminescence images showing average flame structure at power of 21.4 kW and
/ ¼ 0:71 for different O2 mole fractions.
3388 P. Kutne et al. / Proceedings of the Combustion Institute 33 (2011) 3383–3390

discussed which is very close to the flame in Fig. 3b. region (see Fig. 3b) suggests that these gas mixtures
Figure 4 shows radial profiles of the temperature have just reacted. Alternatively, they could have
from this flame at different heights, ensemble aver- been convected as burned gas from further down-
aged over 300 laser shots at each measuring loca- stream within the IRZ. The samples with interme-
tion. Near the nozzle, at h = 5 and 10 mm, the diate temperature contain CH4 and O2 and
temperature in the inlet region (r  5–15 mm) is probably represent mixtures of burned and
close to room temperature. The scatter plot of tem- unburned gas which has not reacted yet due to igni-
perature and mixture fraction from this region, dis- tion delay. The radial temperature profiles in Fig. 4
played in Fig. 5, yields a more detailed view of the show a rather flat profile with T  1400 K at h = 5
thermo-chemical state of the flame. Here, each and 10 mm for r > 20 mm (Raman measurements
symbol represents the result from a single-shot could be performed up to r = 32.5 mm; the remain-
measurement at h = 10 mm, r = 10 mm and the ing 10 mm to the combustor wall were not accessi-
dashed line shows the result from an adiabatic ble due to restrictions of the optical arrangement).
equilibrium calculation. The vertical line indicates This region is supposed to belong to the ORZ. A
the global mixture fraction deduced from the mass scatter plot from this region (Fig. 5) exhibits a rel-
flow rates. All four measurement positions of Fig. 5 atively homogeneous gas mixture close to the glo-
are marked in Fig. 3b by crosses. There is a large bal mixture fraction, and an inspection of the
scatter in mixture fraction reaching from f = 0 molecular composition confirms that it is burned
(pure oxidiser) to about f = 0.15 (corresponding gas. The temperature is, however, approximately
to / ¼ 2:96) that reflects the degree of unmixedness 300 K below the adiabatic temperature. This tem-
in this partially premixed flame. Most of the sam- perature loss might be caused, to some extent, by
ples exhibit temperatures around 300 K, i.e., they thermal radiation and certainly to heat loss at the
have not reacted yet. The samples with increased combustor wall and burner plate. Similar effects
temperature represent mixtures of fresh gas with have been observed before in CH4 =air flames [10]
some admixtures of exhaust gas from the recircula- and are typical for samples from the ORZ. The
tion zones. Going back to the radial profiles of radial temperature profiles in Fig. 4 further show
Fig. 4, one can see an increase in temperature from that the temperature level increases with height,
the inflow towards the flame axis. Along the axis, as expected, due to reaction progress and reaches
the temperature gradually increases from T = the final state at h = 60 mm. It is, however, inter-
800 K at h = 5 mm to T = 1900 K at h = 60 mm. esting to note that the reaction progress is typically
From velocity measurements in CH4 =air flames not characterized by one step from unburned to
performed previously in this burner it can be burned mixtures, but rather through intermediate
assumed that this region lies within the IRZ [9]. states. This is illustrated in the scatter plot from
The scatter plot from h = 10 mm, r = 0 mm, dis- h = 35 mm, r = 23 mm, in Fig. 5. The samples
played in Fig. 5, shows that the mixtures are rela- exhibit temperatures between T = 500 K and
tively fuel rich with a large variation in T = 2100 K and almost the complete range of reac-
temperature. The samples with high temperatures tion progress from non-reacted to completely
(T  1700 K) are completely reacted and an reacted. Again, the intermediate temperatures are
inspection of the composition reveals that there is most likely interpreted as mixtures of burned and
no CH4 present, but small amounts of O2 . The high unburned gas which have not fully reacted yet
chemiluminescence intensity observed in this due to ignition delay.
A qualitative comparison of these results with
CH4 =air flames from previous studies [9,10]
reveals some differences. The temperature level,
and thus the reaction progress, in the central
region close to the nozzle is smaller in this flame.
This is probably due to a different extension of the
IRZ. Here, velocity measurements that have not
yet been performed could certainly yield more
information. Further, the temperature loss in the
ORZ is more pronounced in this flame. This
might be caused by a larger contribution of ther-
mal radiation of CO2 in comparison to N2 and
also by a higher heat transfer rate to the wall of
CO2 compared to N2 .
With regard to gas turbine conditions the flame
temperature plays an important role. Higher tem-
perature implies higher efficiency of the turbine.
Fig. 4. Radial profiles of mean temperature at different This is manly limited by the material used at the
heights above the burner for the flame with turbine inlet, and for today’s gas turbines limited
Pth ¼ 22:7 kW, / ¼ 0:76; 30% O2 . to 1600 K. For oxyfuel flames the adiabatic
 P. Kutne et al. / Proceedings of the Combustion Institute 33 (2011) 3383–3390 3389

Fig. 5. Raman scatter plots for a flame with Pth ¼ 22:7 kW, / ¼ 0:76, 30% O2 , showing the correlation of temperature
and mixture fraction at four different positions in the flame. r is the radial position relating to the burner axis, and h is the
height above the burner. Dashed line represents results from adiabatic equilibrium calculations and vertical line indicates
global mixture fraction.

flame temperature at 21% oxygen is significantly conditions in a gas turbine burner, which was
lower compared to CH4 /air flames at the same optimized for CH4 =air combustion, might not be
stoichiometry. Comparable flame temperatures possible. Other burner configurations, which take
are reached for an oxygen fraction of 30%. into account the low flame speed, could improve
Because of the high cost of air separation and to the flame stabilisation for such cases. But the com-
have less oxygen diluting the flue gas, it is aspired bustion behaviour of the oxyfuel flames can
to burn oxyfuel flames with almost no oxygen in change under pressurized conditions. Therefore
excess near / ¼ 1:0. For those conditions the adi- pressurized experiments will be carried out in
abatic flame temperature for an oxyfuel flame 2010, in the framework of the BIGCO2 project.
with 30% oxygen is 2160 K, much to high for a
turbine. To reach acceptable temperatures the
oxygen level would have to be lowered to 18%, 4. Summary and conclusions
which corresponds to an adiabatic flame tempera-
ture of 1600 K. As stated above, such a low oxy- The combustion behaviour of partially pre-
gen concentration would cause a very low mixed CH4 =oxyfuel flames in a swirl stabilized bur-
laminar flame speed. The results of the measure- ner for O2 fractions of 20–40%, an equivalence
ment show, that even for 30% O2 the flame is less ratio of / ¼ 0:5–1, and thermal powers of
stable than a methane/air flame in this burner. Pth ¼ 10 kW to 30 kW was characterized by using
For lower O2 concentrations a stable operation OH -chemiluminescence to monitor the flame
with this kind of burner is not possible. The structure and flame stability. Flames could be sta-
results from the atmospheric measurements imply, bilized over a wide parameter range. The O2 frac-
that oxyfuel combustion under technical relevant tion had a strong influence on the flame shape,
3390 P. Kutne et al. / Proceedings of the Combustion Institute 33 (2011) 3383–3390

whereas changes in equivalence ratio only had mar- research program Climit. The authors acknowl-
ginal effects. At some of the operating conditions, edge the partners: Statoil, GE Global Research,
the flames exhibited periodic thermo-acoustic Statkraft, Aker Clean Carbon, Shell, TOTAL,
instabilities. For a flame with Pth ¼ 22:7 kW, ConocoPhillips, ALSTOM, the Research Council
/ ¼ 0:76 and O2 concentration of 30%, single-shot of Norway (178004/I30 and 176059/I30) and
laser Raman measurements have been performed Gassnova.
for the simultaneous determination of the major
species concentrations, mixture fraction and tem-
perature. A significant variation in mixture fraction References
and reaction progress was observed with non-
reacted, partially and completely reacted samples. [1] McKinsey and Company, available at: <http://
Compared to CH4 =air flames which have been www.mckinsey.com/-clientservice/ccsi/pdf/
CCS_Assessing_the_Economics.pdf>.
investigated previously in the same burner, the [2] H. Jericha, W. Sanz, K. Woisetschläger, M. Fesha-
CH4 =oxyfuel flames studied here showed a differ- raki, in: CIMAC Conference, Interlaken, Switzer-
ent combustion behaviour. For the stabilization land, 1995.
of these flames, the O2 fraction in CO2 had to be lar- [3] H. Jericha, W. Sanz, E. Göttlich, F. Neumayer,
ger than the 21% in air. The main reasons are obvi- ASME Turbo Expo (2008), GT2008-50515.
ously the larger heat capacity of CO2 compared to [4] K. Foy, E. Yantovski, Int. J. Thermodyn. 9 (2)
N2 that leads to lower combustion temperatures for (2006) 37–63.
the same equivalence ratio and the lower laminar [5] N. Zhang, N. Lior, Energy 33 (2008) 340–351.
flame speed. The Raman measurements revealed [6] T.C. Williams, C.R. Shaddix, R.W. Schefer, Com-
bust. Sci. Technol. 180 (2008) 64–88.
a significant difference between the temperature [7] J.C. Sautet, L. Salentey, M. Ditaranto, J.M.
measured in the outer recirculation zone and the Samaniego, Combust. Sci. Technol. 166 (2001)
adiabatic flame temperature. This temperature loss 131–150.
is probably due to thermal radiation and heat loss [8] M. Ditaranto, J. Hals, Combust. Flame 146 (2006)
to the walls of the combustor. The heat loss was lar- 493–512.
ger than previously observed for CH4 /air flames. [9] P. Weigand, W. Meier, X.R. Duan, W. Stricker,
There were also indications that the inner recircula- M. Aigner, Combust. Flame 144 (2006) 205–224.
tion zone does not extend so far upstream as it was [10] W. Meier, X.R. Duan, P. Weigand, Combust. Flame
measured for CH4 =air. 144 (2006) 225–236.
[11] H. Ax, P. Kutne, W. Meier, K. König, U. Maas,
A. Class, M. Aigner, Appl. Phys. B 94 (2009) 705–
Acknowledgments 714.
[12] J. Warnatz, U. Maas, R.W. Dibble, Combustion,
Springer, 2006, p. 64.
This publication forms a part of the BIGCO2 [13] C.J. Dasch, Appl. Opt. 31 (1992) 1146.
project, performed under the strategic Norwegian