You are on page 1of 11

Proceedings of ASME Turbo Expo 2016

GT2016
June 13-17, 2016, Seoul, South Korea

Draft: GT2016-57223

Validation and Application of Chemical Kinetic Models for Enhancing Gas Turbine
Flexibility

Felix Güthe, Martin Gassner, Stefano Bernero, Thiemo Meeuwissen, Torsten Wind
GE Power
Baden, Switzerland

ABSTRACT This example demonstrates nicely how a development over


In recent years, market trends towards higher power several years starting from fundamental basic research over
generation flexibility are driving gas turbine requirements of experimental validation finally delivers a product for power
operation at stable conditions and below environmental plants. The GT24 / GT26 can not only be operated with H2
emission guarantees over a wide range of operating conditions, containing fuels, but also at very low part load conditions and
such as load, and for changing fuels. In order to achieve these with the integration of H2 from electrolysis (~power to gas
targets, engine components and operation concept need to be ~PTG) the turndown capability can even be further improved.
optimized to minimise emissions (e.g. CO, NOx) and In this way the energy converted at low electricity prices can be
combustion instabilities, as well as to maximize component stored and utilised at later times when it is advantageous to run
lifetime. Therefore the combination of field experience, the GT at lower loads increasing the overall flexibility. This
experimental studies and theoretical modelling of flames with development is well suited to integrate renewable energy at
state of the art tools play a key role in enabling the highly fluctuating availability and price to the energy
development of such solutions. provisioning by co-firing with conventional fuels.
For many applications the relative changes of reactivity
due to changes in operation conditions are important thus in INTRODUCTION
this report a few examples are shown, where chemical kinetics
simulations are used to determine the reactivity and to predict Motivation and objective
engine behaviour. The predicted trends are validated by
In today’s volatile utility markets, the challenges for the
correlating them to validation data from high pressure test rigs
power generation business include not only highly efficient
and real gas turbine operational data. With this approach the operation and strict emission compliance, but also an increasing
full operational range from highest reactivity (flashback) to
demand for flexibility with respect to load and fuels. The
lowest reactivity (blow out or CO emission increase) are integration of renewable energy is expected to even increasing
covered. The study is focused on the sequential combustor the demand for flexible power generation and storage, which
(SEV) of reheat engines and addresses both the safety margins
require balancing measures for load flexibility and energy
with respect to highly reactive fuels and achievable load storage. According to several recent studies on the German and
flexibility with respect to part load CO emissions. European energy market, the fraction of negative residual
The analysis shows that it is necessary to utilize updated
electricity prices will rise from several 100 h per year to over
kinetic mechanisms since older schemes have proved to be 1000 h per year [1, 2, 3].
inaccurate. A version of the mechanism developed at NUI This gives rise to promising business opportunities for
Galway in cooperation with Alstom and Texas A&M was used
energy storage applications. For gas turbine (GT) power plants
and the results are encouraging, since they are well in line with this could include power to gas applications [4, 5], where
experimental test data and can be matched to GT conditions to excess power can be electrolysed to produce H2 that can be
determine, predict, and optimize their operational range.
stored, transported and sold or consumed in GT at times when it

© 2016 by General Electric Technology GmbH. All rights reserved. Information contained in this document is indicative only. No representation or
warranty is given or should be relied on that it is complete or correct or will apply to any particular project. This will depend on the technical and
commercial circumstances. It is provided without liability and is subject to change without notice. Reproduction, use or disclosure to third parties, without
express written authority, is strictly prohibited.
1 Copyright © 2016 by General Electric Technology GmbH
is technically beneficial and economically rewarding [6, 7]. The and load points. This GTs have are to cope with a variety of
economy of such application depends on the value and the fuels and allow for emission-compliant operation at very low
usage of the produced H2. Possible utilisations include the loads [8, 9, 10].
addition to the gas pipeline or usage for automotive mobility. With the objective to demonstrate the utility and benefits of
Another application is to utilise the produced fuel in the gas such an approach, this paper recapitulates the major validation
turbine to reduce fuel costs. steps of an detailed kinetic reaction mechanism for natural gas
This can also be achieved after methanisation, i.e. the use blends including significant amounts of higher hydrocarbons
of H2 for reduction of CO2 to CH4, but direct H2 usage allows (C2+) and hydrogen, and demonstrates the benefits of such an
much higher efficiency of the overall process and additionally approach for the design of new products and upgrades targeting
can be exploited to extend operational flexibility by enabling flexibility.
GTs to operate at conditions where they would normally not This work demonstrates the fuel flexibility with the goal to
operate. It will be shown that the GT operation can be extended further increase the load flexibility utilising the fuel flexible
into the part load range at times when electricity prices are low combustor and the converted renewable energy (PTG-H2) at
and availability at low actual fuel consumption cost is lowest cost to operate independently over an even wider range
rewarding. of energy prices while remaining connected to the grid.
Integrating the PTG- conversion into the scope of power
plant operation offers some interesting economic opportunities Development steps
for highly flexible power plants become interesting. A The development of new combustion features can be
prerequisite for this is a GT which is highly flexible with separated into several steps:
respect to load and fuel variations, which is capable to produce 1. As first step of the fuel flexible GT pants development the
energy from natural gas co fired with H2 from electrolysis or combustion behaviour has to be predicted using proper
even from gasification plants. Apart such PTG-applications, chemical kinetics tools including kinetics at GT condition
combusting fuels of varying composition is a relevant challenge for natural gas and hydrogen fuels, since many older
for syngas applications (e.g. co-firing of refinery gases, or chemical mechanisms are not accurate enough. The
generated through gasification), and for power plants that are chemical parameters of interest are calculated for GT
affected by major changes of the available gas (e.g. due to conditions [11].
seasonal variation of suppliers through pipelines and/or LNG). 2. The next step includes the validation in the laboratory [12]
The relevant parameters for combustion thereby include and for wider conditions included in the chemical kinetics
not only energetic and volumetric properties such as heating mechanism [13]. The mechanisms are developed in
value and Wobbe index, but also its reactivity (i.e. ignition cooperation with the National University of Galway
properties and flame speeds), which significantly affects (NUIG) and Texas A & M University (TAMU) over the last
combustion stability and emissions. Since these characteristics decade to update the hydrocarbon chemistry for higher
are particularly sensitive to pressure, high pressure tests in alkanes and to include H2 as fuel.
engine conditions are usually considered a mandatory step in 3. A further step in the development is the high pressure (HP)
the development of new combustors or their upgrade. High validation [14] on full scale test rigs, which is highlighted in
pressure testing is yet complex and very costly, and is therefore this work.
to be kept at the strict minimum. In past years, this has 4. As last step (4) the field engine validation and experience in
motivated the development of accurate kinetic reaction models commercial operation can lead to validated co-firing
to capture the behaviour of a variety of fuel blends. The concepts for C2+. Based on the validation results, the
application of such models in the design phase has the potential confidence gained with the implemented tools and process
to reduce development time and the need for extensive testing can enable the extension of such concepts to include H2
of new hardware or alternative operation concepts, and thus fuels.
significant improve cost, quality and time-to-market.
In this scenario combustion in the GT is still crucial to the
OPERATIONAL PRINCIPLES AND BOUNDARIES OF
power generation schemes of the next decades, even at reduced
REHEAT COMBUSTORS
fossil fuel consumption and carbon foot print this technology
will be required for stabilization of the grid to allow the The combustor of reheat engines can be operated over a
integration of renewable energy and their optimised utilisation. wide range of conditions, such as fuel composition and power
Well suited for such application are modern GTs like reheat output at very high efficiency. Limiting the operation of a gas
GT26 / GT24 with sequential combustion [8]. They have been turbine combustor are the lean blow-off limit (often indicated
described earlier and consist of two consecutive combustors by increasing pulsations or CO emissions as a precursor and
separated by a high pressure turbine. The two combustor finally leading to flame blow out) and at the rich limit flashback
operate in different regimes and some detail of the burner events which risk to reduce the lifetime of combustor and
development has been reported [9]. They are developed for low burner hardware. In reheat engines, the two combustors are
emissions and high flexibility and a load change between the coupled [8] through the fact that the first combustor’s output
combustors allows stable operation over a wide range of fuels (temperature and composition) is determining the sequential

2 Copyright © 2016 by General Electric Technology GmbH


combustor’s input. Consequently both combustors have to be
optimized and kept in the stable operation range, but through
their coupling both follow the same trend (with fuel reactivity)
allowing an elegant compensation and the fuel split between
first and second fuel flow even offers additional freedom.

Figure 1: EV Flashback. Increased reactivity in the EV


burner flow field causes the flame to move upstream

First (EV) combustor: Flashback and Lean Blow-Off


The first combustor is usually operated at lower
temperature and uses an aerodynamic stabilization mechanism
like a vortex stabilized premixed combustion (EnVironmental
vortex - EV) and produce hot gas as inlet into the high pressure
turbine (HPT) and sequential combustor (SEV). The reactivity
of the EV is dominated by the turbulent flame speed (st) and
parameters that influence that, like turbulence and laminar
flame speed (sl). The latter is changing with fuel composition
and with hot gas temperature Thg. Mixing quality and stability
can also be influenced by the staging ratio of the two fuel
stages within the EV burner. An increase of reactivity causes
the flame to move upstream against the gas flow as Figure 2 NOx values for interpolated to fixed Thg = Tref .
schematically indicated in Figure 1. plotted vs. fuel content (C2+ & H2) and plotted vs. Wobbe
For aerodynamically stabilized flames, when increasing index
reactivity an operation point can be reached where the flame While the data plotted vs. Wobbe index seem to be
jumps to a new location, as indicated by thermocouple uncorrelated, the plot vs. C2+&H2 reveals some correlation
readings, changes in pressure drop and sudden increase of NOx with higher NOx for fuels of higher reactivity, assuming a linear
formation in the performed tests that have been reported earlier relationship to C2+&H2 and similar contribution of C2+ and H2
[14]. H2 contents of over >50%vol have been tested at base to reactivity. A more accurate description could eventually
load and part load conditions and largely increased Thg improve the match, but within the required accuracy and for the
(+200K). Although signs of flame movements were observable sake of simplicity a linear index is preferred for concentrations
at high Thg, no damage on burners was observed and the H2 < 60% C2+&H2. The NOx increase can be compensated by
content was limited by the fuel supply system rather than the reducing Thg, which is also required for the reduction of SEV
burner performance. inlet conditions. Not that also the lean blow out (LBO) limit
The EV burner has proven to be robust against reactivity shift with fuel reactivity.
changes due to fuel composition variation. NOx emissions seem On the low reactivity end the burner limit is given by the
to scale with reactivity due to changes in flame position but no flame moving away from the stabilizing vortex out of the
clear impact of changes in the Wobbe index can be identified burner into the combustor. This can be accompanied by
(Figure 2). Plotted are data from series of reference fuel eventual pulsations and CO increase before actual blow out.
(natural gas) and additions of H2, N2, C2+ and C2+&H2, where Both limits can be related to the turbulent flame speed st and
C2+ is the sum of all higher hydrocarbons in the fuel are subject to ongoing studies for a wide range of conditions,
composition and H2 is the hydrogen content. The NOx shown fuels [15, 16] and pressures.
are not directly measured but interpolated from HP data series The weak correlation of the Wobbe index with the NOx
to correspond to the same Thg. demonstrates the small impact on the mixing quality of the EV

3 Copyright © 2016 by General Electric Technology GmbH


compared to the flame position and fuel reactivity. The Wobbe also the thermocouple reading (Figure 3 middle left) increases
index is useful to assess the impact of fuel changes on fuel proportional to the increase of heat input by the vitiated air, but
pressure drop for given burners and fuel handling systems. The also tign decreases so that the flame approaches the
same quantity also impacts jet penetration when fuel is injected thermocouple and additional heat input increases the
into the burner, but any impact on burner mixing and NOx thermocouple reading indicating the vicinity of the flame tip.
emission of the EV-burner is secondary compared to the flame This is visible as kink in the TC reading. The Delta TC curve
moving according to its reactivity. (Figure 3 lower left) shows schematically the difference
The correlation to C2+&H2 is not perfect but shows a between the expected temperature caused only by TSEVin
reasonable trend. Note also that the scaling of fuel composition increase and the measurement. Although no damages to the
with reactivity does not have to be linear, nor has the addition burner have been detected this point was used as boundary to
of different fuel components (C2, C3, or H2) to be of similar limit the operation of the burner and has been termed flashback.
magnitude. Therefore the scaling of NOx with the combined The lean limit of operation has been determined by the
index C2+&H2 is remarkably good and useful for adjustment of completion of the CO oxidation within the gas turbine. Since
GTs. oxidation is slowed down by cooling, further reaction in the
turbine can be neglected and only the CO oxidation within the
combustor is relevant.
Based on time scales, a qualitative criterion (Equation 1)
was formulated [17] for CO emissions relating the combustor
residence time (tres) to the chemical time scales tign, which is
mainly depending on the inlet temperature and tburnout mainly
depending on the hot gas temperature of the SEV combustor
(~TLPT in):
t res  tign (TSEV _ IN )  tCO_burnout(Thotgas)
Equation 1
In general, CO emissions can be low (near the equilibrium)
when the residence time is longer than the required overall
Figure 3: SEV flashback from [14]. SEV burner and reaction time to equilibrium, i.e. the time for the ignition and
schematic flame position on the left and temperature level oxidation of CO.
on the right
MODELLING APPROACH
SEV combustor: Flashback and CO emissions For the sequential combustion both combustors have to be
The SEV (Sequential EnVironmental) is an auto-igniting kept within the stability limits of each combustion system.
combustor where fuel is injected into the hot exhaust gas of the What at first sight seems to be an additional constraint but
high pressure turbine (see [8] for details). The combustion of actually turns into an advantage is the coupling of the first and
the gas then consists of several (overlapping) phases and time the sequential stage. A change of fuel reactivity can be
scales, which are governed by different parameters: ignition, compensated with a change of fuel split between the two
heat release and CO burnout. The kinetics in the ignition phase burners retaining the combustor stable at constant power.
are principally dependent on the temperature and composition It is assumed that exact matching of flame shape and
of the reactants (i.e. THPT_out ( ~TSEV_in), fuel/air ratio, fuel position is not required for an accurate prediction of trends with
composition and temperature). The ignition time tignition varying fuels and inlet conditions. The hypothesis is rather that
determines the flame position within the combustor. The heat scaling according to physical principal parameters seems useful
release and CO oxidation time scales are mainly governed by to predict GT behaviour and emissions. Other than for NOx
the hot gas temperature (Thg) as well as fuel composition. The predictions, where mixing plays a major role, for the LBO and
inlet temperature influences the auto-ignition delay time and CO predictions premixed conditions appear sufficient in the
therefore the residence time after ignition. This again is reactor calculations coupled to experimental validation. Any a
indirectly controlling the CO burnout. priori quantitative CO prediction would have to include more
The operational limits of the SEV are between flashback physical effects like mixing, leakages of cold streams or
and LBO, which is usually indicated by gradual rise in CO for quenching effects of the CO oxidation.
the sequential combustor, rather than pulsations or The relative reactivities of the two combustors can be
instantaneous flame loss. The latter turns out to be a convenient calculated using chemical kinetics simulations like the laminar
feature of the reheat flame. The flashback has been discussed flame speed solver PREMIX or the perfectly stirred reactor
earlier [14] and is demonstrated in Figure 3. Depending on the (PSR) for the first (EV) combustor operating at lean premixed
flame reactivity the flame moves upstream towards the injector. conditions and a plug flow reactor (PFR) for the reheat
When the inlet temperature is increased (Figure 3 upper left) combustor (SEV). Solvers used are from either the Chemkin

4 Copyright © 2016 by General Electric Technology GmbH


Pro program package from Reaction Design/Ansys or a For a more quantitative assessment of the CO burnout the
corresponding open source software like Cantera. ignition and oxidation times are normalised to a scale
representing the residence time, which provides a
Kinetic models dimensionless quantity that characterises the time required for
The choice of the chemical reaction mechanism proves to complete burnout vs. the time available in the combustor. The
be important as will be shown later. The validation of the reduced reaction time (RTR) is defined in Equation 2. As a
chemical models for the correct conditions is crucial when rough estimate, the residence time scale is assumed
using the model predictions for GT operations. The model used proportional to the ratio of hot gas volume flow at the
here is based on the work done in former years incrementally combustor exit (including also cooling air) and the geometrical
increasing confidence by validation on all levels of maturity volume of the combustor.
[11, 12, 13, 14] with the engine validation demonstrated here. . t ign  t burnout
The kinetic model is under continuous improvement at NUI RTR 
Galway and is referred to as NUIG model. We have used a t res
version of the AramcoMech1.3 (2013) [18], with minor Equation 2
modifications for improved numerical convergence. To give an
impression of the importance of choosing the right model and
of the improvements that has been made to the chemical
kinetics in matching GT conditions some results are compared
to the GRI3.0 model [19], which still serves as standard for
many applied researchers.

Premix combustor model


The reactivity in the first premix combustor (i. e. EV) can
be described by the laminar flame speed sl calculated for
example with the PREMIX code or the PSR extinction time
tPSR_ext , which is defined [14] as the shortest residence time of a
PSR that allows ignition to take place. In turbulent flow fields
the extension to the turbulent flame speed st seems desirable.
For natural gas mixtures st correlations (described in [15, 16])
are useful while they tend to be less accurately validated for Figure 4 CO-profile for the reheat -tignition flame
predictions of H2 and C2+ containing fuels over varying Tinlet
Theoretically, a value of RTR = 1 represents just sufficient
and Thg. Especially for the atmospheric testing and
time for complete burnout, and lower values indicate
extrapolations to higher pressures this leaves some gaps to be
insufficient residence time and thus incomplete conversion of
filled by expensive full scale testing.
CO. Due to the simple representation of the SEV combustor
To understand the influence of different fuels or other
(i.e. a closed, uncooled plug-flow reactor, perfectly mixed inlet
parameters on the stability of the burners the chemical
conditions, usage of average quantities instead of detailed local
reactivity needs to be scaled in the turbulent field.
conditions), this phenomenological approach does yet not
attempt a direct prediction of CO emissions from reaction
Sequential combustor model
kinetics. Instead, it provides a non-dimensional indicator
For the sequential combustor one combustion limit is given quantifying the relative influence of the governing phenomena,
by the highest reactivity or shortest allowable tign, which is to which CO emissions can be correlated.
referred to as flashback, while the lower end of reactivity LBO
is indicated by insufficient CO burnout and the rise CO VALIDATION WITH RIG DATA
emissions. In the simulation within the PFR this is resembled
Experimental validation of the kinetic predictions are done
by the combination of tign and tburnout as indicated by Equation 1.
at full scale combustors at DLR Köln at full pressure and are in
Both quantities can also be predicted as shown in Figure 4.
part described in [14]. Test were done for a wide range of fuels,
The two cases modelled both represent part load
changing C2+, H2, N2 and fuel temperature as well as operating
conditions, but one with higher TSEV_in (and shorther tign) and
conditions.
lower Thg (and longer tburnout) compared to the other extreme.
Two types of tests are highlighted here:
The ignition time tign can be defined on the profile of
1. For the flashback testing the test was started with a given
temperature (i.e. 3/4 of T-rise) or species (i. e. CH, or CH2 ) and
fuel at low TSEVinlet. As explained in Figure 3 TSEVinlet was
the burnout time tburnout can be defined as CO passing a value of
then increased to force the flame moving inward as
1.5 times the equilibrium value. All definitions seem equivalent
observed on the thermocouple reading. From the continuous
but should be used consistently for comparisons.

5 Copyright © 2016 by General Electric Technology GmbH


data recordings (~1 Hz) the corresponding condition were GRI 3.0 seems not appropriate to lay out a combustor operation
extracted and calculated (Figure 5). concept for varying fuels while the NUIG gives satisfactory
2. For the LBO tests the CO were carefully measured for each results without tuning of the results.
fuel at defined TSEVinlet and Thg allowing sufficient time for
equilibrated emission measurements (Figure 6). CO emissions
The second criterion concerns the LBO which is indicated
Flashback behaviour by increased CO emissions. Since the flame is not lost the GT
The flash back test were conducted [14] for natural gas can be operated close to a defined guarantee level at part load.
fuels with up to 45% H2 and 35% C2+ added. With this Experimental data for some selected fuels and conditions are
procedure ca. 20 cases were recorded were the flame was in the shown in Figure 6. CO emissions at part load depend on TSEVin,
vicinity of the thermocouple indicating a similar flame position fuel reactivity and on Thg. Natural gas is taken as reference
and varying TSEVin. These cases (for given fuel and TSEVin) were (green line). Note that two different test runs yield almost
analysed using the PFR model including a given fraction of identical results. At reduced Thg (upper graph blue) the CO
cooling air from lance and burner. The obtained values for tign emissions increase for given TSEVin reducing the CO emission
all correspond to the same flame position. They are plotted in compliant operation.
Figure 5 as histogram (note the log scale for distance) for the
distance that is derived from multiplication of tign with the mean
burner velocity.

Figure 5: Histogram and distribution probability of


ignition distance calculated from the flashback
condition for different fuels (C2+, H2 and N2) and
burner temperature (log spacing)
The position of the thermocouple is given as black square
around 0.22m and very near the centre of the distribution of
ignition events for the NUIG model (events shown in red bars-
centre shown by the green arrow). The GRI30 results are shown
as blue bars and are 2-3 times shorter. Also the accuracy to
predict the ignition position for different fuels show relatively Figure 6: Measured CO values for low load operation
small scatter (10%) with the NUIG model while the GRI data with varied fuel, TAT vs. Thg (lower graph) and vs TAT
seem less consistent (25% scatter). The scatter is given by the (upper graph)
standard deviation relative to the mean. For comparison a
The lower graph shows the CO increase with decreasing
Gaussian distribution from mean and standard deviation is
Thg, which means decreased power output and fuel
plotted as lines (blue –dashed for GRI and red continuous for
consumption. This can be mitigated in part by operating at
NUIG) and actual events are indicated by symbols.
higher TSEVin (red graph) or with more reactive fuel (yellow).
The comparison of the magnitude and the scatter of results
The dilution with N2 is accompanied by cooling and leads to
indicates the importance of the accurate kinetic model for
less reactive conditions and increased CO emissions. The
predictions and of burner behaviour also in GT operations. The

6 Copyright © 2016 by General Electric Technology GmbH


largely reduced CO emission with 20% H2 fuel basically are mechanisms while the GRI data are much more scattered. For
promising for low load operation in compliance with high reactivity the CO values are near the equilibrium values as
environmental regulations. expected. In the logarithmic plot the CO-values seem to start
While some trends are observable, an overall picture is increasing for RTR>1 while the GRI data scatter around half of
maybe difficult to extract from the graph. Only a selection of that values. As indicted by the hand-drawn average line for low
the available measurements has been chosen for the graph since reactivity (high ) the CO values approach a fixed value where
some other data are referring to slightly different conditions and either all fuel is converted to CO or no reaction takes place at
therefore should not be connected by one line. For example all, corresponding to low CO for very high tign.
several C2+ points were not included. This kind of plot enables to interpret all fuel variations in
A better understanding is obtained if all measurement one plot rather treating each fuel separately and therefore uses
points could be included into one larger validation dataset and validation on a large data base. It validates the approach to use
put into one graph. To achieve this, the reduced reaction time chemical kinetics simulations for full scale combustor
RTR (Equation 2) is calculated for measurement points at behaviour and includes C2+, N2, but also H2, for which GT data
varied and fuel compositions and temperatures. are rare. It also highlights the importance to use the validated
For complete burnout the CO concentration is reaching the accurate model since this would not give useful correlations
equilibrium value regardless of the residence time or reaction with the inaccurate model (i. e: the GRI 3.0).
rates. This is an important boundary case to refer the data to. The next step of validation of the approach is the
Since the equilibrium value is changing with Thg the data would application of the model to real GT data and the transfer of the
look more scattered if they were plotted as absolute values. In single burner high pressure with H2 to engine validation.
Figure 7 the CO values are normalised to that equilibrium
value. ASSESSMENT WITH FLEET DATA
As a last validation step of the kinetic mechanism and
engine models, its predictions are assessed with available
adjustment and test data from GT26 fleet engines. Compared to
laboratory experiments (e.g. shock tubes) and combustor rig
tests, engine data can thereby be expected to show larger scatter
due to a number of additional parameters whose influence
cannot be accurately determined or controlled. GT24 / GT26
gas turbines are featured with annular combustion chambers
that unite up to 24 individual burners in a mechanically
sophisticated assembly [20]. Manufacturing and assembly
tolerances, and resulting differences in cooling and leakage
flows thereby cause local flow conditions through the
Figure 7: Measured CO values (normalised to CO individual burners to marginally differ, which can affect the
equilibrium) for low load operation with varied fuel (H2 and overall behaviour of the combustor. CO emissions are thereby
C2+ up to 45%), TAT and Thg vs. reduced reaction time for particularly sensitive due to the strong influence of local
the updated model (NUIG) and the GRI3.0 (inserted) . conditions around the flame zone on the burnout. A part this
direct influence, indirect effects are caused by the need for
For short RTR (and complete combustion) a value of 1 is controlling the engines’ pulsation behaviour. As a specific
expected but for some points actually lower values (by a factor measure, combustor staging, i.e. the creation of temperature
1-3) are measured. This depends on the specific probe and inhomogeneities in the annular combustor to uncouple flame-
sample treatment since the CO concentration is still decreasing structure interactions, results in a few colder burners that
while cooling the extracted combusted gases within the produce over proportionally more CO than compensated by the
sampling probe before it is quenched to a value that is finally hotter-than-average combustion zones.
detected. For hot temperature and high CO values this reduces The following paragraphs illustrate the validation of the
the concentration further, while the effect is less important for kinetic models for an individual engine and a representative
low temperatures and higher CO values like the part load cases subset of the GT26 fleet. The validation is thereby only feasible
considered here, where the CO reaction is quenched before the with respect to CO behaviour and subject to available fuel
combusted gases are sampled. composition on-site. A flashback assessment as carried out in
The CO values vs. RTR are plotted in Figure 7 for all part the rig tests is not shown since fleet engines have not been
load operation point calculated with the NUIG model and the instrumented for this purpose. They are inherently kept safe by
GRI3.0 model (inserted). Shown are data from three different appropriate control measures that ensure sufficient margin.
test runs over all fuels (summarising H2, C2+, N2 and Tfuel
variation) and temperature variation. The coherence of the data
is not as good as for the flashback data but a clear correlation of
CO emissions to RTR can be seen for the validated NUIG

7 Copyright © 2016 by General Electric Technology GmbH


Individual engine characteristics relationship in the data, and results in the formation of several
In analogy to the rig test data presented above, Figure 8 clusters. At the part load emission target of 100 [mg/m3], RTR
shows measured CO emissions of a representative GT26 engine seems to lie a factor 2-3 lower and scattered over a wide range
with respect to the normalised reaction time in the SEV depending on the fuel.
combustor (Equation 2). The latter has been determined from
the operation data and an appropriate model of the engine. The Fleet characteristics
SEV inlet temperature (TSEVin) and the fuel composition (only In order to illustrate the validity of the models in an overall
C2+) are directly measured, whereas the air composition, air fleet context, the assessment of the previous paragraphs is
and fuel flow are derived from design models considering completed here through a comparison of the characteristics of a
ambient conditions and various measured parameters across the representative ensemble of reasonably similar engines. The
engine. units considered for this purpose form a fleet of a specific
GT26 rating with identical turbomachinery components, yet
allowing for minor differences in some details of the combustor
parts and their manufacturing process. Paired with minor
differences in the assembly and accordingly the cooling and
leakage flows, this affects the engines’ emission and pulsation
behaviour. An engine-specific adjustment of the combustor
staging and operational parameters is therefore required, which
takes also site-specific conditions (e.g. ambient conditions, fuel
composition) and requirements (local emission regulations,
commercial operation strategies) into account.

Figure 8: Measured CO emissions vs. normalised reaction


time for a specific engine with large variation of fuel C2+
content (colours). Data points span 3.5 years (28’000 EOH)
of operation.
The data depicted in Figure 8 has been collected at
different occasions spread over a full maintenance interval of
the engine, i.e. from initial new commissioning until its first
overhaul after 28’000 equivalent operation hours (EOH). The
specific site has thereby been subject to large variations of fuel
composition. In particular, the content of higher hydrocarbons
(C2+) is varying between 0% to more than 9%mol, which
significantly affects the reactivity of the fuel. The resulting
impact on the ignition kinetics causes the CO emissions of the
engine to vary by one order of magnitude for identical
operation points. As illustrated by the graph, the engine model
is able to capture such large variations with reasonable
accuracy if an appropriate kinetic model is used. The
application of the NUIG-mechanism collapses the data from
these very different operation points and fuel composition Figure 9: Comparison of fleet data with expected characteristics
within an accuracy of RTR ±0.1, which is a reasonably accurate derived from reaction kinetics. Coloured points indicate fleet
model to be applied with confidence for predicting the engine engines at a specific fixed CO emission limit (~MEL), lines
behaviour. For a reduced reaction time near 1 a typical part load (black: NUIG, grey: GRI3.0) represent design data at a constant
emission target of 100 [mg/m3] seems achievable. The overall RTR of 1 (full lines) ± 0.1 (dotted lines).
correlation is similar to the HP test data (Figure 7). As shown in
comparison in the upper left frame of the figure, the less Figure 9 summarises the CO behaviour of the selected
accurate GRI3.0 mechanism fails to provide a distinctive GT26 fleet engines with respect to the C2+ content of the

8 Copyright © 2016 by General Electric Technology GmbH


available fuel. The lines on the graph indicate the expected fuels. At least with the reheat engines PTG applications appear
evolution of an engine’s relative load at constant RTR for to be feasible.
changing fuel gas reactivity. As demonstrated in the previous
chapters, the RTR can thereby be considered representative for
CO emissions. When fuel C2+ content is high, relative load can
be significantly reduced since the reduction of firing
temperature (and therefore oxidation kinetics) is compensated
by a faster ignition phase of the more reactive fuel components.
The black line, calculated using the NUIG model-
mechanism, sketches a physical limit of the minimum
achievable load level for a given C2+ content and CO emission
limit (i.e. at RTR=1). The grey line represents the same
characteristic calculated with the GRI3.0-mechanisms, which
would predict much lower loads to be compliant with MEL.
These predictions are compared to the final adjustment of
the engines after new commissioning or a major overhaul with Figure 10: schematic reduction of MEL with increased
replacement/reconditioning of hot gas parts. These points are reactivity fuel.
coloured with respect to the SEV inlet temperature (TSEVin),
The advantage of PTG –H2 for low load operation is
which illustrates the differences of the individual adjustment of
briefly sketched in Figure 10. The MEL can be expressed at
the units. Adjusting to higher TSEVin would reduce RTR and
fuel saved while remaining on the grid in fast load response.
allow lower loads (shift to the left). An upper limit at high TSEVin
The low load operation makes particular use produced H2 at
is given for lifetime reasons considering higher metal
operation points, which could not be achieved without the H2.
temperatures. The flashback risk at reduced reactivity is not
This should be economically rewarding if prices fluctuate and if
increased for slightly higher TSEVin. Therefore not all points in
low load parking operation for GT (allowing fast load changes)
the graph have been optimized to low part load behavior1. They
are awarded financially.
are however limited by the red / orange points (Tref+ 20/30K),
which are lying close to the predictions of the NUIG
CONCLUSIONS
mechanism. Despite the scatter2, the data indicate that this
simple combustor model based on the accurate NUIG reaction Key to the successful product development was starting
kinetics captures the overall behaviour of the fleet engines quite with fundamental research extending the validation of the
well when appropriate kinetic schemes are used. The grey line kinetic mechanism. While the GRI3.0 fails to predict the
represents the same characteristic calculated with the GRI3.0- experimental outcome the NUIG model (ARAMCO1.3) can be
mechanisms, which obviously fails to characterise the used to predict GT combustion behaviour for a wide range of
behaviour of the fleet. fuel reactivity including C2+ and H2 fuels.
Following the reasoning form the HP-tests the same The fuel flexibility for GT has been demonstrated on the
prediction can be used for H2 co-firing when replacing C2+ by example of the reheat GT using sequential combustors. In a
a C2+&H2 ordinate. comprehensive development cycle, fundamental studies on
As demonstrated in Figure 9 reheat engines can not only kinetics and the physical combustion process and validation
operate at high reactivities due to C2+; the addition of C2+ can work has been followed by experimental full scale burner
also extend the operational range of the GT in compliance with validation up to engine pressures. Further field engine
emission regulations and effectively reduces the minimum validation for C2+ fuels (where different fuels were available
environmental load (MEL). The comparison with the HP single but not freely selectable) has demonstrated the prediction
burner full scale test can be used to transfer the H2 experience capacity of the models. The missing gap in the validation could
to the GTs and allows estimating the operation for H2 be sufficiently closed by full scale high pressure tests with
containing fuels. It further allows predicting an extended varied H2 content.
operation range utilising the H2 fuel, especially for C2+ poor The fuel flexibility with respect to C2+ and H2 is therefore
demonstrated at least with respect of the reheat combustion
system to be possible with minimal output power changes. With
1
For the non-negligible number of points that are situated in the the systematic characterisation and the improved understanding
suboptimal domain (i.e. on the upper right of the characteristic line), the of the physical combustion processes the fuel flexibility of
colouring of the data suggests that these engines are commissioned to lower other GT frames can be understood and extended. This will be
TSEVin (and thus higher CO). This indicates that due to various possible reasons, useful for several applications of H2 co-combustion. The usage
less emphasis might have been put to operate the engine at lowest CO
emissions in the low part load domain. of H2 from electrolysis and PTG on power plants in co-firing
2
Other engine adjustment parameters (e.g. fuel staging, secondary air with natural gas can be combined with part load extension
flows etc) are known to influence the emission and pulsation behaviour, but are obtaining a further advantage of the H2 which could not be
omitted here since a systematic assessment is difficult to integrate in the simple obtained with less reactive fuels. This way the H2 is not only
models as applied here.

9 Copyright © 2016 by General Electric Technology GmbH


stored energy but also used with a specific advantage when
utilising it by keeping a plant running at very low loads when 5 Karl-Friedrich Ziegahn et al., “Von Kohlehalden und
electricity prices are low and availability must be retained. Wasserstoff Energiespeicher – zentrale Elemente der
A large portion of the required flexibility in the energy Energieversorgung”, Herausgeber: Deutsche
market will have to be realised by gas turbines. Combustion Bunsengesellschaft für Physikalische Chemie, 2013 Frankfurt
will continue to play a big role in the energy market for the next am Main.
decades even when integrating more and more renewables into 6 Jens Hüttenrauch ,“ Zumischung von Wasserstoff zum Erdgas
the energy mix the GT’s will serve as an intermediate storage “, Energie |/ Wasser-Praxis 2010.
medium highly efficient and flexible gas turbines are a must. 7 Gert Müller-Syring et al., Abschlussbericht "Entwicklung von
modularen Konzepten zur Erzeugung, Speicherung und
NOMENCLATURE Einspeisung von Wasserstoff und Methan ins Erdgasnetz" (G
1/07/10) Deutscher Verein des Gas- und Wasserfaches
GT gas turbine
e.V.2013.
PTG Power to gas
8 F. Güthe, J. Hellat, P. Floor: The reheat concept: The proven
EV EnVironmental (burner / combustor) pathway to ultra-low emissions and high efficiency and
SEV Sequential EnVironmental (burner /
flexibility, Journal of Engineering for Gas Turbines and
combustor)
HPT/LPT high pressure turbine / low pressure turbine
Power, 131, 021503 (2009)
9 K. Michael Düsing, Andrea Ciani, Urs Benz, Adnan Eroglu,
Thg Hot gas temperature
Klaus Knapp: Development of GT24 and GT26 (Upgrades
THPT_out Turbine outlet Temperature (HPT: High
2011) Reheat Combustors, Achieving Reduced Emissions And
(~TSEV_in), Pressure Turbine / LPT: Low Pressure
Increased Fuel Flexibility, ASME Turbo Expo 2013, San
Turbine)
Antonio, Texas, USA, GT2013-95437
tign Ignition delay time scale 10 D. Therkorn, M. Gassner, V. Lonneux, M. Zhang,
tburnout CO- oxidation time scale S. Bernero: CCPP operational flexibility extension below 30%
tres Residence time scale load using reheat burner switch-off concept, ASME Turbo
RTR reduced reaction time Expo 2015, Montréal, Canada, GT2015-42446
sl Laminar flame speed 11 M. Brower, E. Petersen, W. Metcalfe, H. Curran, N. Aluri, F.
st Turbulent flame speed Guethe, M. Füri, G. Bourque, "Ignition Delay Time and
LBO lean blow out Laminar Flame Speed Calculations for Natural Gas/Hydrogen
PSR perfectly stirred reactor Blends at Elevated Pressures", Journal of Engineering for Gas
PFR Plug flow reactor Turbines and Power, 135, 021504-1 (2013)
tPSR ext PSR extinction time 12 M. L. Brower, O. Mathieu, E. L. Petersen, N. Donohoe, A.
C2+ relative content of higher alkanes Heufer, W. K. Metcalfe, H. J. Curran, G. Bourque, F. Güthe,
C2+&H2 relative content of higher alkanes and H2 "Ignition Delay Time Experiments For Natural Gas/Hydrogen
MEL Minimum Environmental Load Blends At Elevated Pressures", ASME Turbo Expo 2013, San
Antonio, Texas, USA, GT2013-95151
REFERENCES 13 N. Donohoe, A. Heufer, W. M. Metcalfe, H. J. Curran,
M. L. Davis, O. Mathieu, D. Plichta A. Morones,
E. L. Petersen, F. Güthe: Ignition delay times, laminar flame
1 Kohler, S., Agricola, A., Seidl, H., Höflich, B. et al. – Dena
speeds, and mechanism validation for natural gas/hydrogen
Grid Study II, Integration of Renewable Energy Sources in the
blends at elevated pressure, Combustion and Flame 2014, pp.
German Power Supply System from 2015 – 2020 with an
Outlook to 2025, German Energy Agency, 2010 1432-1443.
2 Philipp Götz, Dr. Johannes Henkel, Thorsten Lenck, Dr. 14 T. Wind, F. Güthe, K. Syed, Co-Firing of Hydrogen and
Konstantin Lenz, “Negative Strompreise”, Energy Brainpool: Natural Gases in Lean Premix Conventional and Reheat
Studie Analyse, Energy Brainpool, for Agora Energiewende, Burners (Alstom GT26), ASME Turbo Expo 2014, Düsseldorf,
2014 Germany, GT2014-25813
3 Philipp Götz, Marie-Louise Heddrich, Dr. Johannes Henkel, 15 S. Ravi, A. Morones, E. L. Petersen, F. Güthe, Effects Of
Tobias Kurth, Thorsten Lenck ”Zukünftige Auswirkungen der Hydrogen Addition On The Flame Speeds Of Natural Gas
Sechs-Stunden-Regelung gemäß § 24 EEG 2014”, Kurzstudie Blends Under Uniform Turbulent Conditions, ASME Turbo
im Auftrag des Bundesverbands WindEnergie e.V, 2014 Expo 2015, Montreal, Canada, GT2015-42903
4 Nadia Grimm, Immo Zoch, Uta Weiß, 16 E. M. Burke, F. Güthe, R. F.D. Monaghan, A Comparison of
Fachbroschuere_Power_to_Gas, “Integration erneuerbaren Turbulent Flame Speed Correlations for Hydrocarbon Fuels at
Stroms in das Erdgasnetz.- Power to Gas – eine innovative Elevated Pressures, ASME Turbo Expo 2016, Seoul, South
Systemlösung für die Energieversorgung von morgen Korea, GT2016-57804
entwickeln.“, Deutsche Energie-Agentur GmbH (dena), 5/2012. 17 F. Guethe, D. Stankovic, F. Genin, K. Syed, D. Winkler,
Flue Gas Recirculation of the Alstom Sequential Gas Turbine

10 Copyright © 2016 by General Electric Technology GmbH


Combustor Tested at high Pressure, ASME Turbo Expo 2011,
Vancouver, British Columbia, Canada, Germany, GT2011-
45379
18 http://c3.nuigalway.ie/mechanisms.html
19 Gregory P. Smith, David M. Golden, Michael Frenklach,
Nigel W. Moriarty, Boris Eiteneer, Mikhail Goldenberg, C.
Thomas Bowman, Ronald K. Hanson, Soonho Song, William
C. Gardiner, Jr., Vitali V. Lissianski, and Zhiwei Qin
http://www.me.berkeley.edu/gri_mech/
20 K. Döbbeling, J. Hellat, H. Koch, 25 Years of
BBC/ABB/Alstom Lean Premix Combustion Technologies, J.
Eng. Gas Tubines Power 2005 (129).

11 Copyright © 2016 by General Electric Technology GmbH

You might also like