Gas Turbine Gas Fuel Composition since there are literally an infinite number of possible composition

mixtures, and since each mixture can have various effects on the
Performance Correction Using turbine systems. Variations in the gas composition tend to affect
the mass flow through the turbine, the enthalpy of the products of
Wobbe Index combustion, thermodynamic changes in turbine component per-
formance, and the associated changes in the response of the tur-
bine control system. As such, performance correction curves for
Bryan Li gas composition can take many forms.
Gas Ttirbine Thermal Performance Engitieer Ideally, the corrections for gas composition would be based on
GE Energy Services, simply running the thermodynamic model with the test fuel and
Schenectady, NY 12306 then again with the reference fuel, and applying the ratio of those
two results to the measured performance. While this model-based
Mike J. Gross correction method is permitted by industry codes, to date, it has
Senior Technical Manager not gained widespread industry acceptance. Instead, most manu-
facturers have provided fuel composition correction curves as a
GE Energy Services, function of fuel heating value and fuel hydrogen to carbon ratio.
Moreland Hills, OH 44022
An alternate format for the fuel gas composition correction is to
express the performance changes as a function of the fuel Wobbe
Thomas P. Schmitt index. Wobbe index represents the energy density of the gas and is
Senior Technical Manager commonly used to gauge the extent to which a given gas is com-
GE Energy Services, patible with a given combustion system design. A majority of
Billerica, MA 01821 literature regarding the Wobbe index has thus focused on variation
of fuel composition and its relation to combustion stability, dy-
namics, and emissions [1,2]. As it is commonly accepted as the
most efficient and robust single index and measure of fuel gas
Gas turbine thermal performance is dependent on many external intercbangeability [3], the Wobbe index has been proposed to be
conditions, including fuel gas composition. Measured perfor- suitable as a means of correlating the perfonnance effects that
mance must be cor-rected to specified reference conditions prior to variation in fuel composition can cause. The purpose of this test
comparison against performance specifications. A performance study is validation of this application of the Wobbe index. The
correction for fuel composition is thus required. One current current method of correction using lower heating value and hy-
method of correction commonly used is to characterize fuel com- drogen to carbon ratio, as well as the method of using Wobbe
position effects as a function of heating value and hydrogen to index, are explained in this paper. Test cases are run using both
carbon ratio. This method has been used in the past within a methods and matched against a thermodynamic model to evaluate
limited range of fuel composition variation around the expected their accuracy.
composition, yielding relatively small correction factors on the
order of ±0.1%. With industry trends suggesting continued expo-
sure of gas turbines to a broader range offuels such as liquefied
natural gas and synthesized low BTU fuel, the corresponding per-
formance effects will be much larger. As a result, a more compre- 2 Lower Heating Value and Hydrogen to Carbon
hensive correction methodology is required to encompass a Ratio
broader range offuel constituents encountered. Analytical studies
Lower heating value and hydrogen to carbon ratio relate to
have been completed with the aid of thermodynamic models to
turbine performance because of their effect on fuel mass flow and
identify the extent to which the Wobbe index can be used to cor-
specific heat of the products of combustion. .The resulting correc-
relate the response of gas turbine performance parameters to fuel
tion curves based on this method are a set of three curves, each at
gas composition. Results suggest that improved performance test
a constant H/C. The graphs are shown in Figs. 1 and 2 for output
accuracy can be achieved by using the Wobbe index compared
and heat consumption, respectively. On the.curve H/C=3.75 in
with the aforementioned conventional fuel characteristics. This
both Figs. 1 and 2, there is an initial point that is based on the
proposed method remains compliant with intent of internationally
design reference condition. This design reference condition is nor-
accepted test codes such as ASME PTC-22, ASME PTC-46, and
mally specified either by contract or expected operational condi-
ISO 2314. [DOI: 10.1115/1.4003453]
tion. Common sense would thus dictate that small variations are
expected from the design point and, consequently, small correc-
tions. The remaining points that make the curves are generated at
a constant H/C above and below the design H/C (4 and 3.56 in
1 Introduction this case), based off the design reference composition.
The primary purpose of correction curves for a performance Upon receiving the fuel analysis of samples taken in the field,
test is to make the corrected test results independent of the bound- LHV and H/C values can be determined. To obtain the correction
ary conditions that persisted at the time of the test. As such, power factor for fuel composition, interpolation along the constant H/C
plant suppliers and customers work together using thermodynamic curves with the test LHV is required. Another interpolation is then
models of the equipment and systems; the resultant model and done between the constant H/C curves at the test H/C. This will
associated correction curve serves all parties and cover a reason- allow the determination of the correction factor to be applied to
ably broad range of possible boundary conditions. Boundary vari- measured performance parameters.
ables such as inlet air temperature and humidity are easily char- Unusual gas compositions, high variation of fuel compositions,
acterized in the thermodynamic models and simply accounted for and high variation from design conditions are becoming more
via conventional correction curves that have been well established commonly experienced. This current method of correction may
in the industry. Fuel gas composition is not as straightforward not be suitable for such situations. Temporary solutions may in-
clude building curves representative of the fuel that is actually
consumed during testing or adding additional points to the exist-
Contributed by the Power Division of ASME for publication in the JOURNAL OF ing correction curve. Rather, an updated correction methodology
ENGINEERING POR GAS TURBINES AND POWER. Manuscript received July 16, 2010; final
manuscript received November 8. 2010; published online May 2, 2011. Editor: Dilip can ensure that the processes with greatest accuracy are utilized
R. Ballal. for thermal perfonnance testing.

Journal of Engineering for Gas Turbines and Power OCTOBER 2011, Vol. 133 / 104501-1
Copyright © 2011 by ASME
 EfliQct of Gas Fuel Composition on Output (LHV-H/C) Effect of Gas Fuel Composition on Output (WI)

1.012
HÄ:=4.00: y= 1.438E-10)('-8.499E-06x+ 1.115

H/C=3.75: y=1.462E-10)f'-8.543E-06x+1.112
1.00S
hrc=3.66: y = 1.451E-10x' - 8.530E-0ex + 1.109
1.004
I 1.004
y=-1.262E-03x+1.067
I
S 1 •S
'HA0=4
o 0.996
0.996
3.75
^ " * ^ HC=3.56
0.992
46 48 50 52 54 56 58
17600 16500 1950C 20500 21500
Wobbe Index, MJ/Nm3
LHV, BTU/lb

Fig. 1 Example fuei gas composition correction curve for out- Fig. 3 Example fuei gas composition correction curve for out-
put based on LHV and H/C ratio P"» "ased on Wobbe index

can then be calculated; the resulting curve for output is plotted in

3 Wobbe Index
The Wobbe index is an indicator of the quality of a fuel gas and
is mathematically described as
HHV
WI = -
It is expressed in this paper in SI units (MJ/N m^), but can also be
expressed in U.S. units (BTU/Sft^). The definition of Eq. (1) says
that for any given orifice size, all gas mixtures with the same
, Wobbe index will deliver the same amount of heat [4]. It is highly
applicable in industrial situations where there is high variation in
fuel composition, such as multiple fuel sources or liquefied natural
gas. An example of its application would be a combustion control
system that modulates fuel flow depending on Wobbe index to
protect the system that may be sensitive to fuel quality [5]. Wobbe
index can be measured real-time using a Wobbe index meter/
analyzer. There are multiple measurement techniques and ap-
proaches to obtain this number, each with their advantages and
disadvantages. But in general, it will consist of a continuous calo-
rimeter coupled with a device that measures the density of the
fuel.
Generating correction curves based on Wobbe index is a con-
tinuation of the correction curves built with the LHV-H/C method.
The fuel composition points that were used to create the LHV-H/C
correction curves are used, in panicular, their fuel constituents and
correction factors. Extra calculations are required to determine the
HHV and SGfuei based on the fuel constituents. The Wobbe index
Effect of Gas Fuel Composition on Heat Consumption (LHV-HC)
1.012
H/C=4.00: y= 1.021E-10x'-6.014E-06x +1.081
HTC =3.75: y = 1.032E- lOx' - 6.048E-06X + 1.079
I 1.008
HC=3.56: y= 1.032E-10x'-6.039E-06x +1.077
I, "^.„..„^^^
MC=3.5e
19500 20500
LHV, BTU/lb
Fig. 2 Example fuel gas composition correction curve for heat
consumption based on LHV and H/C ratio
Fig. 3 (a similar curve is created for heat consumption). The
points that make up the three H/C curves are now collapsed into a
single line that is quite linear. A single interpolation is done at the
test Wobbe index to determine the correction factor.
(1) 4 Test Methodology
This analytical study was done on the basis of a gas turbine
simple cycle, using a thermodynamic model that represents a typi-
cal GE PG9371 gas turbine. As previously mentioned, measured
performance has to be corrected to specified reference conditions.
The ambient reference conditions used in this study were standard
reference conditions. The reference fuel compositions used were
CH4 85.3%, CzHg 7.29%, C3H8 1.73%, C4H,o(i) 0.66%, C5H|2(i)
0.19%, N2 4.59%, and CO2 0.24%.
Two sets of correction curves were generated for this study
using the reference conditions—one set using the Wobbe index
and the other using the LHV-H/C method. Each set consisted of
an output curve and a heat consumption curve. The value on the
y-axis of these graphs is the correction factor; this value is applied
to performance parameters to determine corrected performance.
Any error realized here is directly proponional to the performance
parameter to which it is applied. In this study, output and heat
consumption were evaluated. All correction curves and calculated
data penain only to the PG9371 gas turbine model at the specified
reference conditions. A similar correction curve result would be
expected from other gas turbine models.
To establish the feasibility of using the Wobbe index as a cor-
rection parameter it is necessary to (i) show that the error is re-
duced and, (ii) determine the regions where the highest bias error
will occur. An initial random set of 12 fuel gas compositions was
selected that comprised of a H/C range 3.65-3.95, LHV 19,100-
19,900 BTU/lb, and Wobbe index 50.5-53.5 MJ/N m\ A sys-
tematic investigation of the specific range of different fuel con-
stituents was performed, specifically methane, ethane, nitrogen,
and carbon dioxide. Methane and ethane are normally the greatest
in proponion and have the highest heating values of common
hydrocarbons in natural gas fuel mixtures. Nitrogen and carbon
• WC=4
dioxide are common inens in fuel gas and are normally present in
small amounts; however, they have a significant adverse affect on
the heating value of a fuel. By investigating these specific con-
3.75
stituents' ranges, a wider variation of fuel compositions was used
to test the Wobbe index method, and their resulting effects were
explored. The ranges chosen for these constituents were as fol-
21500
lows: CH4 85.30-98.00%, C2H6 0-7.29%, N2 0-4.83%, and CO2
0-4.00%. The values that are within these ranges reflect what a
normal gas turbine operating on natural gas would experience.
Other higher hydrocarbons present in smaller amounts were also
varied as a result of the way test cases were chosen. As a constitu-

104501-2 / Vol. 133, OCTOBER 2011 Transactions of the ASME

 Table 1 Error (%) from correction curves using Wobbe index and H/C methods (576 test
cases)
Minimum
Output Wl 0.001
LHV-H/C 0.000
Heat consumption WI 0.000
LHV-H/C 0.000
ent range was explored, all other constituents would change by an
equally divided amount to keep the total composition constant.
Each test case in this study had a unique fuel composition. It
simulated a potential fuel analysis that came back from the labo-
ratory in which thermal performance results were to be computed
from. A total of 576 test cases were ran: These test cases covered
a range of LHV 18,000-21,370 BTU/lb, H/G 3.55-4.00, and WI
47.5-57.2 MJ/N m-'. This range represents a maximum deviation
of 9% from the design LHV value.
An example of the process to compare the accuracy of the two
methods is presented below. This example was demonstrated for
output only, since the same process is applicable to heat consump-
tion and can be done easily.
Example. Using fuel compositions of GH4 86.47%, G2H6
6.00%, G3H8 1.73%, G4H,o(i) 0.50%, G5H|2(i) 0.30%, N2 2.00%,
and GO2 3.00%, the following characteristics can be evaluated:
LHV=19,147 BTU/lb, H/G=3.773, and WI=51.47 MJ/N m l
These values are applied to the correction curve for output based
on LHV-H/G (Fig. 1). Interpolation at the test LHV using Eqs.
(2)-(4) gives the conesponding correction factors for the curves
H/G=4.00, H/G=3.75, and H/G=3.56, respectively.
GF= 14,375 X 10-'" (LHV)2-8.4986 X 10"* (LHV) + 1.1149
(2)
GF=14,520X 10-'" (LHV)2-8.5431 X 10"* (LHV)-H 1.1118
(3)
G F = 14,513 X 10-'" (LHV)2-8.5304 X 10-^ (LHV)-1-1.1088
(4)
The correction factors (for curves H/G=4.00, H/G=3.75, and
H/G = 3.56) are 1.00485, 1.00146, and 0.99870, respectively.
Next, interpolation at the test H/G between the curves gives the
final correction factor of 1.00177 for the LHV-H/G method.
The correction factor from the Wobbe index method is then
determined. By applying the test Wobbe index to the correction
curve for output based on Wobbe index (Fig. 3), a single interpo-
lation using Eq. (5) gives the final correction factor of 1.00156.
This method requires fewer interpolations and is overall simpler.
GF = - 1.2620 X 1.0665 (5)
To make a comparison of the accuracies, the correction factor
from each method is compared with a "measured" correction fac-
tor. This measured value is taken as the true value and is deter-
mined from a thermodynamic model prediction. Here, enor is
defined as the difference between the measured value and correc-
tion curve value. By comparing the error from each method, one
test case can be evaluated. In the example test case, the measured
correction factor was 1.00020. The resulting errors for the LHV-
H/G method and the WI method were 0.157% and 0.137%, re-
spectively.
5 Results
Gorrection factors were compiled for all 576 test cases, as was
their enor relative to the measured value for each case. Table 1
shows the comparison of the minimum, maximum, mean, and
Journal of Engineering for Gas Turbines and Power
Maximum Mean Standard deviation
0.185 0.098 0.050
0.251 0.104 0.070
0.051 0.039 0.016
0.112 0.044 0.030
standard deviation of enor between the LHV-H/G and Wobbe in-
dex methods. The maximum error for output was reduced by more
than a quarter from 0.25% to 0.18%, while the mean enor was
slightly reduced from 0.104% to 0.097%. Similar results were
seen with heat consumption. Overall, it was shown that using the
Wobbe index method reduced the maximum enor, mean enor, and
standard deviation.
Enor due to interpolation was quantified to find out whether it
accounted for a significant portion of the error when utilizing the
LHV-H/G method. Refening to the conection curves in Figs. 1
and 2, the conection factors for points on the H/C = 3.75 curve
were determined in two ways. It was taken from the measured
value when the conection curve was generated and also deter-
mined by interpolation. The resulting enor was approximately
0.007% for output and 0.005% for heat consumption. It was con-
cluded that it was not a significant factor in the LHV-H/G method
having a greater enor.
During analysis of the 576 test cases, a clear conelation was
found between the enor and GO2 content utilizing both methods.
High GO2 content of the fuel was observed to drive a higher enor.
This observation was further examined by creating new conection
curves utilizing Wobbe index as a bivariate with GO2, shown in
Fig. 4 for output (a similar curve was created for heat consump-
tion). GO2 is held at a constant percentage when generating each
curve. Each curve represents a 1% increment in the amount of
GO2, starting from 0% to 4%. The same 576 test cases were rerun
using this updated conection curve.
Table 2 shows the results from the Wobbe index conection
curves with lines of constant GO2 (WI-GO2). For comparison,
results are also displayed in the same table from the single Wobbe
index curve and LHV-H/G methods. There was substantial im-
provement in maximum enor, mean enor, and standard deviation.
By adding the GO2 component to the Wobbe index method, maxi-
mum and mean enors for output were reduced from 0.18% to
0.05% and from 0.098% to 0.034%, respectively. Overall, from
the LHV-H/G method to the WI-GO2 method, maximum enor was
Effect of Gas Fuel Composition on Output (WI, Lines of Constant
CO2)
49 51 S3
Wobbo Indax,
Fig, 4 Fuei gas composition correction curve for output based
on Wobbe index witii iines of constant COj
OCTOBER 2011, Vol. 133 / 104501-3
 Table 2 Error (%) from correction curves using LHV-H/C curve, singie WI curve, and WI curve
with iines of constant COj (576 test cases)

Minimum Maximum Mean Standard deviation

Output LHV-H/C 0.000 0.251 0.104 0.070

WI 0.001 0.185 0.098 0.050
WI-CO2 0.000 0.054 0.034 0.012

Heat consumption LHV-H/C 0.000 0.112 0.044 0.030

WI 0.000 0.051 0.039 0.016
WI-CO2 0.000 0.032 0.019 0.006

reduced from 0.25% to 0.05% and mean error was reduced from validate this method on expanded test compositions as well as
0.10% to 0.03%. Similar improvement was seen when applied to other fuel types including low BTU syngas, biofuels, and liquefied
heat consumption. natural gas.
To confirm the significance of these findings, statistical analysis
was done between the LHV-H/C and WI-CO2 methods. Based on Nomenclature
a two-sample t-test and paired t-test, the probability at 95% con- CF = correction factor
fidence interval showed the WI-CO2 yielding a statistically sig- CO2 = carbon dioxide
nificant lower error. CH4 = methane
C2H6 = ethane
C3H8 = propane
C4H10 (i) = isobutane
6 Conclusion
C5H12 (i) = pentane
Significant findings were realized in this analytical study. Fuel H/C = hydrogen to carbon ratio
composition correction curves based on Wobbe index showed de- HHV = higher heating value
creased mean error and maximum error as well as lower standard LHV = lower heating value
deviation compared with the correction curves based on lower MS9001FB = heavy-duty gas turbine manufactured by GE
heating value and hydrogen to carbon ratio. The Wobbe index that operates at 50 Hz frequency
method also simplified the process of determining the final cor- N2 = nitrogen
rection factor by collapsing all data into one curve. This reduced PG9371 = designation used by GE to define the model
the amount of interpolation required to calculate the correction type of gas turbine, and is a type of
factor. The error introduced by interpolation, however, did not MS9001FB
present a significant barrier to using multiple curves. SG = specific gravity
Additional study was done on the Wobbe index method based WI = Wobbe index
on the observation that error increased with higher CO2 content of Wobbe index with lines of constant CO2
WI-CO2 =
the fuel. By utilizing a Wobbe index curve with multiple lines of
constant CO2, the error was further reduced by a significant References
amount. [1] Kurz, R., and Kaiser, R., 2004, "On Fuel Stability for Gas Turbines," Proceed-
The methodologies tested in this study would be useful in up- ings of the 33rd Turbomachinery Symposium, Houston, TX.
dating current practices of fuel composition correction. Utilizing [2] Sträub, D., Ferguson, D., Casleton, K., and Richards, G., 2007, "Effects of
Propane/Natural Gas Blended Fuels on Gas Turbine Pollutant Emissions,"
the Wobbe index with CO2 lines would undoubtedly produce the Fifth U.S. Combustion Meeting, San Diego, CA.
most accurate results, but at the cost of working with more com- [3] Klassen, M., 2005, "White Paper on Natural Gas Interchangeability and Non-
plicated graphs with the need to interpolate results. Whether this Combustion'End Use," NHC-f Interchangeability Work Group, Section C.3,
method or the single Wobbe index curve is used, both have been Power Generation.
[4] AGA Bulletin No. 36.
shown to produce results with a lower error. [5] Vandervort, C , 2000, "Combustion Dynamics Control for Variable Fuel Gas
This study is applicable to a relatively normal range of fuel Composition and Temperature Based on Gas Control Valve Feedback," U.S.
constituents for natural gas. Future objectives would be to test and Patent No. 6082092.

104501-4 / Vol. 133, OOTOBER 2011 Transactions of the ASME

Copyright of Journal of Engineering for Gas Turbines & Power is the property of American Society of
Mechanical Engineers and its content may not be copied or emailed to multiple sites or posted to a listserv
without the copyright holder's express written permission. However, users may print, download, or email
articles for individual use.