Professional Documents
Culture Documents
I. INTRODUCTION
AS turbines are one of the main sources for power generation in countries with natural gas resources and are
installed in numerous places in the world due to their special
characteristics. The needs of a power system analyst for a
mathematical representation of gas turbines in dynamic studies
lead to several publications in this area [1][3]. One of the most
commonly used simplified models was presented by Rowen
[1] taking into account the load-frequency and temperature
control as well as the turbines thermodynamic responses as a
linear function and inlet guide vane effects in a separate work
[2]. Several models with different degrees of simplification
for the representation of gas turbines in dynamic studies were
introduced, amongst which the IEEE model for combined cycle
power plants had deeper sight into internal processes [3]. A
good review of these models is given in [4].
From an educational point of view, there are excellent books
about dynamic studies and gas turbine performance, but few
works have been done on the model and its parameter extraction
for educational purposes [5][7]. Based on the authors experience, while many of electrical engineering students are keen in
Manuscript received October 16, 2008; revised December 10, 2008. First
published May 19, 2009; current version published July 22, 2009. Paper no.
TPWRS-00777-2008.
M. R. Bank Tavakoli and B. Vahidi are with the Department of Electrical
Engineering, Amirkabir University of Technology, Tehran 1587-54413, Iran
(e-mail: banktavakoli@aut.ac.ir; vahidi@aut.ac.ir).
W. Gawlik is with Siemens AG, Erlangen 91058, Germany (e-mail:
wolfgang.gawlik@siemens.com).
Digital Object Identifier 10.1109/TPWRS.2009.2021231
BANK TAVAKOLI et al.: EDUCATIONAL GUIDE TO EXTRACT THE PARAMETERS OF HEAVY DUTY GAS TURBINES MODEL
1367
parameters, the above efficiencies are used and assumed constant in a limited range of operational conditions where the turbine response can be approximated to be linear [1], [3]. Therefore, in the next sections operational data are first used to derive
the above efficiencies to extract turbine parameters.
III. HDGT MODEL FOR STABILITY ANALYSIS
The mathematical representation of an HDGT in dynamic
studies by Rowens model is shown in Fig. 3. There are two
major controls which are shown in this figure. These are loadfrequency and temperature control (originally, in [1] acceleration control have been shown and in [2] inlet guide vanes (IGVs)
have been modeled which are not included here).
Variable IGV (VIGV) are to regulate the air mass flow drawn
into the compressor. Actually the VIGV are in operation in CCP
applications where they regulate the air flow to maintain the
efficiency of gas turbine during operation at partial load. The
VIGV control is mainly affected by exhaust gas temperature. If
this temperature is below the reference (near to rated temperature) then VIGV will open and, in fact, it will become fully
opened in normal operation of around nominal power. Nevertheless, in partial load and start up, the exhaust temperature is
not that high and the VIGV are partly closed. In simple cycle
operation, VIGV control is only active during start up. In this
paper, we do not consider the VIGV and acceleration control as
1368
A. Turbine Parameters
and
and
(5)
(6)
Using (1) and (3), temperature after the compressor is computed as follows:
(7)
(8)
(3)
(4)
(9)
where
(pressure drop in combustor is
neglected) is the cycle pressure ratio.
and
are the cold end (compressor) and hot end (combustor, turbine) ratio of specific heats, respectively. The variable
is defined to make referencing simpler in the next equations.
where
is the specific heat of air at
Generally,
is the specific heat at constant volume.
constant pressure and
It is important to know that the specific heats and ratio vary with
BANK TAVAKOLI et al.: EDUCATIONAL GUIDE TO EXTRACT THE PARAMETERS OF HEAVY DUTY GAS TURBINES MODEL
1369
(11)
is a simplifying symbol of a temperature rise cowhere
efficient. Actually, the internal temperatures and pressure ratio
and overall response of HDGT vary with speed. These responses
are nonlinear and further complicate the models. However, in
Rowens model, a linear model is assumed with respect to speed
by applying the speed constraint of 95% to 107% of nominal
speed; see Fig. 3. We also will assume a linear response with
respect to speed deviations, but first let us assume our HDGT
at nominal speed. In this point, the per unit output torque and
mechanical power would be the same. Then
(12)
By simple mathematics, the above equation is written in the
form of the power block in Fig. 3 using (7), (8), and (11), i.e.
(13)
(14)
(15)
where and are the coefficients of output torque in Fig. 3;
and
are the air and fuel nominal flow rates; and
is the per unit output power which is equal to the p.u. torque. It
is the
should be stated here that the turbine nominal power
base of p.u. for (13)(15).
At nominal speed, the exhaust temperature can be computed
by substituting (11) and (7) in (8). Therefore
(16)
(17)
where is the coefficient of the exhaust temperature block in
is the nominal exhaust temperature of the HDGT.
Fig. 3 and
To extract the parameters, a typical operational condition is selected to derive the turbine and compressor efficiencies and then
all above parameters are computed with available data out of
unit operation.
Now, let us consider the HDGT in nominal fuel flow rate
which corresponds to 1 p.u. output torque. Then the open loop
response of the turbine with respect to speed is
(18)
Fig. 4. Pneumatic valve positioner and valve actuator. (a) Schematic view.
(b) Internal feedback mathematical model.
1370
current position of the valve and eliminates the error between set
point and position signal, see Fig. 4. Therefore, only one time
constant will appear, which is in the valve positioner block of
Fig. 3.
The positioner time constant can be found in the manufacturer
data or similar available data for older units.
Moreover, in liquid fuel systems, there is a bypass way from
the fuel pump output to the pump section. Bypass path is pre. The
sented in Rowens model by the feedback loop gain
is explicitly calculated to force the overall valve
value of
positioner-fuel system loop gain to unity [1].
and (valve positioner block
As a result, the product of
value is zero for
in Fig. 3) should become zero, i.e., the
our large HDGTs if is not. Note that this is not always the
case, especially when inner loop feedback does not exist, see
[1] for more details. Assuming linear response actuators and
valves, the fuel flow will change directly with the output signal
of the valve positioner. However, there is a lag associated with
gas/oil flow in the pipes and fuel system manifold. This lag can
be approximated by the following expression [8]:
(19)
in s is the time constant of the lag associated with
where
the container of volume
in
,
is the average pressure
,
is the steady state mass flow out of the
in Pa
is the density
container in kg/sec and
change due to pressure changes at constant temperature where
is the specific volume. For gas fuels, the lag constant is considerably higher due to higher changes in the specific volume. Knowing the rough estimates of parameters in (19)
makes it possible to obtain rough values for the lag time constant; see Section IV.
C. Time Delays and Discharge Lag
The gas turbine behavior forces its dynamic model to have
small delays and lag time constants. Actually there is a small
time delay between the fuel injection and heat release in
the combustor which is called combustion reaction delay. In
modern systems it is on the order of some ms [10]. This time
delay is implemented in Rowens model as a time delay after
the valve systems. There is also a time delay between the fuel
combustion and exhaust temperature measuring system. This
delay is caused by the exhaust system and turbine to transport
the fluid to the measuring point; see Fig. 1, and is in the order
of some 10 ms depending mainly on the size of the HDGT and
the average fluid speed. A relatively higher time lag exists in
the compressor discharge path to the turbine inlet. It can also
be approximated by (19).
D. Temperature Measurement
Temperature control in HDGTs requires measurement of the
exhaust temperatures which may be composed of thermocouple
and radiation shield [1]. Generally, there are three ways of heat
transfer between materials: conduction, convection and radiation. Here, we are only interested in the exhaust gas temperature
out of the turbine (a convective source) to control the temperature and avoid excessive heating. Nevertheless, the radiation
source, i.e., the turbine itself, will cause errors in the temperature measurement. The radiation shield is therefore used to overcome the problem. The radiation shield is a polished, highly reflective metal shield that is placed around the thermocouple and
reflects most of the radiation away from the thermocouple and
itself; see Fig. 5. This equipment will cause a lag based on its
heat transfer behavior to be presented in the model. Let us assume a simple model of the heat transfer paths of Fig. 5. The
temperature at the thermocouple tip will then be approximated
by
(20)
is the total active area for convection heat transfer
where
is the area effective for convection heat
to the shield head,
transfer to the thermocouple tip, (J/K) is the heat capacity of
the shield head and
is the convection heat transfer
coefficient [11].
Temperature measurement device is the thermocouple which
has a typical lag with a time constant based on its type and
design. Time constant of thermocouple can be easily extracted
from its time response documents.
IV. STUDY CASE: 172-MW HDGT
Operating data are always available in an existing power
plant. For land based engines, performance data are frequently
quoted at the single point ISO conditions (i.e., 1 atm ambient
and 60% relative humidity [10]). These data
pressure, 15
are given for our 172-MW turbine in Table I. The turbine is
a single shaft HDGT with natural gas as primary operating
fuel. It should be noted that in the following computations, the
pressure loss in the entrance air filters and also at the combustor
is neglected. Also, IGVs are not modeled based on the model
in Fig. 3.
Let us start with turbine model parameters using the procedure which is described in Section III. In all turbine parameters,
the values of
and are the values given in (5) and (6). In
BANK TAVAKOLI et al.: EDUCATIONAL GUIDE TO EXTRACT THE PARAMETERS OF HEAVY DUTY GAS TURBINES MODEL
1371
TABLE I
NOMINAL DATA OF HDGT SELECTED FOR MODELING
(27)
From now on, the parameters are computed for nominal opand
should be updated as
eration (Table I). Therefore,
follows:
(28)
TABLE II
TYPICAL OPERATING DATA FOR COMPUTING
TURBINE AND COMPRESSOR EFFICIENCIES
(29)
Now (14) and (15) can be used to extract the parameters of
turbine mechanical power blocks in Fig. 3 as follows:
(30)
advance, we need to know the turbine and compressor efficiencies. To do so, a typical operating point is selected and tabulated
in Table II. In advance, it is assumed that our model should be
designated to represent the HDGT at nominal point. Here we inherently assumed constant turbine and compressor efficiencies
in power output near nominal.
Turbine Efficiency: From (4), and the data in Tables I and II,
the temperature after turbine on an isentropic process
is
computed as follows:
(31)
and
(21)
(22)
where index stands for operating conditions of Table II. Note
that rated pressure ratio is decreased based on lower operating
is converted into K in (3). Turbine effimass flow. Also
ciency according to (2) is
(32)
Please note that we will need nominal flow to determine the
turbine exhaust temperature. From (17) exhaust temperature parameter is computed as follows:
(33)
(23)
Compressor Efficiency: The temperature after compressor in
isentropic process is calculated from (3)
Based on the discussion in Section III, speed sensitivity coefficients ( and ) are chosen to be
and
. The data for no load operation and minimum
fuel flow rate to maintain combustor flame are given at Table III.
Therefore
(24)
(34)
(25)
Here we need the compressor outlet air temperature which is
computed from (11)
and
(35)
(26)
Fuel System: The lag time constant of the fuel system can be
approximated if thermodynamic conditions of the fuel system
1372
TABLE III
ESTIMATED MINIMUM FUEL FLOW AND NO LOAD CONSUMPTION
TABLE VI
DATA OF RADIATION SHIELD
TABLE IV
OPERATIONAL DATA FOR FUEL SYSTEM LAG TIME ESTIMATION
TABLE V
OPERATIONAL DATA FOR COMPRESSOR DISCHARGE LAG TIME ESTIMATION
Please note that pressure and mass flow are taken from
Table I.
Temperature Measurement Blocks: The radiation shield
also has a dynamic behavior as described in (20). The data in
Table VI of the radiation shield can be utilized for modeling
purposes. The heat capacity and convection transfer coefficient
values are adopted form [11].
Let us assume convective heat transfer to the thermocouple
tip to be a cylinder with a length of 2 times the length of thermocouple tip inside shield in each side, then from (20), the following values are estimated for radiation shield in Fig. 3:
(37)
(40)
Note the nominal flow is used for the calculation, and temperature is also required to find the specific volume changes of
Methane.
The operating time of the valve positioner based on manufacturer data is 200 ms. Based on our model and assuming 5
times (which is a typical value) integrating time constants for the
signal to reach steady state values, in Fig. 3 would be 40 ms.
Time Delays and Discharge Lags: We have assumed that
there is a time of 5 ms between the fuel injection until burning
in the combustor, i.e., a combustion reaction delay of 5 ms. A
relatively conservative value of 40-ms delay for air and combustion products transfer to the temperature measuring point is
.
assumed, i.e.,
There is also a lag time constant due to compressor discharge. Using the thermodynamic properties of air and the data
in Table V, (19) is used to approximate the lag time constant
[12] as follows:
(38)
An average temperature of 1050 K is assumed trough the path
to turbine inlet [see (28) and Table II]. It should be noted that the
approximation is inevitable when using the table and graphs of
thermodynamic properties because not all the operating points
(41)
According to manufacturer information, the time response of
thermocouple is 1.7 s. Therefore, thermocouple time constant in
it block of Fig. 3 should be 1.7 s.
Finally it is good to mention here that controller parameters
in the model of Fig. 3 may be extracted from the controller set
points. The values for load frequency control and temperature
controller blocks are only assumed for the simulations in next
section. All obtained values for the parameters are in the range
of the typical values which are usually used for HDGT dynamic
models [1], [13].
V. SIMULATION
The derived and assumed parameters in previous section are
used to simulate the behavior of 172-MW HDGT. In Table VII
all parameters of model in Fig. 3 are reviewed and listed again.
The model of our HDGT is simulated against two scenarios
of 0.1% and 0.3% speed step when operating in nominal
in
conditions. The speed-governor droop of 4% (i.e.,
Table VII) is assumed for the simulation. In Fig. 6, the mechanical output power of the model against 0.1% speed deviations
BANK TAVAKOLI et al.: EDUCATIONAL GUIDE TO EXTRACT THE PARAMETERS OF HEAVY DUTY GAS TURBINES MODEL
1373
TABLE VII
PARAMETERS OF HDGTS MODEL IN FIG. 3
00.1%.
00.3%.
00.3%.
00.1%.
1374