Вы находитесь на странице: 1из 180

MODELING AND CONTROL OF A SUPERCRITICAL

COAL FIRED BOILER


by
ROHIT DAMODAR PARANJAPE, B.Chem.Engg., M.S.Ch.E
A DISSERTATION
IN
CHEMICAL ENGINEERING
Submitted to the Graduate Faculty
of Texas Tech University in
Partial Fulfillment of
the Requirements for
the Degree of
DOCTOR OF PHILOSOPHY
AnnrnvpH

May, 1996

^ 0 5 (J?

/ll^i
fr^

Copyright 1996, Rohit Paranjape

^/.

'P^

ACKNOWLEDGEMENTS

I dedicate this dissertation to my mother and to the loving memory of my father. I


would not have achieved anything without their love, support and guidance throughout
my career and life. I owe special thanks to my wife Mrinal for her love, encouragement
and all kinds of support throughout my Ph.D. work. I am thankful to my brother, sisterin-law, my niece Anuya, father-in-law, mother-in-law and Amit for their love and
encouragement. I am truly grateful to Uncle and Aunty for being my parents in the U.S.A.
I am deeply indebted to Dr. J. B. Riggs, the chairman of my committee, for his
guidance, encouragement, support and everything else that made this research possible. I
extend my sincere thanks to the other committee members. Dr. R. R. Rhinehart, Dr. R. W.
Tock, and Dr. . . Anderson for their invaluable suggestions and assistance in this work.
I owe special thanks to my "Boiler Gurus" Tom Calle and Danny Crow from Texas
Utilities Electric, Dallas and Dennis Buchanan from Southwestern Public Company for
their unconditional and timely assistance in this work.
And last but not the least, my special thanks are extended to my co-workers in the
research group and all myfriendsin the Chemical Engineering Department for their vital
advice and moral support during this work.

J'
TABLE OF CONTENTS

ACKNOWLEDGEMENTS

ii

ABSTRACT

iv

LIST OF TABLES

vi

LISTOFHGURES

vii

CHAPTER
I

INTRODUCTION

LITERATURE SURVEY

MODELING OF A SUPERCRITICAL
COAL FIRED BOILER

16

IV

COORDINATED CONTROL SYSTEM

43

NONLINEAR PROCESS MODEL


BASED CONTROL

59

VI

CONTROL STUDIES

69

Vn

CONCLUSIONS AND RECOMMENDATIONS

161

BIBLIOGRAPHY

163

t,-.

A * *.'^#>^-^) :

ABSTRACT

A boiler or a steam generating unit is an integral part of any electric utility plant. It
requires a source of heat at a sufficient temperature level to produce steam. The fossil
fuels are burnt in the furnace of the boiler for this purpose. The type of steam generating
unit considered in this project was a coalfiredunit.
An advantage of using supercritical conditions is reduction in physical size of the
boiler and steam piping for the same heat carrying capacity and a greater overall efficiency
of the unit.
The steam is superheated and reheated in order to generate electric power with a
turbogenerator. The heat of superheat is all recoverable in the turbine. A variation in the
steam temperature and pressure, etc., may cause an unequal expansion and contractions in
the turbine parts. Rapid and excessive changes can result in damage to the turbine.
Moreover, such variations also cause a change in the unit electric generation. In the
electric utility plants, the objectives are to produce required units of electricity
continuously and make load changes as and when required and as quickly as possible.
Thus, control of steam temperature, pressure, etc., or control of boiler-turbine system is a
very important and challenging problem in electric utility plants.
A dynamic simulator for the supercritical coal fired unit was developed as a first
part of this project. It was based on the Cromby 2 model developed by McDonald,
Kwatny and Spare of the Philadelphia Electric Company.

This simulator was bench-

marked against plant data provided by Texas Utilities Electric on a steady state basis.
In the control studies on this supercritical unit, a coordinated control system with
all the loops having conventional controllers was developed. It was thought that some of
the loops in the coordinated control system might be nonlinear; thus it might be

IV

advantageous to use nonlinear models for control purposes in those loops. From the
literature survey, it was found that no such attempt was made previously to identify
nonlinear loops and apply nonlinear models for control purposes in a coordinated control
system. Thus, in this research work, the nonlinear loops were identified and nonlinear
models were developed for control purposes, for the first time.

The coordinated

controller having these nonlinear models was in general found to perform better or at least
as good as the coordinated controller having all the conventional controller loops for
steam temperature and throttle pressure control.

flWt

Ksv

LIST OF TABLES

3.1

Data receivedfromTU Electric

39

3.2

The fuel specification receivedfromTU Electric

40

3.3

Approximate flue gas temperature

40

3.4

Comparison of the values calculated by the simulator with


the plant data

41

3.5

Values of the parameters used in the simulator

42

4.1

Tuning parameters for the Coordinated Control System

57

4.2

Gain Comparison 1

58

4.3

Gain Comparison 2

58

5.1

Tuning parameters for GMC controller for superheater spray

68

5.2

Tuning parameters for GMC controller for


firing rate to feed water ratio
Measure of goodness of control

68
160

6.1

VI

i^itir^MKiat

LIST OF HGURES
3.1

Process Schematic for the feed water-steam side


of the boiler

37

3.2

A typical flue gas path

38

4.1

Boiler following mode

51

4.2

Turbine following mode

52

4.3

Boiler-Turbine coordinated control

53

4.4

Block diagram of Boiler-Turbine valve-Turbine

54

4.5

Turbine master controller

54

4.6

Boiler master controller

55

4.7

Superheat spray flow controller

56

4.8

Reheat temperature controller

56

5.1

Superheater Pendent Section

67

6.1

Response ofpower output for test 1

82

6.2

Response of throttle pressure for test 1

83

6.3

Response of superheated steam temperature for test 1

84

6.4

Response of reheated steam temperature for test 1

85

6.5

Superheat spray for test 1

86

6.6

Fuelfortest 1

87

6.7

Feed water for test 1

88

6.8

Response of power output for test 2

89

6.9

Response of throttle pressure for test 2

90

6.10

Response of superheated steam temperature for test 2

91

6.11

Response of reheated steam temperature for test 2

92

6.12

Superheat spray for test 2

93

vu

^S^^MBltv

6.13

Fuel for test 2

94

6.14

Feed water for test 2

95

6.15

Response of power output for test 3

96

6.16

Response of throttle pressure for test 3

97

6.17

Response of superheated steam temperature for test 3

98

6.18

Response ofreheated steam temperature for test 3

99

6.19

Superheat spray for test 3

100

6.20

Fuel for test 3

101

6.21

Feed water for test 3

102

6.22

Response ofpower output for test 4

103

6.23

Response of throttle pressure for test 4

104

6.24

Response of superheated steam temperature for test 4

105

6.25

Response ofreheated steam temperature for test 4

106

6.26

Superheat spray for test 4

107

6.27

Fuel for test 4

108

6.28

Feed water for test 4

109

6.29

Response of power output for test 5

110

6.30

Response of throttle pressure for test 5

Ill

6.31

Response of superheated steam temperature for test 5

112

6.32

Response of reheated steam temperature for test 5

113

6.33

Superheat spray for test 5

114

6.34

Fuel for test 5

115

6.35

Feed water for test 5

116

6.36

Response ofpower output for test 6

117

6.37

Response of throttle pressure for test 6

118

vm

'-3S

6.38

Response of superheated steam temperature for test 6

119

6.39

Response of reheated steam temperature for test 6

120

6.40

Superheat spray for test 6

121

6.41

Fuel for test 6

122

6.42

Feed water for test 6

123

6.43

Response ofpower output for test 7

124

6.44

Response of throttle pressure for test 7

125

6.45

Response of superheated steam temperature for test 7

126

6.46

Response of reheated steam temperature for test 7

127

6.47

Superheat spray for test 7

128

6.48

Fuel for test 7

129

6.49

Feed water for test 7

130

6.50

Response of power output for test 8

131

6.51

Response of throttle pressure for test 8

132

6.52

Response of superheated steam temperature for test 8

133

6.53

Response of reheated steam temperature for test 8

134

6.54

Superheat spray for test 8

135

6.55

Fuel for test 8

136

6.56

Feed water for test 8

137

6.57

Response of power output for test 9

138

6.58

Response of throttle pressure for test 9

139

6.59

Response of superheated steam temperature for test 9

140

6.60

Response ofreheated steam temperature for test 9

141

6.61

Superheat spray for test 9

142

6.62

Fuel for test 9

143

IX

6.63

Feed water for test 9

144

6.64

Response ofpower output for test 10

145

6.65

Response of throttle pressure for test 10

146

6.66

Response of superheated steam temperature for test 10

147

6.67

Response of reheated steam temperature for test 10

148

6.68

Superheat spray for test 10

149

6.69

Fuelfortest 10

150

6.70

Feed water for test 10

151

6.71

Response of power output for test 11

152

6.72

Response of throttle pressure for test 11

153

6.73

Response of superheated steam temperature for test 11

154

6.74

Response of reheated steam temperature for test 11

155

6.75

Superheat spray for test 11

156

6.76

Fuelfortest 11

157

6.77

Feed water for test 11

158

6.78

A typical ATV test

159

CHAPTER I
INTRODUCTION

A boiler or a steam generating unit is an integral part of any electric utility plant. It
requires a source of heat at a sufficient temperature level to produce steam. Fossil fuel
utilized for the generation of steam is generally burned directly for this purpose in the
furnace of the boiler. The fossil fuels can be coal, oil, gas, wood, and their derivatives.
The type of steam generating unit considered in this project is a coalfiredunit.
This boiler can be operated at subcritical or supercritical conditions. Modem
utility boilers operate at pressures in the range of 2000 to 4000 psig, while their industrial
counterparts are generally in the range of 100 to 1000 psig. Advantages of using high
pressure are the reduction in physical size of the boiler and steam piping for the same heat
carrying capacity and the improvement in the overall efficiency of the unit.
In generating electric power with a turbogenerator, it is much more efficient to use
steam that has been superheated and reheated as is done in the typical electric utility plant.
The general practice with the industrial boiler is to use saturated steam or only a small
amount of superheat unless the electric power is being generated in the industrial plant. A
turbine generally transforms the heat of superheat into work without forming moisture.
The heat of superheat is all recoverable in the turbine (Dukelow, 1991). A variation in the
steam temperature, pressure, etc., may cause unequal expansion and contraction in the
turbine parts. Rapid and excessive changes in temperature can result in damage to the
turbine. Steam temperatures that are significantly higher than the design temperature can
shorten the life of the turbine metal parts. Such temperature variations also cause a
change in the unit electrical generation.

In electric utility plants, the objective is to

produce required units of electricity continuously; make load changes as and when

required and as quickly as possible. Thus, control of steam temperature, pressure etc. or
control of the boiler-turbine system is a very important and challenging problem in Electric
utility plants.
A supercritical steam generating unit is the one which operates at a pressure above
the critical pressure of 3208 psia. When water at a supercritical pressure is heated, it does
not boil and does not produce a two-phase mixture of water and steam. Instead, the fluid
undergoes a transition in the enthalpy range of approximately 850 to 1050 btu/lb, during
which the physical characteristics (such as density, compressibility, viscosity, etc.) change
continuously and homogeneously from those of a liquid to those of a gas or a vapor. The
nature of the process rules out the use of a boiler drum to separate steam from water and
drumless steam generators have been universally adopted for supercritical units. At the
boiler's inlet the high pressure feed water is forced into the boiler tubes. It is heated as it
passes through them and finally is ejected from the boiler's main outlet (secondary
superheater outlet) as a main steam.
A Combustion Engineering (CE) supercritical unit at Texas Utilities (TU) Electric,
Dallas was chosen for this simulation and control study. The unit under consideration was
designed to deliver superheated steam at a rate of 4003.678 Klbs/hr (maximum
continuous) at 1005^ and 3850 psig (superheat outlet) to a 575 MW turt>ogenerator.
The reheater was designed to handle 3465 Klbs/hr of steam reheated from 5 6 9 ^ to
IOO5OF.
A dynamic simulator was developed as afirstpart of this project. It was based on
the Cromby 2 model developed by McDonald, Kwatny and Spare of the Philadelphia
Electric Company (McDonald et al., 1971). The Cromby 2 model was modified for the
considered supercritical unit. The Cromby 2 model was selected because it covered the
major elements of plants that are of interest, and it had been validated from 50 to 100%

load. The model is well documented and is available in the public domain (Riggs et al.,
1995).
The steady state data at different operating loads was provided by TU Electric,
Dallas from one of their supercritical units. That plant data was provided as the initial
conditions for the dynamic simulator.

The plant data provided include temperature,

pressure of the steam at various locations in the unit, amount of fiiel, fuel heating value,
steam flow rate, spray flow rate, flow rate, temperature and pressure of the steam at the
various stages within the turbine, megawatt output, etc. Then the dynamic simulator was
bench marked against the actual data on a steady state basis.
Control requirements of these supercritical units are quite different from those of a
drum type unit. The interactions between steam generator inputs (feed water, fiiel, air),
output (steam flow, pressure, temperature) and turbine-generator output (Kilowatts) are
more direct and immediate in the supercritical pressure unit than in the subcritical drum
type unit. The changes in the electrical load on the turbine-generator are felt more quickly
(in the form of pressure and temperature changes) throughout the supercritical unit since
there is less heat storage than in a drum type unit.
Since the supercritical unit has differences in response to the input and the output
variables, a different approach is necessary in the operation of the unit.

A different

approach is also required in the design of a control system to adequately control the unit
and to take advantage of its inherent greater responsiveness.
A control system should coordinate all the inputs and outputs directly as a function
of the desired unit output load.

A demand signal should be applied directly and

simultaneously to all the manipulated variables, to produce the most rapid, stable
response.

Thus, the next part of the project was to develop a coordinated control system for
this supercritical unit. Initially, only the conventional controllers were used in all the
loops. A coordinated control system has 10-15 different loops. All these loops make use
of one or more of the control principles, e.g., feedback control, feedforward control, ratio
control, cascade control, etc. It is a complicated control problem. It was thought that one
or more of these loops might be nonlinear. Hence, nonlinear models could be used for
control purposes in these loops. From the thorough literature survey, it was found that no
attempt was made previously to identify the nonlinear loops in a coordinated control
system and to apply nonlinear models for control purposes in those loops. The gains were
calculated for each of the loops within the coordinated control system by giving positive
and negative setpoint changes. Two loops involving the control of the superheated steam
temperature were found to be nonlinear. The nonlinear models were developed for these
two loops in the coordinated control system. This was the first attempt in an open
literature to test the nonlinear models in the coordinated control system. The performance
of the coordinated control system having all the conventional controllers was compared
with the coordinated controller having nonlinear models in two loops.
Development of the process simulator, development of the coordinated control
strategy, nonlinear controller models in the coordinated control strategy, control results
and comparisons are discussed in the following chapters.

ie^iw^VAwfa^\^HiJG^ft

CHAPTER n
LITERATURE SURVEY

Babcock and Wilcox (1972) have discussed the fundamentals of steam, generation
and use in the 39^^ edition of SteamIts Generation and Use. They have described the
application of the fundamentals of steam generation to the design of boiler, superheaters,
etc., to generate steam using chemical or fossil fiiel. After fiiel selection and determination
of the heat requirement, the fiimace is designed for proper combustion of fiiel with
appropriate provisions for handling the ash in the case of solid fiiels. Furnace waterwall
surfaces are proportioned to reduce combustion gas temperature to the level desired at the
entrance to the convection banks. The convection surfaces in the boiler, superheater and
reheater are then designed. The variation of steam temperature with load is discussed for
superheaters in various locations. The need for steam temperature adjustment and control
is explained and the means of accomplishment are set forth, including consideration of a
number of different methods of control such as attemperation, burner selection, excess air,
divided furnace differentially fired, etc.
Then they deal with fossil-fiiel equipment for the electric utilities. The principal
factors involved in the selection of steam generating equipment are listed and examined.
These include fuels, steam requirements, space, and geographical considerations, power
for driving auxiliaries and guarantees. The B & W radiant and universal pressure boiler
types for pulverized coal, cyclone fiimace, natural gas and oil firing are described. Thus,
SteamIts Generation and Use gives an excellent overview about steam generation, fiiels,
steam generating equipments, application of steam generation, etc.

Boiler Modeling Literature


Suzuki et al. (1979) have constmcted a dynamic mathematical model of a
supercritical once-through boiler presently operating at Kainan Thermal Power Station of
Kansai Electric Power Company. The model is based on the nonlinear partial differential
equations which describe the physical phenomena in the boiler. The dynamics of the boiler
can be derived from the three nonlinear partial differential equations for continuity,
enthalpy and pressure (conservation of mass, energy and momentum, respectively). The
equation of pressure drop is static.
Although a number of thin tubes are arranged in parallel in the actual boiler,
Suzuki et al. (1979) have considered them lumped into a single equivalent tube with the
same total cross-sectional area. Each tube in the series represents one section of the
boiler. The length and cross-sectional area of each tube is appropriate to the section of
the boiler it represents and then, equations for continuity, enthalpy and pressure are
written for each section of the boiler.
The equations of the continuity and enthalpy are first solved simultaneously by
utilizing the piecewise-linear relationships for enthalpy and specific volume and then the
pressure equation is solved. In the model, the feed water flow depends on both the
govemor valve's position and on the steam flow to the feed water pump turbine.
Because of the assumptions necessary for solving the equations and the
discrepancy between the design specifications and real values, mathematical models which
are derived analytically generally cannot describe actual behavior.

Therefore, this

analytical model was fitted to the actual system by adjusting parameters. The values of
these unknown parameters were determined by adjusting them so as to minimize the sum
of the squares of the differences between the actual boiler output and the simulated
results. The simulation was performed using Euler integration with time increments of

'I

">w^^^a

one second. The Fletcher-Reeves optimization method (Edgar et al., 1988) was used for
parameter adjustment.
Suzuki et al. (1979) have also shown some simulation results. The model has
three manipulated variables namely govemor valve position, fiiel flow and steam flow to
the feed water pump turbine, and three controlled variables namely main steam flow,
pressure and temperature. Suzuki et al. (1979) have shown the simulated open loop
responses of the controlled variables to the step changes in each of the manipulated
variables.
Volkov et. al (1992) have presented a physical-mathematical predictive model of
combustion in the firing of polydisperse pulverized coal in the fiimace chamber of utility
boilers, which includes a specific use of an inertial-diffiision model for description of
particle dynamics and a technique for taking into account fiiel polydispersity. Volkov et
al. (1992) claim that this model is applicable to large-scale plants in a wide range of
variation of geometrical parameters and operating conditions.
Bauer et al. (1989) have developed a mathematical model for a 320 megawatt, wet
bottom, coal fired utility boiler to determine boiler performance and considerable design
information at any load. The model which consists of many stand-alone submodels can
predict the boiler performance parameters such as fiimace exit gas temperature, steam
temperature, spray flows, combustibles in ash, boiler efficiency, air preheater outlet gas
temperature, tube metal temperatures, gas flows and resident times, flame temperature.
Bauer et al. (1989) also describe an algorithm that can be utilized for on-line monitoring of
boiler performance and to offer advice to boiler operators on a real time basis.
Chawdhry (1993) has described the application of system identification techniques
to obtain models of boilers in power stations. He has used the method of recursive least
squares (RLS) to estimate parameters. Before proceeding to test on a real plant, a

detailed boiler model was simulated. The nonlinear analytical model used in the simulation
was developed by McDonald et al. to represent a boiler at the Cromby power plant of
Philadelphia Electric Co. The simulation program represented the plant during steady
state conditions. (Chawdhry [1993] has added local PI controllers for closed loop
simulation. The inputs that were controlled, were feed water, fiiel flow, air flow, bumer
tilt.) The computer based data acquisition system was used to perform three sets of tests
in two power stations. Pseudorandom binary sequence (PRBS) input signals were used to
perturb plant set points while doing the tests.

A two-stage recursive least squares

algorithm was developed to estimate initial operating conditions and model parameters. In
these tests, models of the plant were derived and validated. The results indicated that
models of the order 11-16 (number of state variables) depending upon the selection of
variables are suitable for representing significant plant dynamics over a long period.
Sekoguchi (1970) has proposed an analytical method using the theoretical
relationships and transfer fiinctions for predicting plant dynamics from design data. To
make it easy to calculate, some of the transfer fiinctions are reduced to integrated forms.
He has included both anal3^ical results and field test results (step responses), which were
performed using three public utility plants with Sulzer and Benson type boilers. A fairly
good agreement between analytical results andfieldtest results suggests the validity of the
analytical methods. Further, Sekoguchi has observed substantial differences between the
process dynamics of both plants with Sulzer and Benson type boilers.

Boiler Control Literature


Dukelow (1991) has discussed the basic ideas involved in boiler control. He has
extensively covered the boiler process aspect before covering the control aspects.

In the boiler process aspects, Dukelow (1991) has discussed the basic boiler, and
its steaming process, mass and heat balance across various parts, heat recovery from flue
gases, efficiency calculations, various types of fiiels, and their handling.
In the control aspects, Dukelow (1991) says that a carefiil examination of most of
the boiler control applications will show that the overall control system is an
interconnected matrix of the four types of control applications namely simple feedback
control, feedforward plus feedback control, cascade control and ratio control.
Then he has discussed the important control loops in the boiler such as feed water
control system, main steam and reheat steam temperature control, andfiringrate demand
for utility boilers. In firing rate demand, he discusses different modes such as boiler
following, turbine following, and boiler-turbine coordinated control which will be
discussed in a chapter for "Coordinated Control System."
Dukelow (1991) has discussed pumping andfiringrate, steam temperature control
and feed water control for a once through boiler. Most of the once through boilers
operate in the supercritical pressure range. The typical throttle pressure of such units is
3500 psig. In a once through boiler, combustion is controlled by ratioing the air flow to
the fiiel flow. Steam temperature is controlled by ratioing fiiel and combustion air to feed
water. Because of the need to obtain more responsive control of steam temperature
during transient, a parallel proportional control using spray water is incorporated.
Garrett (1967) has discussed the control system objectives. Proper system design
can provide maximum unit efficiency and optimum dynamic response, while insuring
properly balanced boiler inputs and boiler-turbine power levels.
Deviations of the dependent variables of pressure, temperature or oxygen result
from imbalances between feed water, fiiel, air or between the production of steam by the
boiler and the requirements for steam by the turbine. It is important to maintain the

^^f5?"

proper relationships between the boiler and the turbine generator. To accomplish this end,
four basic control objectives resulted.
1.

A common proportional feedforward signal is generated for the independent


variables of feed water, fiiel, and air.

2.

This same signal is used to establish a requirement for the govemor valve area.

3.

Integrating actionfromerrors in the dependent variables of pressure, temperature,


flue gas or megawatt bias the basic demand for feed water, fiiel, air in a nonintegrating marmer.

4.

The feedforward demand signal is properly conditioned to include capacity and


dynamic limitations of any of the major equipment and their associated auxiliaries.
In the coordinated control system for a supercritical coal fired boiler, the

feedforward demand signal is modified by integration action from process variable errors
to establish setpoints for the independent variables like feed water flow, fuel, air flow and
the govemor valve area.

These integration actions are applied in a non-interacting

manner. For instance, megawatt and pressure errors are arithmetically summed, before
being applied to the boiler inputs and the same errors are subtracted before being applied
to the govemor valve area demand. As an example, consider the action if both power
output and pressure are high. The arithmetic sum of these two errors calls for reduced
boiler input and any action by the govemor valve would tend to aggravate one condition
or the other. Consequently, subtraction of the variable errors results in no action by the
govemor valves and all corrective action is taken by the boiler inputs.
Garrett (1967) also talks about the five basic control modes for the supercritical
boiler namely Manual, Turbine Follow, Base input, DEB (Direct Energy Balance), and
DEB-Load Control. The feedforward demand signal is developed in the boiler-turbine
govemor. Mode selection, limits, mnbacks, rate of change, andfrequencybias are located

10

tWMWmm:.

in the boiler-turbine govemor. The rate of change setter limits the rate at which the target
load is fed forward into the boiler input and to the turbine govemor valves. The frequency
bias is included in order that the action of the govemor valves in following the speed
changes can be fed forward into boiler demand.
Garrett (1967) has presented a steam temperature control configuration for a once
through supercritical coal fired boiler.

He used a cascade arrangement where a PI

controller controls the attemperator spray flow rate which is cascaded to a PI controller
that controls the steam temperature. Garrett has discussed reheat temperature control
configuration. The bumer tilts are positioned to maintain the reheat temperature setpoint.
Spray type desuperheaters are also provided as a part of the reheat steam temperature
control but are used only during emergencies and abnormal operating conditions.
Laubli et al. (1970) have described the dynamic characteristics of the supercritical
pressure boiler and its coordinated control system. They have given computer simulation
studies and actual open-and closed-loop tests of supercritical boiler plants.
The coordinated control system makes use of both feedforward and feedback
control concepts. In the feedforward part of the control system, a load demand signal
(required output) is sent to each of the four flow control loops. This feedforward signal
can be properly interpreted as the overall boiler-turbine plant setpoint for electric
generation. Conventional feedback controllers with P (proportional), I (integral), and D
(derivative) actions are used to correct for steady state errors in the controlled variables;
megawatts, throttle pressure, steam temperature, and oxygen.
To fiirther clarify the coordinated control system concepts, Laubli et al. (1970)
have included actual field test data to illustrate certain basic dynamic characteristics of the
supercritical boiler when compared with the dmm type boiler. The coordination of the fuel
and feed water flow loops are more important because of their direct mutual influence on

11

^^I

steam temperature. In fact, the ratio of fiiel flow to feed water flow determines the final
superheat steam temperature during steady state conditions.
Several field test results given by Laubli et al. (1970) illustrate that feedforward
control can be used to supplement the feedback control to improve the load changing
capability of the boiler-turbine unit. The supercritical boiler is inherently very stable
following a load change if the feedforward rate signals are properly adjusted.
The control system tuning philosophy used by Laubli et al. (1970) was based on
the belief that the principal tuning criterion should be power output response.

Heat

storage within the boiler should be used to aid in achieving rapid generation response by
permitting temporary deviations in throttle pressure.
Laubli et al. (1970) have described the behavior of a supercritical boiler, especially
with regard to its capability to generate instantaneous standby capacity. The most rapid
generation of instantaneous standby capacity can be achieved by reducing the turbine
throttle steam pressure. Another possibility of increasing the generator output temporarily
by reducing the energy stored in the boiler is to increase the feed water flow.
Riggs et al. (1995) have presented a nonlinear PMBC controller for steam
temperature control and compared it with a State Variable Controller (SVC) and a PI
controller, for a dmm type coalfiredboiler. The simulation was based upon the Cromby 2
model (McDonald et al., 1971) and was used to test each controller for ten different upset
scenarios that included soot blowing, load ramps, and change in bumer tilt.
Nonlinear Process-model based control (Nonlinear PMBC) uses a nonlinear
approximate model directly for control purposes. The controller model does not have to
be rigorous but should contain the major characteristics of the process. Riggs et al.
(1995) have used the Generic Model Control (GMC) developed by Lee and Sullivan
(1988). For a nonlinear approximate model, an unsteady state energy balance around the

12

>-,.

* * > *

>*:'(^*

^^mmmmm^

secondary superheater and the attemperator was written. The approximate model was
parameterized solving for the effective heat transfer coefficient using a steady state energy
balance on the secondary superheater.
The nonlinear PMBC clearly outperformed the SVC controller which offered only
a marginal advantage over the PI controller. The nonlinear PMBC controller has an ability
to adapt to process gain changes. While the PI controller and SVC controller, both are
fixed gain controllers, show significant cycling, whereas, a PMBC controller, due to its
adaptive gain nature, was able to eliminate the cycling.
Kaya (1990) presents an overview of boiler controls and presents conventional
schemes which use various arrangements of Proportional Integral (PI) controllers. For
steam temperature control, he presents a three-element controller which uses cascade
control to a flow control loop on the attemperator spray flow rate and feedforwards the
load to the attemperator spray flow rate.
Dattatreya (1982) presents an adaptive gain controller for steam temperature and
pressure control on a utility coal fired boiler. Maximum deviations from setpoints were
reduced by a factor of three using a controller with an adaptive gain compared with a
static gain controller.

Nonlinear Process Model Based Control


Nonlinear Process Model Based Control (NPMBC) uses a nonlinear approximate
model directly for control purposes. The approximate model does not have to be a
rigorous model but should contain the major characteristics of the process.
The Generic Model Control (GMC), one of several model based control
techniques, was proposed by Lee and Sullivan (1988).

The control law employs a

nonlinear process model directly within the controller. The control law also includes the

13

i^ -..'iTjfcp-xwnKKVHmBBaa

>}^^"^^^

^ ^ ^ ^ ^

iLJ^^5!5cg

feedback term such that the closed-loop response exhibits zero offset, the process model
is usually a nonlinear description of the process derivedfromfiindamental mass and energy
balance considerations.

GMC is a control framework for both linear and nonlinear

systems in the time domain. All the equations and structure of GMC will be explained in
the chapter for "Nonlinear PMBC Controller."
Economou et al. (1986) extended Intemal Model Control (IMC) to include
nonlinear models. This approach is called Nonlinear Intemal Model Control (NLIMC)
and uses an interactive integration of the approximate model for its control law. Rhinehart
and Riggs (1990) compared GMC and NLIMC for a wide range of exothermic CSTR
control problems and found that GMC control law is an explicit numerical formulation and
NLIMC is an implicit one. GMC is easier to apply and implement and requires less
computational efforts.
Nonlinear Model Predictive Control (NLMPC) (Parrish and Brosilow, 1988) uses
the intemal model stmcture but assigns any process/model mismatch to unmeasured
disturbances. Bequette (1989) presented a version of NLMPC that used a single stepahead control law with continuous model parameter update. Patwardhan et al. (1988)
applied NLMPC for the startup of an open-loop unstable, exothermic CSTR.

Riggs

(1990) applied a version of GMC to the same problem considered by Patwardhan et al.
and found that GMC and NLMPC gave an equivalent performance. Thus, there is not
much difference between the performance of various nonlinear PMBC methods when the
same approximate model is used. The basic difference between the various nonlinear
PMBC methods is the way in which offset is removed, e.g., in GMC integral term is used;
while in NLMPC adjustment of disturbances is used to remove the offset.
A variety of nonlinear models have been investigated using the differential
geometric approach. Nonlinear decoupling via feedback using this approach has been

14

-yj'.i*t-'--:-':-

!"'"'^^^

explored by Isodori et al. (1981). An overview of nonlinear geometric methods have been
summarized by Kantor (1987).

Ha and Gilbert (1983) have outlined a complete

characterization of decoupling laws for a general class of nonlinear systems.


Henson and Seborg (1989) reviewed the field of differential geometric control
strategies. By studying a CSTR and a pH control problem, they found that static methods,
of which nonlinear PMBC is a subset, provided the best control performance and were
relatively insensitive to process/model mismatch. They state that GMC is only applicable
to a very restrictive class of problems for which the manipulated variable appears explicitly
in the dynamic model for the output variable. While this statement is true in a strict sense,
it does not pose a practical limitation to GMC since the manipulated variable can usually
be expressed as an explicit function of one or more variables that do appear in the model
equations.
GMC has been successfiilly applied to processes such as coal gasification (Pandit
et al., 1989), pH (Rhinehart and Choi, 1988; Williams et al., 1990; Mahuli et al., 1993),
propylene

sidestream

distillation

(Riggs,

1990),

supercritical

fluid

extraction

(Ramchandran et al., 1992), heat exchanger (Paruchuri et al., 1994), plasma reaction
(Subawalla et al., 1994), and nonideal binary distillation column (Pandit et al.,1992).

15

Ih^MI^^BMnill

I ' l l

'

i*"i-,,*i.>t-.'.-*-^'*', " ^ i A " ^ ' - .

iiiwu.jjj

CHAPTER m
MODELING OF A SUPERCRITICAL COAL FIRED BOILER

A supercritical steam generating unit is the one which operates at a pressure above
the critical pressure of 3208 psia. When water at a supercritical pressure is heated, it does
not boil and does not produce a two-phase mixture of water and steam. Instead, the fluid
undergoes a transition in the enthalpy range of approximately 850 to 1050 btu/lb, during
which the physical characteristics (such as density, compressibility, viscosity etc.) change
continuously and homogeneously from those of a liquid to those of a gas or vapor. The
nature of the process mles out the use of a boiler dmm to separate steam from water and
dmmless steam generators have been universally adopted for supercritical units.
At the boiler's inlet, the high pressure feed water is forced into the boiler tubes. It
is heated as it passes through them and finally is ejected from the boiler's main outlet
(secondary superheater outlet) as the main steam.
The first part of this project was to develop a dynamic simulator for a supercritical
coal fired boiler and bench-mark it against the actual plant data. Texas Utilities (TU)
Electric, Dallas had provided us with the actual data on one of their supercritical units
which we used to benchmark our model.
The supercritical unit for which the data has been provided by TU Electric, Dallas
is designed to deliver superheated steam at a rate of 4003.678 Klbs/hr (maximum
continuous) at 1005^ and 3850 psig (superheat outlet) to a 575 MW turbogenerator.
The reheater is designed to handle 3465 Klbs/hr (maximum continuous) of steam reheated
from 5690F to 1005^. The unit being a divided fiimace design has each furnace half fired
through four tilting tangential windbox assemblies (waterwall dividing the fiimace into
two halves).

16

f'.vji.. - ^ - - . s ^ - ^ > * -

*"-^"'

'

-*.*!

.11. ijijjgmt^ga

W-m

The simulator developed in this work was based on the Cromby 2 model
developed by McDonald, Kwatny and Spare of the Philadelphia Electric Company.
(McDonald et al., 1971). As mentioned above a supercritical boDer is a dmmless unit.
The Cromby 2 model was modified for the dmmless unit. The Cromby 2 model was
selected because it covered the major elements of plants that are of interest, and it had
been validated from 50 to 100% of load.
The boiler simulation software consists of separate modules, each representing a
segment of the boiler system. The boiler was divided into the following segments:

Feed water valve and boiler feed pumps,

Mills,

Fumace,

Waterwall,

Backpass,

Superheater which consists of the following different modules: Divisional panels.


Superheater platen, Desuperheater, Pendent,

Reheater which consists of the following different modules: Desuperheater,


Reheater platen, Reheater horizontal and vertical,

Turbine governing stage,

High pressure turbine,

Low pressure turbine.


Figure 3.1 shows a process schematic for the feed water-steam side of the boiler.

Two general assumptions were made in the development of the model by McDonald,
Kwatny and Spare (1971). First, it is assumed that the dynamics of the steam and flue gas
could be neglected, since the time constants of the steam and flue gas dynamics are much
smaller than the other process time constants. Second, it is assumed that the economizer

17

.4i-y....----'---

:- - = ; * ' - : ; ' ^ = T ^ - : " -

, ^ : ^ J

"'"^^J

and feed water heater dynamics can be neglected. To account for changes in feed water
enthalpy, which occurs with changes in feed water flow, a curve fitting was done on the
actual plant data.
Now, the description of the model equations will be given for the various units
modeled.

Mills
There are a total of eight mills for the entire boiler. The mills are the Raymond
Bowl type manufactured by Combustion Engineering.

The mill is modeled as two

elements, the first represents the mass of the cmshed coal (M^) stored in the mill and the
second represents the deflector and the exhauster. The rate of change of the stored mass
(MCT) is equal to the difference between the inflow and outflow of the coal. The input
flow rate is proportional to the mass of coal stored.

^=K^I^-K^M,

(3.1)

where
K^f = proportionality constant between coal flow rate out of cmsher and mass of the
cmshed coal stored, per second,
Kfs = proportionality constant between coal flow into mill and the feeder stroke,per
second,
Ifg = normalized feeder stroke, lbs,
MCT = mass of coal in the cmsher zone of the mill, lb.
The flow of primary air through the mill is proportional to the feeder stroke (K^^ \^ and
the air-coal mixture fills the entire volume representing the deflector and the exhauster.
The mass balance on the fraction of coal (X^f) in the volume yields.

18

..Sfc"-*%fc-

> ' ' "

^tTT'

mm

sa

dX <f

dt

p^V

(K^M^-W,),

(3.2)

where, the coal flow to the fumace is given by.


(3.3)

and
^poo

-^ao'^

^wfs^fs >

(3.4)

where
^wfs ~ proportionality constant between primary air flow and feeder stroke, per second,
V = mill volume, cubic ft,
Wao = primary air flow with zero coal flow, lbs/sec,
Wf = mass rate of coal flow to the boiler, lbs/sec,
Wpao = primary air flow through the mills, lbs/sec.
Pa = density of the primary air, lb/cubic ft,
Pcf ^ density of the coal, lb/cubic ft.
The process of combustion of pulverized coal in a fiimace is extremely complex.
A macroscopic view of the process indicates that if Wf lbs/sec of coal having a heating
value of T' is bumt with a thermal efficiency of TI, the gas temperature with no heat loss
would be given by the following equation. (McDonald et al., 1971)

(3.5)
gf

where
Q

= specific heat of combustion products at the averageflametemperature,


btu/lb/degree R,

Tf = adiabaticflametemperature, degree R,
T/ = air heater outlet temperature, degree R,
19

l.iuttM

Wg = mass rate of gas flow through the boiler, lbs/sec.


Since, (W^ /Wf = 5.3317, for the given fiiel composition)
Wa = 5.3317 *Wf* pet,

(3.6)

where
Pet = weight percent excess air,
Wg = mass rate of the air flow through the boiler, lbs/sec.
The total gas flow is the sum of the fiiel and the air flow.
Wg = Wf+W3.

(3.7)

The specific heat of the combustion products at the calculated range of gas temperature
was found to be very close to the specific heat of nitrogen alone. The specific heat of
combustion products (Cgf) as a fiinction of coal flow can be given as the following
equation.
Cgf = 0.2918 + 0.8828 e "^ * Wf

(3.8)

Fumace
The unit considered has a divided fiimace design (center waterwall divides the
fiimace into two halves). Each fiimace half is identical and is fired through four tilting
tangential assemblies.
The majority of the boiler heating surface is contained in the fiimace walls where
the tubes are spaced so as to prevent a tube to tube wall constmction. The large mass
contained in the total waterwall tubes represents a significant heat capacitance. In the
fumace, the rate of change of metal temperature (Tm^ is obtained by the energy balance.

A^.c,^=e^-e.>

(3 9)

20

' - - - &

where
^pm ^ specific heat of a metal in the waterwaU, btu/lb/degree R,
M^ = total effective mass of the waterwall, Klb (includes metal mass and water mass),
Qnvs "^ I'^te of heat transfer from the gas to the metal in the waterwall, Kbtu/sec,
Qws ^ rate of heat transfer from the metal to steam in the superheat waterwall, Kbtu/sec.
These units must be designed to operate in the nucleate boiling range (Babcock and
Wilcox, 1972; McAdams, 1954). In the nucleate boiling range, the heat transfer to the
steam is given by the following equation. (McAdams, 1954; Rohsenow et al., 1985)
Q^=K(T^-T,)\

(3.10)

where
k^ = heat transfer coefficient between tube metal and steam in waterwall,
Kbtu/(degree R)Vsec,
Tjj = average steam temperature in waterwall, degree R.
The heat removed from the hot gas is given by the equation
O
\-rws

=k, HP -P ) ,
\w \ gas

nrws f

(3.11)
^

'

where
kj^' = overall heat transfer coefficient between gas and tube metal in waterwalls,
Kbtu/ (degree R)4/sec,
T = mid section gas temperature of the waterwall, degree R.
gas

'^

The average and the exit gas temperatures are obtained by assuming that changes in the
energy stored in the fiimace gas is negligible and writing a fiimace gas energy balance
^^=7>-^.
(3.12)
T^ = n.-Tf,
C =0.2989+ 0.0002235 * > ,

21

(313)
(3.14)

fc'wfa'

where
Cg = specific heat of the combustion products at the averagefiimacetemperature,
btu/lb/degree R,
Tgs = fiimace exit gas temperature, degree R.
The effect of bumer locations and bumer tilt positions can be taken into account by
recognizing that the heat transfer coefficient k,^' is proportional to the waterwall surface
area exposed to the radiating gases (McDonald et al., 1971). This area is approximately
the exposed waterwall area between the flame center and the fiimace exit. The flow
center location moves in proportion to tan G^ where 9^ is the superheat fiimace bumer tilt
position. Consequently, the heat transfer coefficient can be expressed as
*iw' = ^,w"(l-^2wtan^,),

(3.15)

where
kj^" = basic heat transfer coefficient between gas and tube metal including the effect of
bumer geometry, Kbtu/ (degree R)4/sec,
^2w = coefficient relating a change in bumer tilts on the waterwall heat transfer,
Kbtu/ (degree R)4/sec.
The eight mills supply eight corresponding (four comers of the each half of the fiimace)
vertically arranged rows of bumers, the lowest to the highest being respectively designated
a, b, c, d, e, f, g, h. A is the vertical distance from bumer row H to fiimace exit and 6 is
the vertical distance separating successive bumer rows. The effect of bumer location can
be included by defining

di-\-(a*lS^b*65-^c*SS-\-d*4S-^e*3S+f*lS->t^g*S
a+b-\-c-\-d+e-k-f-\-g-\-h

, _,
^iw ~'^\w

>

A + 72

(3.16)

where a, b etc. are 1 or 0 according to operating or non-operating status of A mill, B mill


etc., and
22

i,]^'.-.;.

~ aa4;;",Viitfas*.2; mjtMm*m.i.'.:fi^-f'

__,^__

^sm

ki^ = basic heat transfer coefficient between gas and tube metal in the fiimace,
Kbtu/ (degree Kflsec,
A = vertical distancefromthe bumer row H to fiimace exit, ft,
6 = vertical distance separating successive rows, ft.
The flue gas after coming out of a fiimace goes through the shell side of different
heat exchangers of the superheater and the reheater. The path it traverses is given below.
Fumace exit > Divisional panel > Superheater platen > Reheater platen >
Superheater pendent > Backpass > Reheater horizontal and vertical > Economizer
> Air preheater. Figure 3.2 shows the typical flue gas path as block diagram. As it
travels along this path, its temperature decreases. The reduction in the temperature of the
flue gas as it passes through the shell side of the heat exchanger is given by the following
equation.

where
fg = flow rate of the flue gas, lbs/sec,
Q = heat transferredfromthe flue gas to the steam in the heat exchanger, btu/sec.
The average temperature of the flue gas across the heat exchanger is given by the
following equation.
T

T +T
=-^^,

where
Tgin = temperature of the flue gas at the heat exchanger inlet, degree F,
Tgout = temperature of the flue gas at the heat exchanger outlet, degree F,
Tgavg = average temperature of the flue gas across the heat exchanger, degree F.

23

iSft5Ai*~l

(3.18)

-^" ""TPiw

-.rrj^

Superheaters and Reheaters


For waterwall, backpass, divisional panels, platen, pendent, reheater platen,
reheater horizontal and vertical steam side, mass, energy, and momentum balance
equations are written.
Mass balance:
y ^

= K-W^

(3.19)

Energy balance:
dh
MS!ii-=Q-Wh
ju

-W M

"tn"tn

"out'*out

(3 20)
V-'"*'^/

Momentum balance:
Pin-P,.t=L ,

(3.21)

Pin

where
fgs = fiiction coefficient for steam flow in a particular unit, [ Psi * lb/ft3/(lb/s)2],
hjn = enthalpy of incoming steam to the unit, Btu/lb,
hout = enthalpy of outgoing steam to the unit, Btu/lb,
M = effective steam mass in that unit, (includes metal mass and actual steam mass), lbs,
Pjn = pressure of incoming steam, psia,
Pout ^ pressure of outgoing steam, psia,
Q = heat transferred from flue gases to steam, Btu/sec,
V = steam storage volume of that unit, ft^.
Win = steam flow rate of the inlet of the unit, lb/sec,
Wout = steam flow rate of the outlet of the unit, lb/sec,
Pjn = density of incoming steam, Ib/ft^,
Pout ^ density of steam going out of the unit, Ib/ft^.

24

.,.>,<.'."->-^

->>-r*v---'a-.-.:'

^s^

- ^ - . ^

The empirical relationships are written for the temperature and the pressure leaving that
particular unit. They can be represented as,
Taut^f

i^out^out ^Kut)-

Pout=fihou,^Paut^Pau,T^t)-

(3.22)
(3-23)

These relationships are derived empirically and using the steam table for the temperature
and pressure range in which T^^t and P^^t change for the different loads. Thus, the two
state variables are enthalpy (h) and density (p) of the outgoing steam.
In the energy balance equation,
e = ^*^*A/,

(3.24)

where At^ (log mean temperature difference) can be given by the following relationships.
(On the gas side, tj = temperature of the incoming gas, ii ^ temperature of the outgoing
gas. On the steam side, tj' = temperature of the incoming steam, X-^ = temperature of the
outgoing steam.)
For the cocurrent flow,

^^^(LzlDz^lZhl,
h

(3.25)

^2-^2*

For the counter current flow.


(3.26)

ini^
^2-^l'

The heat transfer from the flue gases to the steam takes place by two modes: the
convection and the radiation.

So both of these factors should be embedded in the

expression for V. The expression for IT is given by.


U=

(3.27)

{U^^U^^US

25

l|Bfciia6*^iXii:'''.ii..? -iistisi;**'"^-'

"TfW-

where,
Ugg = convective heat transfer coefficient on the gas side or hotterfluidconvection
conductance, Btu/(ft2hi^),
Ups = convective heat transfer coefficient on the steam side or colderfluidconvection
conductance, Btu/(ft2hr<^),
Ujg = radiative heat transfer coefficient.
For the turbulent flow inside the long tubes.
U =U

FF

(3.28)

where
Fpp = physical properties factor, unitless,
Ff = temperature factor, unitless,
Ucs = basic convection conductance, btu/(ft2 hr ^ ) .
Equation (3.28) applies to both the heating and cooling of the fluids inside the clean
conduits.
In the steam boiler, the most important application of convection is in the transfer
of heat from hot combustion gases to the various heat absorbing tubular surfaces.
Compared with the extensive research on heat transfer for fluids flowing inside the tubes,
little has been done to establish convection heat transfer coefficients for crossflow over the
tube banks. Perhaps the most authoritative and complete data on crossflow heat transfer
are those obtained as a part of a Babcock and Wilcox research program.
For turbulent flow outside tubes,
Ucg=U'c^FFpPaF.^
where
Fg = arrangement factor, unitless,
Fd = depth factor, unitless.

26

^l5-.^v^^--v- - rr*..^--.

'^^-'-^^

(3.29)

mmmmmmm.

Fpp = physical properties factor, unitless,


Ugg = basic convection conductance, btu/(ft2 hr ^ ) .
The arrangement factor, F^, depends upon the tube arrangement, the ratios of the tube
spacing to tube diameter and the Reynold's number.
While,
Urg=U'rK,

(3.30)

where
Uj = basic radiative conductance, btu/(ft2 hr ^ ) .
Factor "K' takes into account the effect of mean radiative length and fiiel on radiation
conductance or radiative heat transfer coefficient.

When an exchange of energy is

considered between any source and any receiver by radiative mode, all the radiation may
not be seen to fall on the receiver. The receiving surface, such as the banks of tubes, is
cylindrical and may partially obscure some of the surface from "seeing" the source. Thus,
a geometric factor or view factor has to be incorporated in the calculations. It has been
stated that the elements of the fiimace are best handled individually, and in evaluating the
effective or "seen" surface of the tube rows, the assumption is made that the heat source is
a radiating plane parallel to the tube row. The end effects are eliminated by assuming both
the plane of the tubes and the radiating plane to be infinite in extent. All surfaces are
assumed to be black. For such surfaces the view factor is equal to 1 (Kern, 1950).
In Equation (3.27), TJ represents the combined or overall heat transfer coefficient
for most boilers and superheaters where the resistance to heat flow through the metal is
very small and may be neglected. Only the convective and radiative parts of the heat
transfer coefficient are significant and are considered (Babcock and Wilcox, 1972).

27

^ S ^ ' S i i i ^ .V.-.Sttfe

Superheater and Reheater Desuperheaters


The primary method of throttle temperature control is the use of superheater spray
in which sub cooled liquidfeed water from the discharge of the boiler feed pumpis
sprayed into the steam as it flows from the superheater platen to the superheater pendent.
The feed water is introduced through a spray nozzle at the throat of a venturi section
within the steam line. The spray action at the nozzle and the high velocity through the
venturi assures that the feed water is completely nuxed as it mixes with the main steam
flow. The venturi is surrounded by a thermal sleeve designed to avoid thermal shock of
the high pressure piping.
The process is considered to be typical of adiabatic mixing of the two streams and
mass and energy balance equations are written.
Mass Balance:
W^t = W,+W^,

(3.31)

Energy Balance:
W^fiout = W,X^W^r<^^,^^

(3.32)

where
hin = enthalpy of the steam coming in into the desuperheater, btu/lb,
hout = enthalpy of the steam coming out of the desuperheater, btu/lb,
hgpray = enthalpy of the spray water, btu/lb.
Win = "^^^ s^^"* ^^^ ^*^ ^ ^^ desuperheater, lbs/sec,
Wout = outlet steam flow rate to the desuperheater, lbs/sec,
Wspray = flow rate of the spray water, lbs/sec.
Feed water taken from the intermediate stage of the boiler feed pump can be
sprayed into the steam before it enters the reheater horizontal (after it comes out from the
high pressure turbine) in a similar fashion to the superheater spray. The final reheater

28

?|Mt.'!>.>'".

temperature can be controlled in this manner. The mass and the energy balance equations
are written in a similar way as in the superheater.
In order to fiirther improve the realism of the simulation, hysteresis in the
attemperator spray control valves (5%), lag in the temperature sensor (5 seconds time
constant) was added.

High Pressure Turbine


After passing through the turbine governing stage, the steam passes through the
high pressure turbine. The flow conditions in the high pressure turbine are such that sonic
velocities are assumed to exist over the entire load range so that flow and first stage
pressure are related by
W^=K^*P^.

(3-33)

where
Kjjp = flow coefficient for reaction stages of high pressure turbine,
Pjj = steam pressure at the exit of the throttle valve, (first stage pressure), psia,
Wsso = steam flow rate at the secondary superheater outlet, lbs/sec.
The temperature at the high pressure turbine inlet can be obtained using the state
relationship

which was determined using the steam table.


where
hhp = steam enthalpy in the impulse chamber, btu/lb,
Thp = impulse chamber (first stage) temperature, degree R.
The ideal high pressure turbine outlet temperature is computed assuming an isentropic
expansion through the turbine (McDonald et al., 1971)

29

t = T,A'\

(3.35)

where
k = ratio ofspecific heats (at constant pressure/at constant volume).
Per = steam pressure at the discharge of high pressure turbine, psia,
TCT*

= ideal steam temperature at the discharge of high pressure turbine, degree R.

The ideal cold reheat enthalpy can be obtained using the state relationship
Kr=f(.PcrX)-

(3.36)

The actual cold reheat enthalpy is given by the following equation


K=K-1xpiK-f'lr)<

(3.37)

where
her ^ steam enthalpy at the discharge of high pressure turbine, btu/lb,
h^* = ideal steam enthalpy at the discharge of high pressure turbine, btu/lb,
h^p = steam enthalpy in the impulse chamber, btu/lb,
T||jp = overall efficiency for the reaction blading.
The total high pressure turbine power generation is given by the following equation.
MW, = W^{h^ -h)\.Q542.

(3.38)

(The factor 1.0542 is a unit conversion factor.)

Low Pressure Turbine


After leaving the reheater, the steam enters the intermediate pressure (IP) turbine
followed by the low pressure turbine (LP). The IP and LP turbine stages are considered
together as the low pressure turbineand the exhaust hood conditions are assumed to be
constant. As in the case of high pressure turbine, sonic velocity is assumed to exist and
the flow equation can be given as
30

'--* " ' " ^ ' ^ ~

',^

.. _s'.

':.

"t^-

"*9m

mmwmmmn

^n,=-^.

(3.39)

where
Pro = steam pressure at the reheater outlet, psia,
Tro = steam temperature at the reheater outlet, degree R,
Wro = reheater outlet steam flow rate, lbs/sec.
The energy equation relating the power output to the decrease in energy of the steam as it
passes through the turbine is given by the following equation.
MW, = rj,(h-hJW^*l0542,

(3.40)

Here also, the factor 1.0542 is used for the unit conversions,
where
h^j^ = exhaust enthalpy, btu/lb,
Tij = low pressure turbine isentropic efficiency.

Turbine Groveming Stage


The flow of steam from the secondary superheater (pendent) bifurcates and each
half after passing through a normally open throttle valve, enters a steam chest on either
side of the turbine. Each steam chest contains four governing valves that are opened
sequentially to control the flow of steam to the turbine. Each valve discharges the steam
to its own section of the first stage nozzle ring.
The valves are of the single seated plug type and are operated by a hydraulic servo
motor which is positioned by the turbine govemor. The servo motor raises or lowers a
valve bar which opens and closes the valves in a predetermined sequence, each valve
opens when the bar engages a nut on the threaded stem of the valve.
The steam leaving each governing valve passes through an associated convergent
nozzle designed to increase fluid velocities and chaimel the flow into the first stage
31

..Iw.S.,I..

A i * K-Tf

blading. The steam then flows through the impulse blading which, ideally, converts the
kinetic energy into work.
It has been observed that the flow through the valve-nozzle combination may
achieve sonic velocity. If this is the case, a shock wave will exist and the flow can be
computed from the valve inlet conditions using the standard formula,
W^ = C-^A,

(3.41)

where
A = governing valve area,
C = flow coefficient for the sonic flow through the valve-nozzle combination,
Psso ~ pressure at the secondary superheater outlet, psia,
Tgso = temperature at the secondary superheater outlet, degree F
Wgso = steam flow rate at the secondary superheater outlet, lbs/sec.

Feed water valve


Condensate is delivered to the suction pipe of the three boiler feed pumpsfromthe
first four feed water heaters.

The pressure and temperature of the feed water are

increased in the boiler feed pumps and the flow from the pumps is regulated by the feed
water valve. In the model, the feed water flow is govemed by Bemoulli's equation
restricted by the following assumptions:
1. No change in elevation.
2. Inlet velocity at the feed water valve equal to zero.
The flow through the feed water valve is related to the pressure drop by the
relation
W^^=K^A}(P,-P^),

32

(3.42)

'-r^

where
Af = normalized feed water valve flow area,
Kf = flow coefficient for the feed water valve,
^ecoo = pressure at the economizer outlet, psia,
Ph = boiler feed pump discharge pressure, psia.
The superheater and reheater spray flow is extracted from the boiler feed pump
discharge header which is after the feed water control valves. The actual flow to the
boiler is given by the following relationship.

W =W -W*pray -W
' rsprayy

(3.43)

where
Wg = feed water flow rate to the boiler, lbs/sec,
Wgp = water flow through the feed water valve, lbs/sec,
Wspray ^ Superheater spray flow rate, lbs/sec,
Wrepray ^ reheater spray flow rate, lbs/sec.
In order to improve the realism of simulation, hysteresis was added in the feed
water valve (0.5%).

Model Bench-Marking
The dynamic simulator was bench-marked against the steady state plant data
provided by TU Electric, Dallas (Calle, 1995). The plant data provided include power
output; temperature, pressure of the steam at various locations in the unit, amount of fuel
fuel heating value, steam flow rate, spray flow rate; flow rate, temperature and pressure of
the steam at the various stages within the turbine etc. This data was provided at different
power outputs (loads). The data received is listed in Table 3.1 and Table 3.2.

33

Jit

There are several heat exchangers at various locations within the superheater
section, e.g., division panels, platen, pendent etc.

Similarly, there are several heat

exchangers within the reheater section. The steam properties, i.e., steam temperature,
pressure, flow, etc., were not available at the inlet and outlet of each of these heat
exchangers. These properties were available only at the certain intermediate locations e.g.
superheater pendent outlet, cold reheat, hot reheat, economizer inlet, boiler exit etc.
Normally, in an actual utility, thermocouples are not installed at all the locations to
measure temperatures.

But, TU Electric provided us with the approximate flue gas

temperatures at different locations within the superheater and reheater section. Table 3.3
shows these values. From the amount of fiiel fired, fiiel composition, excess oxygen
percentage, the amount of flue gas produced was calculated. The values of flue gas flow
rate are listed in Table 3.5. From that, the heat loads for each heat exchanger were
estimated. Normally, thefireballtemperature is not measured in an actual utility. But the
adiabaticflametemperature seems to be very close to the actualflametemperature. Thus,
by knowing the flame temperature and the heat gained by each heat exchanger, a
temperature profile of the flue gas was estimated from the fireball to exit past the
economizer.
Similarly, from the knowledge of heat loads of different heat exchangers, steam
properties (e.g., temperature, pressure, density, enthalpy) were estimated on the steam
side. In Equation 3.21, pressure drop across the unit is related to inlet steam flow rate to
the unit. While doing bench-marking, the values of 'f^' or fiiction coefficient for steam
flow were adjusted to get the proper value of the inlet steam flow rate to the unit. The
values of T ' for different units at different load conditions are listed in Table 3.5. These
values o f f were regressed as a function of current values of steam flow rate and square
of current values of steam flow rate. Thus, as load changes, steam flow rate changes and

34

PJ'-WMi.-i-:

wmmmmmtmmmmi^

SO f^^ also changes. The initial values of the heat transfer coefficients were estimated using
the steady state version of Equation 3.20. The values obtained for different units are listed
in Table 3.5. Similarly, for the fiimace, k^ and kj^' were estimated assuming that at
steady state Q ^ = Qj^. The values obtained are listed in Table 3.5. The properties that
were provided by TU Electric were used as a point of reference for these estimations.
These kind of calculations were repeated at several different power outputs for which data
was provided by TU Electric.
The data provided by TU Electric and the estimated property data were used as
the input for the dynamic simulator. The dynamic simulator was mn to steady state using
the input data. The steady state values of different properties such as temperatures,
pressures, flow rates at different locations within the system, power output of the system
were found to be very close (within 5%) to the input or the design data.
A general program was written where a user can select the unit load at which the
system to be mn. The program automatically selects the inlet conditions by regressing the
data given in Tables 3.1 and 3.5 and mns the dynamic simulator to steady state. The input
variables to the simulator which are obtained by regression of the input data are fiiel rate,
feed water flow rate, spray rates, etc.

Table 3.4 shows comparison of the values

calculated by the simulator with the plant data at 473 megawatts power output. The
relative percentage error shows a close fit between the plant data and the values calculated
by the simulator.
Some open-loop tests (with no controller on) were performed on a simulator by
giving step changes in the manipulated variables like superheat spray flow, bumer tilt
position, fiiel flow, feed water flow, govemor valve position, etc. The responses of the
controlled variables, such as power output, throttle pressure, superheated steam
temperature, reheated steam temperature, were observed for these step changes in the

35

, K .> . i . " -" liSk-U'^-i^i'af^.i

f^^mm

manipulated variables. No dynamic data was available from a supercritical unit to bench
mark these open-loop responses. However, some open-loop responses from a dmm type
of unit at Southwestem Public Service Company, Lubbock were obtained (Buchanan,
1995). Open-loop responses were qualitatively bench marked against those responses
observed industrially.

36

""~T;sja

Governor valve

/f^

Feed pump and


Feed water valve

]Nomenclature

Economizer
Waterwall
Back pass
Divisional panel
Superheater platen
Superheater desuperheater
Pendent
High pressure turbine
Reheater platen
Reheater desuperheater
Reheater horizontal and vertical
Intermediate pressure turbine
Low pressure turbine
Condenser

Econo
WW
BP
DP
Platen
Desup
Pend
HP
Plat
Des
Hor
IP
LP
Con

Figure 3.1 Process Schematic for the feed water-steam side of the boiler

37

DP
/

N
}

Plat

N
}

RH
Plat

- ^

Pend

\
/

\ y

RH
Hor

s.

Flue gas
and flame

Econo
\ y

APH

Divisional Panel
Superheater Platen
Reheater Platen
Pendent
Reheater horizontal and vertical
Economizer
Air preheater

DP
Plat
RHPlat
Pend
RHHor
Econo
APH

Figure 3.2 : A typical flue gas path

38

mmmmmmmmm
'-

Table 3.1: Data received from TU Electric

Properties

Units

Klbs/hr

For 327
megawatts
power output
2234.78

For 473
megawatts
power output
3277.19

For 502.5
megawatts
power output
3472.56

Feed water
flow
Superheat
spray flow
Reheat spray
flow
Throttle flow
Throttle
pressure
Cold reheat
pressure
Hot reheat
pressure
First stage
pressure
Main steam
temperature
Cold reheat
temperature
Hot reheat
temperature
Lignite flow
Heat content
Heat input
Excess oxygen
HP turbine
efficiency
IP turbine
efficiency
LP turbine
efficiency
Boiler
efficiency
Used energy
endpoint
Economizer
gas out

Klbs/hr

44.55

62.55

78.65

Klbs/hr

0.0

0.0

0.0

Klbs/hr
psia

2234.78
3514.9

3277.19
3515.6

3472.56
3513.8

psia

374.5

543.1

575.4

psia

343.1

497.7

527.5

psia

1383

2041.2

2170.4

degF

998.6

996.7

997.7

degF

508.4

545.3

552.4

degF

975.2

1003

1009.5

tons/hr
Btu/lb
MMBtu/hr
%

242
6741.7
3262.98
3.71
70.89

351.4
6775.7
4761.96
3.29
78.91

376.5
6501.3
4895.52
2.73
80.62

90.42

90.55

90.42

82

81.93

82.39

83.76

82.46

82.78

Btu/lb

1070.3

1073.4

1072.2

degF

648

727

752

39

js:j.d,z^i^iMaaitSS^^

mmm

Table 3.2 : The fiiel specification receivedfromTU Electric


Species
C
H
0
N
S
Ash
UjO

Percentage
39.2
2.99
11.04
0.57
0.61
14.31
31.27

Table 3.3: Approximate flue gas temperature


Sections
Entering superheater
section
Around divisional panel
Around reheater platen
Around back pass
Around reheater
horiozontal and vertical
Leaving economizer

Units
degF

Temperature
2300

degF
degF
degF
degF

2200
2000
1600
1300

degF

850

40

Table 3.4: Comparison of the values calculated by the simulator with the plant data
Relative %
error

473
3280
3515
540

Values
obtained by
simulator
472
3295
3503
545

psia
DegF

497
995

500
988

0.6%
0.7%

DegF

545

538

1.2%

DegF

1000

995

0.5%

Properties

Units

DatafromTU
Electric

Power
Throttle flow
Throttle pressure
Cold reheat
pressure
Hot reheat pressure
Main steam
temperature
Cold reheat
temperature
Hot reheat
temperature

MW
Klbs/hr
psia
psia

41

-^*'v<'*#il

0.2%
0.5%
0.3%
0.9%

mmm
-

y-

Table 3.5: Values of the parameters used in the simulator


Properties

For 327 megawatts


power output

For 473 megawatts


power output

1.85e-3

8.65e-4

For 502.5
megawatts power
output
7.59e-4

fgg for back pass


(psia*lb/ft3)/
(lb/sec)2
fgg for divisional
panel
fgg for waterwall

1.52e-3

7.1e-4

6.24e-4

0.04053

0.01913

0.01732

fgg for platen

1.271e-3

5.93e-4

5.22e-4

fgg for pendent

7.61e-4

3.53e-4

3.11e-4

fgg for reheater


horiozontal
fgg for reheater
platen
Flue gasflowrate
(Klbs/hr)
Heat tranfer
coefficient for back
pass (Btu/hr deg F)
Heat tranfer
coefficient for
divisional panel
Heat tranfer
coefficient for
platen
Heat tranfer
coefficient for
pendent
Heat tranfer
coefficient for
reheater platen
Heat tranfer
coefficient for
reheater horiozontal

4.81e-5

4.73e-5

4.87e-5

1.52e-5

1.5e-5

1.41e-5

3836.882

5434.296

5729.97

400.6414

498.71

505.42

20.1396

27.9974

29.4589

21.98

30.4

31.9973

55.5

75.52

79.08

31.53

44.47

44.55

122.42

170.4228

174.34

6.5154e-5
1.788e-12

2.316e-5
2.4454e-12

1.8977e-5
2.62e-12

if^

42

J..

..

,^-,-#j

mmm

CHAPTER IV
COORDINATED CONTROL SYSTEM

There are different modes in which the control system for a boiler-turbine can
operate, e.g., boiler following, turbine following, boiler-turbine coordinated control
system.
In the boiler following control, the control systems for the boiler and turbine are
separate and uncoupled.

Starting with the steady state loading, any control system

demand for more electric power is applied only to the turbogenerator.

Either from

additional load on the electric system or from a remote demand signal, the turbine
govemor valves open. The result is that the turbine asks for additional energy input in the
form of superheat steam. Since the boiler was previously producing an amount of steam
with a lower total energy level, the pressure will begin to drop. As the pressure drops,
some steam will be produced due to release of energy from boiler energy storage (energy
stored in the boiler fumace mass due to the high temperature and pressure in the system).
The drop in throttle pressure and the change in steam flow requirement activate the
combustion control system to increase the firing rate of the boiler and bring the steam
pressure back to its setpoint. The effect of the increased firing rate starts to be effective
about 20 to 40 seconds after the demand for additional electric generation. Stability with
pressure at setpoint and power output at the desired level, is not achieved until several
minutes have passed. Figure 4.1 shows the typical schematic for boiler following system
and power output, throttle pressure response for the same.
Of the various front end control arrangements, boiler following is the most
responsive to a change in electric energy demand but also is the most unstable. The
turbine valves open immediately to produce the required additional generation although

43

MBMtoArfirv^-vV^-

-T-:-VSC^i'^

* >4 a^a..d

the pressure may be below the setpoint. As the pressure starts to retum to the setpoint,
the electrical generation starts to increase above the desired level, causing the turbine
steam valves to begin closing. This causes boiler pressure to increase. The ringing of the
boiler following scheme arises from this interaction between the turbine valve control and
the boiler control (Dukelow, 1991). A greater feedforward precision is needed in a utility
boiler control system than is required for industrial boiler because it is more important to
maintain the steam temperature close to setpoint at all times.
The slowest but the most stable arrangement is "turbine following" development of
thefiringrate demand. The demand for an increased power output is used to increase the
firing rate to the boiler. As the additional steam energy is obtained from the increased
firing rate, the throttle steam pressure tends to rise. This causes the turbine throttle
backpressure control to open the turbine valves. Only then is there any addition to the
energy generation rate. This system does an excellent job of pressure control but does not
allow the borrowing of energy from, or depositing of energy to, boiler energy storage
(energy stored in the boiler fiimace mass due to the high temperature and pressure in the
system) during a load change. The steam is thus denied the benefit of using boiler energy
storage to assist in making load changes. The response time is several minutes for this
mode. Figure 4.2 shows the typical schematic for Turbine following mode and power
output, throttle pressure response for the same.
By coordinating the action of the boiler firing and the turbine valve action into a
single subsystem (boiler-turbine coordinated control), the firing rate demand control can
be improved. The major portion of the responsiveness of the boiler following method and
no ringing of the turbine following scheme are thus combined. In this arrangement, the
power demand is applied to both the turbine valves and the firing rate demand as a
feedforward signal. The sum of the power output error and the steam pressure error is the

44

aai^ ^ ^ v w * i K * |

total error input to the firing rate demand proportional plus integral (PI) controller, which
is the final control of the firing rate demand. Thefiringrate demand is used to drive the
total error (error in power output plus error in steam pressure) to zero. Load can be
picked up or dropped without stretching the boiler stability. In the end, a sustained load
change in either direction can be handled more smoothly and at a faster rate (Dukelow,
1991). Figure 4.3 shows a schematic for boiler-turbine coordinated control system and
power output, throttle pressure response for the same. The effect of increased firing rate
starts to be effective immediately after demand for additional electrical generation.
Stability of pressure at setpoint and power at the desired level is also achieved very
quickly as can be seen from Figure 4.3. Keeping these features in mind an attempt was
made to develop a coordinated control system for the supercritical coal fired boiler under
consideration.
A pressure feedback to this system is pressure at the turbine throttle. Since a fixed
setpoint reference at the turbine throttle is needed, the controller requires both
proportional plus integral action. Thus, in this coordinated control system, the boiler
master and the turbine master will control on throttle pressure error plus power output
error using a (PI) mode, with a unit load demand as a feedforward signal. Table 4.1 gives
tuning parameters for feedback and feedforward controllers for boiler master as well as
tuning parameters for feedback and feedforward controllers for turbine master.

Turbine Master Control


The purpose of the turbine master controller is to control the amount of steam
going to the turbine, i.e., to control the turbine valve position. The steam flow to the
turbogenerator can be determined by measuring the first stage turbine pressure. Refer to
Figure 4.4. It shows the exact place at which first stage pressure can be measured in the

45

l l i l l i i i f i f i H i r ' ' '^

system. The relation between energy flow to the turbine and first stage pressure is nearly
a straight line. The output of the turbine master controller is a setpoint for first stage to
throttle pressure ratio. The turbine master controller output is actually a percentage
signal. It is transformed into a first stage to throttle pressure ratio demand or setpoint
signal once it is directed into that loop. This setpoint is compared to an actual first stage
to throttle pressure ratio (the turbine energy input should be ratioed to the throttle
pressure to avoid the positive feedback), and the error is used to calculate the new turbine
valve position. The new turbine valve position is compared with the old turbine valve
position and the error in valve position is calculated.

A proportional plus integral

controller is used to raise or lower the turbine valve in order to allow more or less steam
to pass through to the turbine. Table 4.1 gives tuning parameters for this slave loop of the
turbine master controller. Figure 4.4 shows arrangement of boiler, turbine valve and
turbine in a block diagram. Figure 4.5 shows a control diagram for the turbine master
controller.

Boiler Master Control


The basic purpose of a boiler master controller is to control the amount of feed
water, fiiel, air used by the boiler to generate the steam. The boiler master output is the
basic demand or setpoint for all major boiler variables (fiiel, air and water). The output
signal is directed to the fiiel and feed water loops in parallel; however, these signals can be
modified by the firing rate to feed water ratio. However, the effect of firing rate to feed
water ratio is very small. Figure 4.6 shows a complete control diagram for the boiler
master controller.
The firing rate to feed water ratio is set by a Proportional plus Integral plus
Derivative (PID) controller driven by waterwall temperature error. Table 4.1 gives tuning

46

wmm
4

parameters for this loop. The waterwall temperature setpoint (T^g^tpoint is obtained by a
(PI) controller driven by superheat temperature error and the rate of change of superheat
temperature. Table 4.1 gives tuning parameters for this loop. When a value offiringrate
to feed water ratio given by the controller falls within the certain limits (1% above or
below the base case value), the boiler master output signal is passed unaltered to the fiiel,
air and feed water flow control loops. When a value of firing rate to feed water ratio
given by the controller is greater than the upper bound of the limit, the demand for fiiel
and air flow is ratioed up above the boiler master output; and the demand for feed water
flow is ratioed down below the boiler master output. In other words, the setpoint for the
firing rate loop is modified by a factor. That factor is obtained by calculating the relative
percentage change offiringrate to feed water ratio given by the controller with respect to
the base case value and 10 to 15% of this value is taken to modify the setpoint. The
setpoint for the feed water loop is modified by the same factor only in the opposite
direction to firing rate setpoint change. In a similar manner, the fiiel and air demand are
ratioed down and feed water ratioed up for firing rate to feed water ratio less than the
lower bound of the limit.
The boiler master controller output is a percentage signal. It is transformed into a
feed water demand signal and fiiel demand signal once it is directed into that particular
loop. The boiler master output, modified by the firing rate to feed water ratio, is cross
limited by available air flow to become the demand for total fiiel flow. When the firing
rate demand increases, air flow goes up first followed by fiiel flow. This avoids the
formation of carbon monoxide due to the presence of insufficient amount of oxygen. As
the firing rate demand decreases, fiiel goes down first followed by air flow, again for the
same reasons. This is a cross limiting effect.

47

A proportional plus integral controller

compares total fuel demand and total fuel flow and gives the new value of the feeder
stroke. Table 4.1 gives the tuning parameters for the fiiel loop.
Similarly, the boiler master output, modified by the firing rate to feed water ratio
becomes the demand for feed water flow. A proportional plus integral controller
compares feed water demand and actual feed water flow and gives the value of feed water
valve position. Table 4.1 gives the tuning parameters for feed water loop.
Superheat spray is used to control the superheated steam temperature. Thus,
whenever superheated steam temperature goes above the setpoint, more superheat spray is
needed to bring the temperature back to its setpoint. Similarly, when superheated steam
temperature goes below the setpoint, less superheat spray is needed to bring the
temperature back to its setpoint. The boiler master output signal is a percentage signal. It
is transformed into a superheat steam temperature setpoint when it goes into that loop
(see Figure 4.7). The feed water flow demand described above is used as the basic spray
flow demand signal. A proportional plus integral controller output driven by superheat
temperature error (ejj) and the rate of change of temperature error is used to ratio this
signal into a final spray flow demand. Thisfinalspray flow demand is compared with the
current spray flow and the error is used in a proportional plus integral controller to give
the final spray valve position. Table 4.1 gives the tuning parameters for master and slave
loop ofthis spray flow controller. It is important to keep the spray in the mid range. An
additional feedback loop is introduced to do that. Whenever spray flow changes to
control the temperature, an error in the spray valve position is calculated. It is passed
through a proportional controUer (Kc = 5). The output of the controller is used to modify
firing and feed water so that the spray can be brought back to the mid range.
The reheat outlet temperature is controlled by sequencing of bumer tilts and reheat
spray valves. The boiler master output signal is a percentage signal. It is transformed into

48

_._.

,_fr~^.ATT "

Mi

a reheat steam temperature setpoint when it goes into that loop.

Whenever reheat

temperature is above the setpoint, bumers are tilted in the downward direction in the
fiimace in order to cut down the heat supplied to the superheater, reheater section.
Similarly, when reheat temperature is below the setpoint, bumers are tilted in the upward
direction in the fiimace in order to raise the heat supplied to the superheater, reheater
section (see Figure 4.8). The reheat temperature setpoint is compared with the actual
reheat temperature; and the error is used in a proportional plus integral controller to give
the final bumer tilt position. Table 4.1 gives tuning parameters for this loop. Reheat
spray flow can also be used to control the reheat outlet temperature. But generally, reheat
spray flow is equal to zero in an actual utility plant. It is only used in an emergency
situation. Otherwise, bumer tilts are sufficient for reheat temperature control. If reheat
spray is to be used, it is controlled in a similar fashion to superheat spray flow.
All the loops described above were tuned using autotune variation testing (Astrom
et al., 1983). Initially, all the master loops were tumed off'and all the slave loops were
tuned first, e.g., loop which calculates new turbine valve position, feed stroke, feed water
valve position, superheat spray valve position, bumer tilt position.

After this was

accomplished, all these loops were kept on, while proceeding to tune master loops. Thus,
tuning was done from the bottom of the pyramid towards the top. The boiler master
controller and turbine master controller which gave the value of the setpoint for all the
other loops were tuned at the end.
Whenever there is a load change or any other kind of disturbance in the system,
firing rate and feed water are manipulated to get temperature and pressure near the
setpoint, while the bumer tilts and the spray flows are used as afinecontrol.
After tuning the coordinated control system, using all the conventional loops, the
tests were performed to check the nonlinearities in different loops of this coordinated

49

control system. This was achieved byfindingthe gain information of these different loops.
Each loop was tested individually by giving the setpoint changes in the controlled variable
of the loop in a positive and a negative direction. The changes in the value of the
manipulated variable required to achieve those positive and negative setpoint changes
were found.
calculated.

Then the gains for the positive and negative setpoint changes were
If the values of gains for positive and negative setpoint changes are

approximately equal, then the loop is approximately linear in that operating region. Table
4.2 shows gain for reheated steam temperature controlled by bumer tilt. Since the gains in
the positive and negative directions are approximately equal, this loop is linear. The gain
information in each of these different loops showed that most of these loops are linear
except the loops involving the superheated steam temperature. In the loops involving
superheat steam temperature, the gains differ by a factor of two. Table 4.3 shows gain for
the main steam temperature controlled by the superheat spray, so these loops are
nonlinear. Hence, there is a potential that the performance of these loops can be improved
using nonlinear models. Thus, an attempt was made to introduce nonlinear models into
the loops involving superheat steam temperature control. It will be discussed in a chapter
for the nonlinear process model based control.

50

ry

y MW Y ^

\5/
Gen.

Steam
pressure

MW Generation
B'

I- -

Set
Point

Throttle Pressure
.

. - "

3
4
Time, min

Figure 4.1 Boiler following mode

51

"A

mmmmmmmfmmm

!>-*^-r

Load
Demand

MW Generation
B'

Set
Poin

Throttle Pressure

3
4
Time, min

Figure 4.2 Turbine following mode

52

{III iiiiij inaiiijiiii

PKPVP

^.j

'y* r

Coordinated
Unit
Control

Gen.

MW Generation

B'

z.
Set
Point

Throttle Pressure

~
\

3
4
Time, min

Figure 4.3 Boiler-Turbine coordinated control

53

ftrnm

wmm
r

Turbine
Valve
Boiler

Figure 4.4 Block diagram of Boiler-Turbine valve-Turbine

First stage pressure


Throttle Pressure
+ e.

Feedforward
I signal

Turbine
Master
Controller

Setpoint

\/ Error
Slave loop
new turbine valve position
Figure 4.5 Turbine master controller

54

-I-

megawatt

throttle pressure

Boiler

Feedforward

Matter

signal

Controller

\/ (T
^

^ W W 'setpoint

WW

0^-^r

Feed water

>

!S

<

Airflow
Firing rate
setpoint

setpoint
Fuel
Controller

Feed

N^

water flow
error
P +l

Feed water
valve position

Figure 4.6 Boiler Master Controller

55

controller

master loop
P + I
spray flow
Ratio
error

> X
Feed water
setpoint

P + I
>

Final
spray flow
demand

superheat spray
valve position

Figure 4.7 Superheat spray flow controller

Boiler Master
Output
Reheat
temperature

Reheat
temperature
setpoint

error
P + I
Bumer t i l t
position

Figure 4.8 Reheat temperature controller

56

slave
loop

Table 4.1 Tuning parameters for the Coordinated Control System

k.

Xi

Feedback of Boiler Master

0.015

7.5

Feedforward of Boiler Master

0.086

Feedback of Turbine Master

0.015

Feedforward of Turbine Master


Slave of Turbine Master

"^r

TH

200

100

7.5

0.086

200

100

71.04

291.666

1000

0.001

0.1

1000

0.001

Fuel loop

3.2848 * 10-3

260.4

Feed water loop

4.45 * 10^

10

Master loop of sprayflowcontroller

0.01

250

0.000

Slave loop givingfiringrate to feed water 0.01

ratio
Master loop giving
water wall temperature setpoint

1
Slave loop of sprayflowcontroller

0.0383

291.66

Bumer tilt position

1.2

300

0.001

57

Table 4.2 Gain Comparison 1

Gain

AT/AO in deg F/deg tilt

Positive setpoint change in

Negative setpoint change in

reheat steam temperature

reheat steam temperature

0.8564

0.83

angle

Table 4.3 Gain Comparison 2

Gain

AT/A(spray to feed water

Positive setpoint change in

Negative setpoint change in

main steam temperature

main steam temperature

-409.584

-216.696

ratio) in deg F/(lbs/s)/(lbs/s)

58

^HPi

CHAPTER V
NONLINEAR PROCESS MODEL BASED CONTROL

Nonlinear process model based control uses a nonlinear approximate process


model directly for control purposes. The controller model does not have to be rigorous
but should contain the major characteristics of the process. The nonlinear approximate
model used here is a steady state model, so it provides a steady state gain and decoupling
information for the controUer.
The Generic Model Control (GMC), one of the several model based control
techniques, was proposed by Lee and Sullivan (1988). One stmcture of it uses steady
state models to describe the process gains and interactions and the assumption of low
order dynamics to describe the process transient behavior. The steady state model is
usually a nonlinear description of the process derived from fiindamental mass and energy
balance considerations.

To understand GMC, consider a SISO (single input, single

output) process described by the following dynamic model.


^ = f{y,u,d,k),
at

(5.1)

where
d = vector of measured disturbances,
k = vector of parameters,
u = manipulated variables,
y = output variables.
Assuming that y has a value of yo and it is desired to move the process from yo to ygp in
some time interval x, then equation (5.1) can be approximated using the forward
difference approximation of the derivative.

59

^^^^ = /0'.,M.,*).

(5.2)

Equation (5.2) can be solved directly to determine u, the control action, if we know do, k.
The time interval x is the tuning parameter. If x is small, rapid response is required; if x is
large, a more sluggish response is required.
Since equation (5.2) uses an approximate model, using this control law will result
in a steady state offset. To eliminate this offset, Lee and sullivan added an integral term
(analogous to a PI controller), resulting in the GMC control law given by equation (5.3).
f{y..u,d,,k)-^K,(j^,-y^)^K,\{y-y^)dt

= Q,

(5.3)

where Kj represents 1/x in the Equation (5.2). Kj and K2 are the tuning parameters.
Thus, Equation (5.3) can be solved to calculate the control action.
In many applications, a dynamic model of the form described by Equation (5.1) is
not available. In such cases a steady state model is used. But since the implementation of
GMC requires a dynamic model, the steady state controller model is converted into a
dynamic model assuming linearfirst-orderdynamics.
^=iy.-y\
dt

(5.4)

T^

where, ygs is the solution of the steady state approximate model based upon the current
values of the manipulated variables. Combining the GMC control law with Equation (5.4)
we obtain,
0

where, Kj' is equal to XpKj and K2' is equal to XpK2. This control law yields the steady
state target yss which can be used in the steady state model to yield the required control
action.

60

Henson and Seborg (1989) stated that GMC is only applicable to a very restrictive
class of control problems for which the manipulated variable appears explicitly in the
dynamic model for the output variable. While the statement is tme in a strict sense, Riggs
et al. (1995) reported that it does not pose a practical limitation to GMC, since the
manipulated variable can usually be expressed as an explicit fiinction of one or more
variables that do appear in the model equation.
Consider the superheater pendent section shown in Figure 5.1. Applying an
energy balance around the superheater pendent and superheater attemperator results in:

where
Cpm = heat capacity of the metal, Btu/lb/degree F,
KTpiout) = steam enthalpy at temperature Tpj^ut, Btu/lb,
h(Tpnout) = steam enthalpy at temperature Tpno^, Btu/lb,
hp = enthalpy of water used by the attemperator, Btu/lb,
M^ = mass of the metal in the pendent section, lbs,
0 , ^ = heat transfer rate from thefluegas to the steam in the superheater pendent,
Btu/sec,
t = time, seconds,
Tpiout = outlet steam temperature leaving superheater platen, degree F,
Tpnout = outlet steam temperature leaving superheater pendent, degree F,
W lout = flow rate of the steam out of superheater platen, lbs/sec,
W = flow rate of the water spray in the attemperator (manipulated variable), lbs/sec.
sp

Note that the steam and metal tube in the pendent are lumped together in this model.
Also, Q

is modeled using an overall heat transfer coefficient, hj.


Qp.n=K{T^-TX
61

(5-7)

where
hj = effective heat transfer coefficient between the flue gas and the superheated steam,
Btu/(sec deg F),
Qpcn = heat transfer ratefromthe flue gas to the steam in the superheater pendent,
Btu/sec,
Tg = average gas temperature, degree F,
Tg = average steam temperature, degree F.
Each time the controller is called, a steady state energy balance around the pendent
(the steady state version of Equation [5.7]) is used to calculate hy and this value is filtered
with afirstorder exponentialfilterto provide the value of hj used by the controller. Also,
it is assumed that Tg is constant and equal to 1931^. The major mismatch between
Equation (5.7) (model) and the process (dynamic simulator) is that Tg varies between
1850 to 2000^ in the simulator and the simulator dynamically models the heat transfer to
the pendent while Equation (5.7) is a steady state model.
The manipulated variable is an attemperator spray flow rate or desuperheater spray
flow rate. Next, the GMC control law was used to calculate the new attemperator spray
rate using Equation (5.3).

Q^+W^MT^.)^ wji, - (W^,+w^)h(T^) = k,(T^.-T^)-^k,\(T^,-T^)dt+k,^^

= 0.
(5.8)

where g ^

=hj(Tg-T^).

Since Wgp affects Tg, solution ofthis equation requires an iterative numerical procedure.
The GMC controller uses the secant search method to find the solution to this equation
on-line.

62

wsmmmfww^

The approximate model is parameterized (the value of a model parameter is


adjusted to make a model output match the process) solving for an effective heat transfer
coefficient using a steady state energy balance on the superheater pendent.
/!,=

^^ ,

(5.9)

where, Qpe^ is calculated by performing an energy balance around the superheater


pendent. This on-line parameterization should eliminate any process-model mismatch and
should allow the user not to use the integral term in the GMC control law. However, it
was found that even after on-line parameterization the offset was not eliminated. So, the
integral term was also used in the GMC control law along with the on-line
parameterization.
There are four parameters that must be tuned for this controller: kj, k2, k3 and the
filter factor 'f for the heat transfer coefficient hj-. Nominal values were chosen for kj, k2,
k3 and f such that stable control was obtained. The tuning parameters were optimized
using ISE. The tuning parameters used are listed in Table 5.1.
Now consider the waterwall section. Applying the energy balance results in the
following equation:
M,C^^

= Q^^W^h^ -W^T^),

(5.10)

where
Cpw = heat capacity of the metal, Btu/lb/deg F,
h^,^,,;^ = steam enthalpy at the inlet of the waterwall, Btu/lb,
h(Tbpui) = steam enthalpy at temperature T^pin, Btu/lb,
M^ = effective mass of the metal in the waterwall section, lbs,
0

= heat transfer rate from thefluegas to the steam in the waterwall section,Btu/sec,

t = time, seconds.

63

^bpin ~ outlet Steam temperature leaving waterwall, degree F,


^bpin ~ flow rate of steam out of the waterwall, lbs/sec,
^wwin ^ flow rate of steam into the waterwall, lbs/sec.
Note that the steam and metal tube in the waterwall are lumped together in this model.
Q^^wv is modeled using an overall heat transfer coefficient, h^
Q^=K(T^-LX

(5.11)

where
h^ = effective heat transfer coefficient between the flue gas and the steam, Btu/sec/deg F,
Q^;^^ = heat transfer rate from the flue gas to the steam in the waterwall section, Btu/sec,
Tgww ^ average gas temperature in the waterwall, degree F,
Tjj = average steam temperature in the waterwall, degree F.
*wwin

(economizer outlet enthalpy) in Equation (5.10) can be expressed as a fiinction of

feed water flow, i.e..


^^n =a-i-bW^+cW^ -\-dW^ +eW^,

(5.12)

where
Wg = actual feed water flow to the waterwall, lbs/sec,
a, b, c, d, e = coefficients of the curve fitting.
Wg' can be expressed in terms of firing rate/feed water ratio.
0.98Fr.
7
W=-

(5.13)

where
FR/FW

= firing rate to feed water ratio (manipulated variable),

Wf = fuel flow rate, lbs/sec.


Substituting Equation (5.13) in Equation (5.12), we get equation (5.12) in terms of
(FR^W)

which is a manipulated variable.

64

mi^mm

At each sampling period, each time the controller is called, a steady state energy
balance around the waterwall (the steady state version of Equation [5.11]) is used to
calculate h^ and this value is filtered with a first order exponential filter to provide the
value of h^ used by the controller. Also, it is assumed that Tg^^^^ is constant and equal to
27850F. The major mismatch between Equation (5.11), i.e., model, and the dynamic
simulator of a process is that Tg varies between 2700 to 2850^ in the simulator and the
simulator dynamically models the heat transfer to the waterwall, while Equation (5.11) is a
steady state model.
The manipulated variable is firing rate to feed water ratio (FR/F^V)- Next, the
GMC control law (minus the integral term) was used to calculate the new firing rate to
feed water ratio using Equation (5.3).

Qww "^ ^wMin "wwin

W^j^h{ l^j^ )

MJ2^

= K(T^-T,^)^k,,^

= 0,

(5.14)

where Q^ = h^{Tg^ - 7^).


The solution of this equation requires an iterative numerical procedure.

The GMC

controller uses the secant search method tofindthe solution to this equation on-line.
For this system, the integral term in the GMC control law was not required
because the on-line calculation of h^ eliminated any process/model mismatch.
The approximate model was parameterized solving for the effective heat transfer
coefficient using a steady state energy balance on the waterwall.
1.

T^WW

"' " 71...


-L
'gww

(5.15)

where Q^w is calculated by performing an energy balance around the waterwall. There
are three parameters kj,, kjj and the filter factor Tj' for the heat transfer coefficient h^.
Nominal values were chosen for kjj, ki3 and fj such that stable control was obtained. The

65

-' r

tuning parameters were optimized using ISE. The tuning parameters used are listed in
Table 5.2.

66

iflP"^i^^^"ipwwBiit*5a

r
I
Wplout
Tplout

Wsp

Vv/VAJVL/VVVt

Tpnout

Qpen

L
Figure 5.1 Superheater Pendent Section

67

mrm

Table 5.1 Tuning parameters for GMC controller for superheater spray
Parameters
k,
k.
k.
f

Values
7.5
0.1
200
0.5

Table 5.2 Tuning parameters for GMC controller forfiringrate to feed water ratio
Values
0.8

Parameters
ku

kii

0.9

68

mi^mmmrn

CHAPTER VI
CONTROL STUDIES

Control studies were performed on a simulation of a supercritical coal fired boiler.


The performance of the coordinated controller with all the loops having conventional
controllers was compared with the coordinated controller with the nonlinear models in
certain loops (described in the chapter for nonlinear process model based control). Both
of these controllers were tested for load changes (power changes) as well as disturbances
in the system.

Autotune Variation Testing


Autotune variation (Astrom et al., 1983) is an extremely useful procedure for the
calculation of the ultimate gain and uhimate period of a control loop.

The method

involves closed-loop testing of the process by replacing the feedback controller with a
relay element. The relay switches the manipulated variable between its preselected high
and low limits whenever the controlled variable crosses its setpoint. Use of the relay helps
set up sustained oscillations in the controlled variable with the manipulated variable
oscillating in a square wave. Figure 6.78 shows a typical ATV test (Luyben, 1989). The
ultimate gain and period of the control loop are determinedfromthe following relations.
(6.1)
tr_4h

(6.2)

P =T
* u

>

69

where
a = height of the controlled variable oscillations (amplitude), units of controlled variable,
h = height of the manipulated variable oscillation, units of manipulated variable,
K^ = ultimate gain, units of manipulated variable/units of controlled variable,
m = manipulated variable (see Figure 6.78),
Pu = ultimate period, seconds,
T = time period of the controlled variable oscillations, seconds,
X = controlled variable (see Figure 6.78).
The relay height 'h' is found by trial-and-error so as to get equal amplitude in both the
directions. The value of 'h' usually ranges between 1-6% of the nominal value of the
manipulated variable, depending on the gain of the process. Next, using the ultimate gain
(Ku) and ultimate period (P^) obtained from the ATV test, the Ziegler-Nichols (ZN)
tuning constants (Ziegler et al., 1942) for PI controllers are calculated.

^zN

''ZN

(6.3)

2.2

(6.4)
1.2

where, K^ and P^ are the ultimate gain and ultimate period respectively and K^^, ^ZN, are
the ZN tuning constants (proportional gain and reset time respectively) for that particular
loop. Due to the interaction between different loops in a multiloop controller, each loop
has to be detuned in order to get stable and good performance. The detuning process
involves dividing the controller gain and multiplying the controller reset time of the loop
with the same detuning factor (F) (Toijala and Fagervik, 1972).

70

K =^ .

(6.5)

F
T, = T^*F.

(6.6)

where, K^. is the proportional gain and Xj is the reset time for the PI controller. The
detuning factor can be determined on-line by giving setpoint changes in the controller
loop.
This ATV tuning method was employed in all the PI controller loops to obtain K^.
and Xj. As described in the previous chapter, the tuning procedure was started for all the
slave loops first. Then keeping those slave loops on, tuning was done for all the master
loops (next level). The boiler master controller loop and turbine master controller loop
which gave the setpoint for the rest of the loops were tuned at the end. The detuning
factor was optimized using integral of square of error (ISE) and integral of absolute error
(lAE) in response to the setpoint changes. The general formula for ISE is,
ISE = je^dt,

(6.7)

where e = error between the controlled variable and the setpoint.


The general formula for lAE is,
t

IAE=\\e\it.

(6.8)

In the loops in which GMC controller was used, initial tuning parameters were set
by trial-and-error and subsequently adjustments were made to get minimum ISE.

Results and Discussion


Several tests were performed on both the controllers: the coordinated controller
with all the loops having conventional controllers; and the coordinated controller with the

71

nonlinear models in certain loops and the comparative performances can be seen from the
figures at the end ofthis chapter which will be stated in the forthcoming discussion.
The set of tests are as follows:
1.

Ramp increase in power output: The load demand (power output demand) is
increased 25 megawatts/min. from 420 megawatts to 490 megawatts at which
point load demand remains constant.

2.

Ramp load profile: The load is ramped up at 25 megawatts/min. from 420


megawatts to 490 megawatts and after some time it is ramped down at 25
megawatts/min. until it reaches 340 megawatts where it is maintained for the
duration of the test.

3.

Ramp decrease in power output: The load demand is decreased 25 megawatts/min.


from 420 megawatts to 340 megawatts at which point load demand remains
constant.

4.

Step increase in power output: The load demand is increased instantaneously from
420 megawatts to 460 megawatts at which point load demand remains
constant.

5.

Step decrease in power output: The load demand is decreased instantaneously


from 420 megawatts to 380 megawatts at which point load demand remains
constant.

6.

Removal of a set of bumers: A set of bumers in the fiimace is removed.

7.

Fumace wall soot blowing-conduction change: The fiimace wall soot blowing
operation modeled by decreasing ash thickness on the fumace wall.

8.

Fumace wall soot blowing-radiation change: The fiimace wall soot blowing
operation is modeled by decreasing the radiation heat transfer coefficient for the
fumace wall.

72

^^^^W"^9l"""

9.

Primary superheater soot blowing: The primary superheater soot blowing


operation is modeled by a decrease in the ash thickness on the primary superheater
tubes.

10.

Secondary superheater soot blowing: The secondary superheater soot blowing


operation is modeled by a decrease in the ash thickness on the secondary
superheater tubes.

11.

One half percent excess air increase: The excess air is increased 50% in the form
of pulse lasting 2 minutes.
In an actual utility plant, approximately 5% of the maximum load limit is used as

the maximum ramp rate (Crow, 1995a). Hence, in first three tests a ramp rate of 25
megawatts/min. was chosen which is approximately 5% of maximum load for the unit
considered. In tests 4 and 5, a step change of 40 megawatts was used. This step change
was chosen in order to test the robustness of the controller (Crow, 1995b). The rest of
the tests were selected as stringent tests of steam temperature and pressure control system
performance. These tests include the interactive effects of the controllers and the physical
elements in the boiler for events likely to occur in normal operation (Riggs et al., 1995).
In all the tests mentioned above, the initial conditions were lined out for operation at
418.6 megawatts, 987.47 ^F superheated steam temperature (main steam or high pressure
steam temperature), 3499.53 psia throttle pressure and 993.3 ^F reheat steam temperature.
The control interval was 0.5 seconds.
Figures 6.1 through 6.4 compare the performance of both the controUers for four
variables, namely power output, throttle pressure, superheated steam temperature and
reheated steam temperature for test number 1.

Figures 6.5, 6.6 and 6.7 show the

superheat spray, fuel and feed water respectively given by both the controllers for this test.
(These three manipulated variables are given since two loops involving these three

73

manipulated variables show nonlinear behavior, as described in the previous two chapters.
Nonlinear models are applied in these two loops only. The nonlinear model, which gives
firing rate to feed water ratio, affects both fiiel and feed water, although the effect is very
small as described in the previous chapter.) At 7000 seconds, the load demand was
ramped up from 420 to 490 megawatts at 25 megawatts/min. (Henceforth, a coordinated
controller with all the loops having a conventional controller will be referred to as
controller 1, and a coordinated controller with the nonlinear models in certain loops will
be referred to as controller 2. In all of the plots, controller 1 performance is denoted by
thin solid line, while controller 2 performance is denoted by thick solid line.) As shown in
Figure 6.1, both the controllers do well to bring the load from 420 megawatts to 490
megawatts very quickly. There is only a small difference in the performance for power
output for both controllers in this test. The plots are actually overlapping. Figure 6.2
shows the performance of throttle pressure for this test. As the power output increases,
throttle pressure goes down initially, but recovers back to its setpoint. Figure 6.3 shows
the performance of superheated steam temperature. Controller 2 clearly outperforms
controller 1 by showing about three times smaller total deviation and a shorter settling
time. As shown in Figure 6.5, superheat spray goes ahnost to zero momentarily. But it
recovers because of the feedback loop which altersfiringand feed water. There are more
oscillations in spray in case of controller 2 as compared to controller 1. Figure 6.4 shows
the performance ofreheated steam temperature. Controller 1 and controller 2 do equally
well in this case. As expected, fuel as well as feed water go up to generate the necessary
power in this case (Figure 6.6 and 6.7).
Figures 6.8 through 6.11 compare the performance of both controllers for power
output, throttle pressure, superheated steam temperature, reheated steam temperature,
respectively, for test number 2. Figures 6.12, 6.13 and 6.14 show the superheat spray.

74

fiiel and feed water given by both the controllers for this test. At 7000 seconds, the load
demand was ramped up from 420 megawatts to 490 megawatts at 25 megawatts/min. It
was held at 490 megawatts till 8000 seconds at which point the load was ramped down at
25 megawatts/min. until it reaches 340 megawatts, where it was maintained for the
duration of the test.
As shown in Figure 6.8, both controllers perform equally well for power output.
The plots are overiapping. Similarly, in the case of throttle pressure in Figure 6.9, both
controllers perform very good. Figure 6.10 shows the behavior of the superheated steam
temperature for this test. In the case of controller 2, steam temperature settles to its
setpoint very fast and shows three times lower total variability as compared to controller
1. Figure 6.12 shows that both controllers are trying to keep the spray in the mid range.
As expected, fuel and feed water go up initially to bring the power up to 490, and then go
down simultaneously to take the power down to 340 (Figures 6.13, 6.14). Controller 2
shows a slightly lower variability as compared to controller 1 in case of reheat steam
temperature (Figure 6.11).
Figures 6.15 through 6.18 compare the performance of both the controllers for
power output, throttle pressure, superheated steam temperature and reheated steam
temperature, respectively, for test number 3. Figures 6.19, 6.20 and 6.21 show the
superheat spray fiiel and feed water, respectively, given by both controllers for this test.
At 7000 seconds, the load demand was ramped down from 420 megawatts to 340
megawatts at 25 megawatts/min.
As shown in Figure 6.15, both controllers perform equally well for power output.
The plots are actually overlapping.

Figure 6.16 shows the performance of throttle

pressure for this test. As power output decreases, pressure goes up initially, but recovers
back to its setpoint. Controller 1 as well as controller 2 show an equal performance here.

75

ipl^pp

Figure 6.17 shows the performance of the superheated steam temperature for this test.
Note a slightly longer settling time shown for the steam temperature using controller 1 as
compared to controller 2. When power output starts decreasing initially, the firing rate
also has to decrease, but it takes a short time to do so. Thus, temperature increases
initially, but it is brought back to its setpoint. Figure 6.19 shows that superheat spray
given by controller 2 is slightly smoother than controller 1. Both controllers are able to
keep the spray in the mid range. Reheated steam temperature (Figure 6.18) essentially
shows equal performance for both controllers.
Figures 6.22 through 6.25 compare the performance of both controllers for power
output, throttle pressure, superheated steam temperature and reheated steam temperature,
respectively, for test number 4. Figures 6.26, 6.27 and 6.28 show the superheat spray,
fiiel and feed water, respectively, given by both controllers for this test. At 7000 seconds,
the load demand was stepped upfrom420 megawatts to 460 megawatts instantaneously.
As shown in Figure 6.22, both controllers performed equally well for power
output.

The plots are overlapping.

Figure 6.23 shows the performance of throttle

pressure for this test. As the power output increases, pressure goes down initially, but
recovers back quickly to its setpoint. Controller 1 as well as controller 2 show an equal
performance here. The kind of initial sagging shown by the pressure is very close to the
one observed industrially (Crow, 1995c). Figure 6.24 shows the performance of the
superheated steam temperature for this test. When the load demand increases, the firing
rate also has to increase, but it takes a little longer to do so. Thus, the temperature
decreases initially, but it should come back to its setpoint.

The initial drop in the

temperature shown by controller 2 is slightly smaller as compared to controller 1.


Controller 1 shows a little more oscillations as compared to controller 2. Reheated steam
temperature (Figure 6.25) essentially shows an equal performance in the case of controller

76

1 as well as controller 2. The superheat spray given by both the controllers is essentially
showing the same behavior (Figure 6.26). Both of them are oscillating. As expected fiiel
and feed water have gone up to generate the necessary power (Figures 6.27, 6.28).
Figures 6.29 through 6.32 compare the performance of both controllers for power
output, throttle pressure, superheated steam temperature and reheated steam temperature,
respectively, for test number 5. Figures 6.33, 6.34 and 6.35 show the superheat spray,
fiiel and feed water, respectively, given by both controllers for this test. At 7000 seconds,
the load demand was stepped down from 420 megawatts to 380 megawatts
instantaneously.
As shown in Figure 6.29, both the controllers perform equally well for power
output. The plots are again overlapping in this case. Figure 6.30 shows the performance
of the throttle pressure for this test. As the power output decreases, pressure goes up
initially, but recovers back to its setpoint very fast. Controller 1 as well as controller 2
show an equal performance here. The kind of initial overshoot shown by the pressure is
very close to the one observed industrially (Crow, 1995c).

Figure 6.31 shows the

performance of the superheated steam temperature for this test. When the load demand
decreases, the firing rate also has to decrease, but it takes a little longer to do so. Thus,
temperature increases initially, but it is brought back to its setpoint after a long time in the
case of controller 1. Controller 2 is able to adapt to these changes quickly and takes rapid
actions.

Thus, it brings the temperature back to its setpoint faster as compared to

controller 1. Controller 1 shows more oscillations as compared to controller 2. Figure


6.33 shows that superheat spray given by controller 2 is slightly smoother as compared to
controller 1. Both controllers are able to keep the spray in the mid range. The reheated
steam temperature (Figure 6.32) essentially shows an equal performance for controUer 1
as well as controller 2.

77

Figures 6.36 through 6.39 compare the performance of the controllers for power
output, throttle pressure, superheated steam temperature and reheated steam temperature,
respectively, for test number 6. Figures 6.40, 6.41 and 6.42 show the superheat spray,
fiiel and feed water, respectively, given by both controllers for this test. At 7000 seconds,
the lower set of bumers are removed.
As shown in Figures 6.36 and 6.37, both the controllers perform equally well for
power output as well as throttle pressure, respectively.

Figure 6.38 shows the

performance of the superheated steam temperature for this test. Note the higher initial
variability and settling time shown by temperature in case of controller 1. Controller 2 is
able to adapt to the process change quickly than controller 1. But controller 2 shows
bigger oscillations as compared to controller 1. Figure 6.40 shows a slightly smoother
superheat spray given by controller 2 as compared to controller 1. The reheated steam
temperature (Figure 6.39) essentially shows an equal performance for controller 1 as well
as controller 2.
Figures 6.43 through 6.46 compare the performance of both controllers for power
output, throttle pressure, superheated steam temperature and reheated steam temperature,
respectively, for test number 7. Figures 6.47, 6.48 and 6.49 show the superheat spray,
fiiel and feed water, respectively, given by both controllers for this test.
As shown in Figures 6.43 and 6.44, both controllers perform equally well for
power output as well as throttle pressure, respectively.

Figure 6.45 shows the

performance of the superheated steam temperature for this test. Note the significantly
higher initial variability and a higher settling time shown by temperature for controller 1 as
compared to controller 2. The soot blowing increases the gain of the process. Controller
2 is able to adapt to these changes better than controller 1.

The superheat spray is

brought back in the mid range by the feedback loop for both controllers. After that time.

78

w^mmrfi^

I III! . > ! ' J ( 3

the temperature cycles a little in the case of both controllers. This can be attributed to the
slight cycling of fiiel and feed water (Figures 6.48, 6.49). The reheated steam temperature
(Figure 6.46) essentially shows an equal performance in the case of controller 1 as well as
controller 2.
Figures 6.50 through 6.53 compare the performance of both controllers for power
output, throttle pressure, superheated steam temperature and reheated steam temperature,
respectively, for test number 8. Figures 6.54, 6.55 and 6.56 show the superheat spray,
fiiel and feed water, respectively, given by both controllers for this test.
As shown in Figure 6.50, both controllers perform equally well for power output.
Both plots are overlapping.

In Figures 6.51 and 6.52, throttle pressure and the

superheated steam temperature show an equal performance.

The reheated steam

temperature (Figure 6.53) essentially shows an equal performance for controller 1 as well
as controller 2.
Figures 6.57 through 6.60 compare the performance of both controllers for power
output, throttle pressure, superheated steam temperature and reheated steam temperature,
respectively for test number 9. Figures 6.61, 6.62 and 6.63 show the superheat spray, fuel
and feed water, respectively, given by both controllers for this test.
As shown in Figure 6.57, both controllers perform equally well for power output.
In Figure 6.59, controller 2 is able to adapt to the process change much better than
controller 1. Note the significantly higher settling time and cycling shown by controller 1
as compared to controller 2. As shown in Figure 6.61, both controllers are trying to bring
the spray back to the mid range. Controller 2 shows a smoother response. Controller 2
clearly outperforms controller 1 in this case. In Figure 6.60, reheated steam temperature
shows an equal performance for both controllers. In Figure 6.63, feed water shows a lot
of oscillations in case of controller 1 as compared to controller 2.

79

Figures 6.64 through 6.67 compare the performance of both controllers for power
output, throttle pressure, superheated steam temperature and reheated steam temperature,
respectively, for test number 10. Figures 6.68, 6.69 and 6.70 show the superheat spray,
fiiel and feed water, respectively, given by both controllers for this test.
As shown in Figures 6.64, both controllers show an equal performance for power
output. In Figure 6.66, controller 2 is able to adapt to the process change much better
than controller 1.

Note the significantly higher variability, settling time shown by

controller 1 as compared controller 2. As shown in Figure 6.68, both controllers are


trying to bring the spray back to the mid range. Controller 2 shows a better performance.
Controller 2 clearly outperforms controller 1 in this case also. In Figure 6.67, the reheated
steam temperature shows an equal performance for both controllers.
Figures 6.71 through 6.74 compare the performance of both controllers for power
output, throttle pressure, superheated steam temperature and reheated steam temperature,
respectively, for test number 11. Figures 6.75, 6.76 and 6.77 show the superheat spray,
fuel and feed water, respectively, given by both the controllers for this test.
As shown in Figure 6.71, both controllers show an equal performance for power
output. In Figure 6.73, initially, temperature shows a lesser deviation and lower settling
time for controller 2 as compared to controller 1. Both sprays are brought back to the
mid range and settle around 8500 seconds (Figure 6.75). However, controller 2 shows
more oscillations as compared to controller 1. Reheated steam temperature (Figure 6.74)
shows essentially the same performance for controller 1 and controller 2.
Table 6.1 shows the values of ISE and lAE for controller 1 and controller 2 for
tests 1 through 11, observing superheated steam temperature. These ISE and lAE values
were calculated for power output as well as superheated steam temperature. However,
ISE and lAE values were very close for controller 1 and controller 2, for power output in

80

i^Wff<^'^T'

all of the tests. This is well justified since it was observed that the power output response
was almost identical for both controllers in all the tests. From Table 6.1, it can be seen
that lAE values for controller 2 are 2-5% lower than controller 1. In case of test 1 and 2,
the lAE values for controller 2 are ahnost 15-20% lower than controller 1. ISE values are
also lower for controller 2 than controller 1. Thus, in all of the tests described above,
controller 2, in general, shows a better or at least an equal performance as compared to
controller 1. Thus, it is advantageous to use nonlinear models for control purposes
because of their ability to adapt to process gain changes.

81

j,^|..i

J .^;ij"*|

'

" I'jaw

500
490
480 H
470
460
450440430-

420
410
7

10

11

12

time in seconds
(Thousands)
setpoint

Controller 1

Figure 6.1 Response ofpower output for test 1

82

^ ' "I'lrayifi

Controller 2

"IBP"

jte;

3600

'^

3550

0)

3500

""-e?V5S'--V5>WVVs/W

3450

8
9
10
time in seconds
(Thousands)
Controller 1

11

Controller 2

Figure 6.2 Response of throttle pressure for test 1

83

r^'^^

j..J3^^B^

12

iW"PjpPiif5??

990
989.5
989988.5Q)

988987.5-

E
CD

987
986.5
986
985.5
985

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

11

Controller 2

Figure 6.3 Response of superheated steam temperature for test 1

84

12

1040
1030
u.

1020-

1010

Q. 1000
E
CD

990
980
970
8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

11

Controller 2

Figure 6.4 Response ofreheated steam temperature for test 1

85

12

,<^../wg--"-*'

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.5 Superheat spray for test 1

86

U^iMPJtH

^ -. :wt! W ^ ' ^ - ^

250

o 200CD
CO

Si

CD

150-

100

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.6 Fuel for test 1

87

11

12

a^r^m

1000
o
CD

CD

0
CD

fi

900-

nrW^T"

-cv^

800

700-

600

8
9
10
time in seconds
(Thousands)

controller 1

controller 2

Figure 6.7 Feed water for test 1

88

^^-.Mtb

11

12

mmmmimim

mill,

4 * ^M.i^K- '-f

500
480
460
440
420
400
380
360
340
320

8
9
10
time In seconds
(Thousands)

setpoint

Controller 1

Figure 6.8 Response ofpower output for test 2

89

" " ^"W31

11

12

Controller 2

3600

CO

'^

3550

3450

8
9
10
time in seconds
(Thousands)
Controller 1

Controller 2

Figure 6.9 Response of throttle pressure for test 2

90

992

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

Controller 2

Figure 6.10 Response of superheated steam temperature for test 2

91

^ ^ " " ^ "

< "'1-'

1040
1030LL

1020

1010H
d. 1000H
E
CD

990 H
980-

970
9
10
time in seconds
(Thousands)
Controller 1

setpoint

Controller 2

Figure 6.11 Response ofreheated steam temperature for test 2

92

%*

.'^1

!W?;p*gglJi^'.':!--Z-:iZ

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.12 Superheat spray for test 2

93

250

200

CD

n
CD

150-

100

10

time in seconds

(Thousands)
controller 1

controller 2

Figure 6.13 Fuel for test 2

94

12

-iJ/j.sTT'

twr

1000
o
CD

900-

800
CO
D

0
0

700

600

10

time in seconds
(Thousands)

controller 1

controller 2

Figure 6.14 Feed water for test 2

95

.^^

'^^\

-IpWWipipppUllJiJiiuiJ, . ^ ^

-J

-j.iui.ijivMiv^

420
410
400
390
380
370
360
350
340 H
330

8
9
10
time in seconds
(Thousands)
Controller 1

setpoint

Figure 6.15 Response ofpower output for test 3

* *

.J^S

11

12

Controller 2

3600

CO

'^

3550

0
C/3

3500

CL

3450

8
9
10
time In seconds
(Thousands)
Controller 1

Controller 2

Figure 6.16 Response of throttle pressure for test 3

97

.-:*! JJI

mmmmmmm im i

.. "^" ;;;;;

990
989"l 988
D)
0
"D

.E 987 H
d.
0 986
985
984

12

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

Controller 2

Figure 6.17 Response of superheated steam temperature for test 3

98

HJg'J.'-IPl*
^^^^^ttmmm

WiMwa

'

"^^Mw.;*-^

^W^?^,|

1040
1030U-

1020

1010H

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

Controller 2

Figure 6.18 Response ofreheated steam temperature for test 3

99

'mm^^

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.19 Superheat spray for test 3

100

d.

11

12

BSBS-!

'/:!

250

o
0

200-

In

150

100

10

time in seconds
(Thousands)
controller 1

controller 2

Figure 6.20 Fuel for test 3

101

'%fif^^

.jy..rv:*^npftiji-w*>.w'5'"^^^'^^
,f>

1000

900

0
JQ

800
5
0
0

700-

600

8
9
10
time in seconds
(Thousands)

controller 1

controller 2

Figure 6.21 Feed water for test 3

102

.-5.

11

12

465
460
455
450445440435
430425420-.
415

^^^i^i^^"^

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

Figure 6.22 Response ofpower output for test 4

103

.&;

11

12

Controller 2

^--'jtfi^^^viassi^'^ssBsa^afss^^

3600

CO
C/3
Q.

3550

0
^

D
(f)
CO

3500

j'-uMui^iy^i'uwtjyv ^Vy^gyv

Q.

3450

10

11

time in seconds
(Thousands)
Controller 1

Controller 2

Figure 6.23 Response of throttle pressure for test 4

104

.14*

12

990
989
^

988H

.^^^>-----.*:?--'^i.Ht^^

. 987
d.
0

986985984

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

11

Controller 2

Figure 6.24 Response of superheated steam temperature for test 4

105

12

1040
1030u.

1020-

1010H

d. 1000H
E
0

990 H
980
970
8
9
10
time In seconds
(Thousands)
Controller 1

setpoint

11

Controller 2

Figure 6.25 Response ofreheated steam temperature for test 4

106

!i.

ar^l

12

1^1

ii

*"

"

40
o
0

Q.
CO

30

20-

Q.

wwflWM

10

D
CO

10

11

time in seconds
(Thousands)
controller 1

controller 2

Figure 6.26 Superheat spray for test 4

107

rt-

''^^^'Wfi&

12

250

200

lo
0

150

100

10

time in seconds
(Thousands)
controller 1

controller 2

Figure 6.27 Fuel for test 4

108

--'1K:

1000
o
0

900-

To

U^ U

Ol

) U " U ' u A u ' I iDLiiiui

~iininui>n,s_,iiY^Myt

800
CO

5
0
0

700-

600

8
9
10
time in seconds
(Thousands)

controller 1

controller 2

Figure 6.28 Feed water for test 4

109

it^-i.

11

.^SBB^^vl

12

mm

420
415
410H
405
400-

395-

390
385
380
375

"T"

setpoint

8
9
10
time in seconds
(Thousands)
Controller 1

Figure 6.29 Response ofpower output for test 5

no

11

12

Controller 2

^fnvfXMmtsz-

*jSf*C*:v---

3600

CO

'^ 3550
0
CO

3500

Q.

3450
8

10

time in seconds
(Thousands)
Controller 1

Controller 2

Figure 6.30 Response of throttle pressure for test 5

111

'

-:^jsteMBvt:--

;-:rtsr<M(Sj;

990
989
9880

.E 9870

986
985
984
8
9
10
time in seconds
(Thousands)
Controller 1

setpoint

Controller 2

Figure 6.31 Response of superheated steam temperature for test 5

112

-4*li

C-'ff'i-c^a-v--*

*.''

1040
1030
u.

1020-

1010

d. 1000
E
0

990
980970
8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

Controller 2

Figure 6.32 Response ofreheated steam temperature for test 5

113

40
o
0

lo

30-

I 20
CO

1^^**
'*=^

-^f^-HWHr-

u ' ' ^''j^

s
0

"i 10
Q.
D
CO

10

11

time in seconds
(Thousands)
controller 1

controller 2

Figure 6.33 Superheat spray for test 5

114

12

250

200

To
0

150-

100

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.34 Fuel for test 5

115

^m:4

11

12

1000

o
0

900-

To
n

800
5
0
0

^-*yV

AA^/A^^/^

. , ^ ^ . A

. . r..

M. . A s ^ ^ ^ iV. > ^ ^ .

700-

600

"T"

8
9
10
time in seconds
(Thousands)

controller 1

11

12

controller 2

Figure 6.35 Feed water for test 5

116

a^^B*

430

425

420-

415

410

8
9
10
time in seconds
(Thousands)
Controller 1

setpoint

11

Controller 2

Figure 6.36 Response ofpower output for test 6

117

r'-^JUU'.M^iilJ

12

3600

CO
CO
Q.

3550

c
0
D
CO
CO
0
Q.
^

3500

3450
8
9
10
time in seconds
(Thousands)
Controller 1

11

Controller 2

Figure 6.37 Response of throttle pressure for test 6

118

12

lill

^^'^^^

988LL.

V y\V^

/^

"^^^"^^"^ \/^^ \/^

-\/

\/^

\/

987-

|-

986-

985-

984^

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

11

Controller 2

Figure 6.38 Response of superheated steam temperature for test 6

119

. -issisa

.^a

12

<^^^^^^^rm

1040
1030-

li.

1020-

1010H

10

11

12

time in seconds
(Thousands)
setpoint

Controller 1

Controller 2

Figure 6.39 Response ofreheated steam temperature for test 6

120

i>i>iB."i'

40
o
0

To

Q.
CO

30

20
n

- I

Jo

10-

CO

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.40 Superheat spray for test 6

121

11

12

185

o
0

To
^ 175
0

165

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.41 Fuel for test 6

122

11

12

850

o
0

To

800

0
0

750

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.42 Feed water for test 6

123

'> (Min M r f i f l

ji T
i I itanWTirrr*

430

425-

420

415-

410

setpoint

8
9
10
time in seconds
(Thousands)
Controller 1

11

Controller 2

Figure 6.43 Response ofpower output for test 7

124

12

yii^&mifi

"^asms^m

3600

CO
CO

Q.

3550

c
0
i
D

CO
CO

0
Q.

3500

3450

8
9
10
time in seconds
(Thousands)
Controller 1

11

Controller 2

Figure 6.44 Response of throttle pressure for test 7

125

12

K-J*'t
- 1 ^ *

990
989
988
0
D

.E 987 H
d.
0 986
985984

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

11

Controller 2

Figure 6.45 Response of superheated steam temperature for test 7

126

12

/J

1030-

LL

1020-

D)
1

8 1010_C

d. 1000E
0

990-

'V^

980970-

setpoint

8
9
10
time in seconds
(Thousands)
Controller 1

11

Controller 2

Figure 6.46 Response ofreheated steam temperature for test 7

127

12

SSORSBBSSSSSH^^vi^:^

8
9
10
time in seconds
(Thousands)
controller 2

controller 1

Figure 6.47 Superheat spray for test 7

128

12

~.J3'.'-7

185

o
0

CO

To

175

165

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.48 Fuel for test 7

129

il W lrf lrf"l

850

o
0

CO

To

^
0

800^^^

5
0
0

750

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.49 Feed water for test 7

130

11

12

r"'

^r^.-^. .

B 8 M i I III

.^

430

425

420-

415-

410

"T"

setpoint

8
9
10
time in seconds
(Thousands)
Controller 1

11

Controller 2

Figure 6.50 Response ofpower output for test 8

131

^mgmmm^m

12

II

3600

CO
CO
Q.

3550

c
0
^
D
CO
CO
0
Q.

3500

3450

I ^^^^^^^*^"*^**

"''*^i^'''^c;:?^'^^^''^^*\.^>=^Vj^^^^

8
9
10
time in seconds
(Thousands)

Controller 1

11

Controller 2

Figure 6.51 Response of throttle pressure for test 8

132

r-^-

TI

12

990
989T;

9S8

0
D

. 987
Q.

986
985984

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

11

Controller 2

Figure 6.52 Response of superheated steam temperature for test 8

133

12

1040
1030
u.

1020-

1010H

8
9
10
time In seconds
(Thousands)
setpoint

Controller 1

11

Controller 2

Figure 6.53 Response ofreheated steam temperature for test 8

134

12

40
o
0

To
n

30-

I 20
CO

CO
0

Q.

10

D
CO

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.54 Superheat spray for test 8

135

"^ISsqwWMsfi^S**?:***

11

12

185

o
0

CO

To

175-

165

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.55 Fuel for test 8

136

11

12

850

o
0

To
n

800

0
0

750

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.56 Feed water for test 8

137

12

.1

430

425

420T^Tv^f^^'^r-y'^rV'^v'

415

410

8
9
10
time in seconds
(Thousands)
Controller 1

setpoint

11

Controller 2

Figure 6.57 Response ofpower output for test 9

138

s^Ws

K-H'SJ*^

mm)^.:yMf-

12

W/mmmmmmem

3600

CO
CO

CI

3550

c
0
u.
D

CO
CO

0
Q.

3500

3450
8
9
10
time in seconds
(Thousands)
Controller 1

Controller 2

Figure 6.58 Response of throttle pressure for test 9

139

i^S5^gaiDlL:.-i..ii'.i

C.^

990

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

Controller 2

Figure 6.59 Response of superheated steam temperature for test 9

140

- ^ f : W 9 * ^ ^ ^ E ^ i ^ ^

tmmmm*

j ^ i | y r'lT f i -

JTOT!^

1040
1030u.

1020-

1010

d. 1000E
0

*-

990
980
970

8
9
10
time in seconds
(Thousands)
Controller 1

setpoint

11

12

Controller 2

Figure 6.60 Response ofreheated steam temperature for test 9

141

.---*H

'T

Vt?1

o
0

To
n

Q.
CO

to
0
0
Q.
D
CO

8
9
10
time in seconds
(Thousands)

11

controller 2

controller 1

Figure 6.61 Superheat spray for test 9

142

12

-'iVs'.,ff|

mm

185

CO

To
n=

175-

165

"T"

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.62 Fuel for test 9

143

11

12

850
o
0

CO

To
^
to

800 H

0
0

750

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.63 Feed water for test 9

144

11

12

BH

430

425

420

415

410
8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

11

Controller 2

Figure 6.64 Response ofpower output for test 10

145

12

,.^li>*M' I

li*'..K^1

3600

CO
CO
Q.

3550

c
0
imm
D
CO
CO
0
Q.

3500

3450

- I

8
9
10
time in seconds
(Thousands)
Controller 1

11

Controller 2

Figure 6.65 Response of throttle pressure for test 10

146

12

M^Mpirl

MMM

990

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

12

Controller 2

Figure 6.66 Response of superheated steam temperature for test 10

147

t . ViJ '." .1 I I

.Mn

1040
1030H
1020
g

1010H

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

11

12

Controller 2

Figure 6.67 Response ofreheated steam temperature for test 10

148

"^PMC* **"

' **

il<l'H'i-ll'tJ

40
o
0

CO

To
n

30-

I 20
CO

CO
0
^.

Q.

10

CO

8
9
10
time in seconds
(Thousands)
controller 1

11

controller 2

Figure 6.68 Superheat spray for test 10

149

::;;j-

"'^ESS'

12

BB

185

o
0

CO

To
n
=

175-

165

8
9
10
time in seconds
(Thousands)
controller 1

11

controller 2

Figure 6.69 Fuel for test 10

150

..i'.t:

IH^^^FJ

12

F.V?^

850
o
0

To
^
0

800
^^^

to
D

0
0

750

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.70 Feed water for test 10

151

i-t

11

12

-W9

430

425-

420

415

410

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

11

Controller 2

Figure 6.71 Response ofpower output for test 11

152

:.m

12

mtmummm

am

au

t. s ^

3600

3450
8
9
10
time in seconds
(Thousands)
Controller 1

Controller 2

Figure 6.72 Response of throttle pressure for test 11

153

12

BiB r i ^ m n

?5^

990
989T ; 988H
0
D

. 987
d.
0

986
985984

8
9
10
time in seconds
(Thousands)
setpoint

Controller 1

11

12

Controller 2

Figure 6.73 Response of superheated steam temperature for test 11

154

jssmm

1040
1030u.

1020

1010H

8
9
10
time in seconds
(Thousands)
Controller 1

setpoint

11

12

Controller 2

Figure 6.74 Response ofreheated steam temperature for test 11

155

.^^it-

MMP

rws

O
0

To

CO

to
0

0
Q.
D
CO

8
9
10
time in seconds
(Thousands)
controller 2

controller 1

Figure 6.75 Superheat spray for test 11

156

12

mssss^s^

^i

185

o
0

CO

To

175-

165

8
9
10
time in seconds
(Thousands)
controller 1

11

controller 2

Figure 6.76 Fuel for test 11

157

*?:..>*'";>''"'''

12

niiifiiiriin T

itv*nsszaBCusw

--n

850

o
0

CO

To

800

\jif^t^^^\2^^

to
"D
0
0

750

8
9
10
time in seconds
(Thousands)
controller 1

controller 2

Figure 6.77 Feed water for test 11

158

Tr^^

11

12

l^;2ciii^SiiggiSSiiiiimimtti.i^m^tg'mi

Time

Time

Figure 6.78 A typical ATV test

159

^.- ^ 1

Table 6.1 Measure of goodness of control


Test#
1
2
3
4
5
6
7
8
9
10
11

ISE for
controller 1
8.1716e6
8.6430e6
7.9373e6
7.8961e6
7.9184e6
7.9259e6
8.0280e6
7.8924e6
7.9224e6
7.9428e6
7.9132e6

ISE for
controller 2
7.9000e6
7.9689e6
7.9148e6
7.8914e6
7.9103e6
7.8932e6
7.9049e6
7.8936e6
7.8909e6
7.8880e6
7.8942e6

lAEfor
controller 1
1.4198e6
1.5880e6
1.2933e6
1.2427e6
1.2729e6
1.2677e6
1.3394e6
1.2361e6
1.2750e6
1.2664e6
1.2594e6

160

KSCj^WWtWK'i

lAEfor
controller 2
1.2360e6
1.3020e6
1.2521e6
1.2190e6
1.2365e6
1.2405e6
1.2465e6
1.2319e6
1.2106e6
1.2059e6
1.2344e6

CHAPTER v n
CONCLUSIONS AND RECOMMENDATIONS

Conclusions

This project had various stages. The results achieved in these various stages are
stated as conclusions ofthis project.
1.

A detailed dynamic simulator of a supercritical coalfiredboiler was developed and


bench-marked against the steady state plant data provided by TU Electric, Dallas.
The simulator was bench marked at various operating loads. Thefinalsteady state
values of the parameters like temperature, pressure, steam flow, etc., were found
to be very close to the steady state plant data.

2.

The process gains showed that the system under consideration is quite linear in
most of the loops. Only the high pressure steam temperature loops showed
nonlinearity.

3.

A coordinated controUer was developed using the conventional controllers in all


the loops.

4.

The approximate models based on the energy balance relationships were written,
and a Generic Model Controller (Nonlinear Process Model Based Controller) was
implemented for the steam temperature control.

5.

The coordinated control system having nonlinear models in certain loops, showed
a better or at least an equivalent performance for steam temperature and throttle
pressure control than the coordinated control system having conventional
controllers in all the loops. It is advantageous to use nonlmear models for control
purposes because of their ability to adapt to process gain changes.

161

1 "

1 ^ ^ ^ * ^ > . - > - - g ^ . . -

Recommendations
Some of the fiiture work that can be done in this area is recommended as under.
1.

An experimental demonstration or an actual utility plant application of nonlinear


models would prove the authenticity ofthis work.

2.

Applications of other advanced techniques such as neural networks can provide


some advantages. The nonlinear approximate models which are used are
phenomenological models. They may not be able to take into account all the
disturbances that may hit the system. In that respect, neural networks may provide
some advantages if properly trained.

162

M4^-><S

^Jfa.i<l>w<MgMMIiaaMh**.

^ t'VnMtdBB^Ei

BiM

BIBLIOGRAPHY

Astrom, K. J.; and Hagglund, T., Proceedings of the IF AC conference, San Francisco,
1983.
Babcock, G. H.; and Wilcox, S., Steam-its generation and use. Babcock and Wilcox
Company, New York, 1972.
Bauer, F. W.; Nayak, R. V.; and Kohanik, W., "Boiler Modeling for performance
improvement," Presented at the joint ASME/IEEE power generation conference,
Dallas, TX, October 22-26, pp. 1-8, 1989.
Bequette, B. W., "A one-step-ahead approach to nonlinear process control," Proc. 1989
National ISA meeting, Philadelphia, PA, 1989.
Bird, R. B.; Stewart, E. W.; and Lightfoot, N. E., Transport Phenomena. John Wiley &
Sons Inc., NY, 1960.
Buchanan, D. J., Personal Communication, Southwestem Public Service Company,
AmariUo, TX, 79170, 1995.
Calle, T., Personal Communication, Texas Utilities Electric, Dallas, TX, 75201, 1995.
Chawdhry, P. K.; Hogg, B. W., "Identification of boiler models," lEE Proceedings, vol.
136, no. 5, September 1989, 261-271.
Collier, J. G., "The design of boilers," Heat exch: Therm. Hvdrauli. Fundam. Des.. 1980,
619-646.
Crow, D., Personal Communication, Texas Utilities Electric, Dallas, TX, 75201, 1995a.
Crow, D., Personal Communication, Texas Utilities Electric, Dallas, TX, 75201, 1995b.
Crow, D., Personal Communication, Texas Utilities Electric, Dallas, TX, 75201, 1995c.
Dattatreya, S., "Adaptive gain improves coalfiredboiler control," Hydrocarbon
Processing. 1982, 261-263.
Dolezal, Richard, Large Boiler Fumaces Theorv. Constmction & Control. Elsevier
Publishing Company, New York, 1967.

163

-.^ I

fK

si,f^.^aiBSTsr

r-ifftiTi V? ^rSr

Dukelow, Sam G., Improving boiler efficiency. Kansas Cooperative Extension Service,
Manhattan, Kansas, 1981.
Dukelow, Sam G., The Control of Boilers, second edition. Instrument Society of
America, 1991.
Economou, G. G.; Morari, M.; and Palsson, B. O., "Intemal model control. F. Extension
to nonlinear systems," Ind. Engng. Chem. Process Des. Dev.. 1986, 403-43.
Edgar, T. F.; and Himmelblau, D. M., Optimization of chemical processes. McGraw-Hill,
New York, 1988.
Fujii, T.; Ohta, J.; Ziang, A. Z.; Kijima, K.; and Kinka, T., "Dynamic Characteristics of
boilers in combined cogeneration plants," Chemical Engineering. 1990, 561-572.
Garcia-Barras, Thomas, Manual for improving Boiler & fiimace performance. Gulf
PubUshing Company, Houston, TX, 1983.
Garrett, C. J., "Mountain Creek Unit No. 8 control system design emphasizes
Dependability and Safety," Presented at the ISA 1967 Power Instmmentation
Symposium, Dallas, TX, 1967.
Ha, I. J.; and Gilbert, E. G., "A complete characterization of decoupling control laws for a
general class of nonlinear systems," TFF.F Tras. Auto. Control. 1986, 32, 763.
Henson, A. M.; and Sd)org, D. E., "Extension of nonlinear coupling methods to include
feedback linearization," Paper 1444A, presented at the 1989 AICHE meeting, San
Francisco, CA, 1989.
Isodori, A.; Krener, A. J.; Gori-Giorgi, C; and Monaco, S., Nonlinear Decoupling via
Feedback: A Differential Geometric Approach, TFEF. Trans. Auto. Control
1981,26,331.
Kakac, S.; Bergles, A. E.; and Mayinger, F., Heat Exchangers: Thermal - Hydraulic
Fundamentals and Design. Hemisphere Publishing Corporation, New York, 1981.
Kantor, J. C , An overview of nonlinear geometric strategies for process control Shell
process control workshop, Butterworth, Stoneham, MA, 1987.
Kaya, A., "A critical review of boiler controls for improved efficiency," Energy Engng,
1990,36-51.

164

P l ^ ^

rSBSaHH^i^HK^ VH

Keenen, J. H.; Keyes, F. G..Thermodynamic Properties of Steam. John Wiley & Sons,
New York, 1947.
Kern, D. Q., Process Heat transfer. McGraw-Hill Book Company, Inc., New York, 1950.
Laubli, F.; Fenton, F. H., "Theflexibilityof the supercritical boiler as a partner in power
system design and operation - part I: Theoretical relationship," IEEE transaction
on power apparatus and systenLl971, 1719-1724.
Laubli, F.; and Fenton, F. H., "Theflexibilityof the supercritical boiler as a partner in
power system design and operation - part II: Application andfieldtest results,"
IEEE transaction on power apparatus and system^ 1971, 1725-1733.
Lee, P. L.; and SuUivan, G. R., "Generic Model Control" Computers Chem Engng.
1988, vol. 12, 573-583.
Luyben, W.L., Process modeling, simulation and control for chemical engineers. Second
edition, McGraw Hill Co., New York, 1989.
Mahuli, S. K.; Rhinehart R. R.; and Riggs, J. B., "Nonlinear Model Based In-line control
of wastewater pH: A Lab Study," ISA Transactions, 1993, 32, 241.
McAdams, W. H., Heat Transmission. McGraw-Hill Book Co., New York, 1954.
McDonald, J. P.; Kwatny, H. G.; and Spare, J. H., "Nonlinear model of a reheat boilerturbine generator system," Philadelphia Electric Company, Research Division
Report No. -196(1971).
Meyer, C. A.; McClintock, R. B.; Silvestri, G. J.; and Spencer, R C , Thermodynamic and
transport properties of steam. American Society of Mechanical Engineers, New
York, 1967.
Pandit, H. G.; Rhinehart, R. R., "Process Model Based Control of afluidizedBed
Reactor: A comparison of two strategies," Advances in Instmmentation ISA. 44,
1989, 703.
Pandit, H. G.; and Rhinehart, R. R., "Experimental demonstration of constrained process
model-based control of a nonideal binary distillation column," in Proc. Amer. Cont.
Conf, Chicago, IL, 1992, 630-631.
Parish, J. R.; and Brosilow, C. B., "Nonlinear inferential control" AICHE J. 1988, 633644.

165

f; m

r.i^

t",-.'?

Pamchuri, V. P.; and Rhinehart, R. R., "Experimental Demonstration of Nonlinear Model


Based Control of a Heat exchanger," 1994 ACC Conf Proceed., Baltimore, MD,
1994.
Patwardhan, A. A.; Rawlings, J. B.; and Edgar, T. F., "Nonlmear prediction control using
solution and optimization," presented at the 1988 AICHE National meeting,
Washington, DC, 1988.
Ramchandran, B.; Riggs, J. B.; Heichelheim, H. R., "Nonlinear Plant-wide Control:
Application to a Supercritical Fluid Extraction Process." Ind. Eng. Chem. Res..
1992,31,290.
Rhinehart, R. R.; and Choi, J. Y., "Process Model Based Control of wastewater pH
Neutralization," Advances in Instmmentation. ISA. 1988, 43, 351.
Rhinehart, R. R.; and Riggs, J. B., "Comparison between Two Nonlinear Process Model
Based Controllers," Comp. Chem. Engg.. 1990, 14, 075.
Riggs, J. B.; Curtner, K.; Foslien, W., "Comparison of two advanced steam temperature
controllers for coal-fired boilers," Comp. Chem. Engg.. 1995, Vol 19 (5), 541550.
Riggs, J. B., "Nonlinear Process Model Based Control of a Propylene Sidestream Draw
Column," Ind. Eng. Chem. Res.. 1990, 29, 2221.
Riggs, J. B., An Introduction to Numerical Methods for Chemical Engineers. Second
edition, Texas Tech University Press, Lubbock, TX, 1994.
Riggs, J. B.; and Rhinehart, R. R., "Comparison between process model-based
controllers," Proc. 1988 Amer. Control Conf Atlanta, GA, 1988.
Rohsenow, W. M.; Hartnett, J. P.; and Ganic, E. N., Handbook of heat transfer
fundamentals, second edition, McGraw Hill Inc., New York, 1985.
Sekoguchi, K.," Analyses andfieldtests of once through boiler dynamics," Bulletin of the
JSME. 1970, vol 13, no. 63, 1084-1095.
Sherwin, K.," Design of a Boiler," Intemational Joumal of Mechanical Engineering
Education, vol 17, no. 2, 1989, 107-116.
Subawalla, H.; and Rhinehart, R. R., "Experimental Comparison of Model Based and
Convention Control of a Plasma Reactor," 1994 ACC Conf Proceed., Balthnore,
MD, 1994.

166

::.vr,*j>;.

"iim^'mi^^m

Suzuki, Y.; Pak, P. S.; and Uchida, Y., "Simulation of a supercritical once-through
boUer," Simulation. 1979, vol. 33, no. 6, 181-193.
Suzuki, Y.; Pak, P. S.; and Uchida, Y., "Simulation of supercritical once through boiler,"
Proceedings Summer Computer Simulation Conference, 1977, 343-352.
Toijala, K.; and Fagervik, K., "A digital simulation study of two-point control of
distillation columns," Kem. Teollisuus. 1972, 29, 5.
Volkov, E. P.; Gusev, I. N.; and Zaichik, L. I., "Mathematical modeling of combustion in
fiimace chambers of coal-fired utility boilers," Izvestiya Rossiiskoi Akademii Nauk
Energetika. 1992, vol. 30, no. 2, 81-89.
Williams, G. L.; Rhinehart, R. R.; and Riggs, J. B., "Inline Process Model Based Control
of Wastewater pH using Dual Base Injection," Ind. Eng. Chem. Res.. 1990, 29,
1254.
Wilson, C. R., "Modem Boiler Economisers - Development and Applications," Heat
Recovery Systems. 1982, vol 2, no. 2, 209-225.
Ziegler, J. G.; Mchols, N. B., "Optimum settings for automatic controllers," Trans.
ASME. 1942, 64, 759.

167

-'I'lMiwirfininiifi I

i iiiilirrriTfi -

NJ

ieiffiHisattaii

.-,/

Вам также может понравиться