Sei sulla pagina 1di 18

Simulation of Model Predictive Control for a Coal Fired Power Plant

Authors: Phillip D. Schnelle Jr. - Principle Consultant, Process Control - E. I. DuPont de


Nemours Co. Inc. and Paul S. Fruehauf* - Principle Consultant, Fruehauf Engineering
Services, Inc.

Keywords

Simulation, Model Predictive Control, Boiler-Turbine Coordinated Control, Coal Fired Power
Plant

Abstract:

In the following, we present a control study preformed to compare Model Predictive Control
(MPC) to the more conventional Single Input Single Output (SISO - PID) as applied to a
dynamic simulation of a coal fired multiple-boiler/single-turbine power plant. Boiler-turbine
coordinated control is chosen because it is a challenging problem for SISO given the difficult
process dynamics (i.e., deadtime, nonminimum phase), multivariable nature and constraint
seeking control requirements. We also discuss the value of using this kind of simulation study to
prove control concepts and demonstrate operational and economic benefits. In addition, this
approach facilitates the training of plant staff (engineers, maintenance and operators) on the
MPC Human Machine Interface and MPC operation.

*Corresponding Author: mailto:fruehaufps@frue-engg-svcs.com; (302) 690-1929

1. Executive Summary

In this paper, we present the modeling and control of a coal fired power plant. We compare the
application of simple PID controls to advanced MPC control. The control objectives are: setpoint
control of boiler drum level, steam header pressure and turbine MW generation (set load) subject
to upsets to the steam pressure, electrical demand and the fuel delivery system. The control
objectives requires turbine load to set production with highest priority on high and low limit
control to maintain the drum level and steam pressure within bounds and if possible, adjust the
steam pressure to get the most efficient turbine operation.

The approach we use in this study is to model the boiler-turbine system using a first principles
dynamic simulation. This simulation includes all relevant upsets and disturbances so that the
control objective can be properly represented. We implement the SISO control design in the
control blocks provided in the simulation tools. We conduct the MPC control evaluation by
hooking the control simulation up to a commercial MPC package using an OPC connection. We
submit both control systems to the same set of upsets and we evaluate both control approaches
against an agreed upon control objective.

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 1
The study conclusions are that there are significant benefits to be obtained by using MPC for
coordinated boiler-turbine control. The model based control provides better regulation to target
because of the natural deadtime compensation and the model based feed forward action provided
by MPC. Better regulation allows the coordinated control to push and hold tighter to the upper
bounds of target limits which may allows for more efficient operation. The simulation and MPC
test environment also provide a good platform for engineer, operator and maintenance training so
that they can get some exposure and checkout on the new proposed control strategy and HMI
development environment.

2. Introduction

In this article, we use a dynamic simulation to compare the performance of different control
strategies for a process with complex dynamics. We assume that the underlying instrumentation
and regulatory control systems work well. This is an important and frequently over looked
assumption.

Our purpose is to suggest a best practice for evaluating the benefits of advanced levels of control.
There are various levels of advanced control, depending on the nature of the opportunity being
sought. In other words, the control objectives, the complexity of the process dynamics and the
magnitude of the savings often dictate the level and function of the advanced control required.
This is one reason that it is so critical to get the control objective well defined and agreed to
before any evaluation begins.

2.1 Important Questions Regarding MPC

When first considering a control improvement project, a control or process engineer may have
many questions. What possible benefits can be expected and realized by using advanced control
methods? How much will it cost? What type of advanced control should be used? How can this
opportunity be investigated in an inexpensive way? How can the engineering and operations staff
be trained without jeopardizing the process utility? These are all very legitimate questions that
process control engineers and operations managers worry about when considering if they should
get involved with an advanced control project.

One way to answer these questions is to use dynamic simulation to study the specific process
operation in question. Modern dynamic simulation tools are very powerful and can be used to
rapidly configure very a realistic dynamic representation of the process.

With the correct dynamic process simulation constructed and validated, the next questions
become: How should the simulation be used to select the best control solution for the problem at
hand? Is there a way to estimate the savings opportunity for this new control approach? How can
the developer maximize the chance of success? To answer these questions it is critical to address
the control objective including constraint and cost issues, the level of process control capability
on or off site and the training needs of the operating staff. All these things are important in
choosing the best control solution.

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 2
2.2 Literature Survey: Application of MPC to Boiler–Turbine Control

In a fairly recent review article, the authors state that advanced controls are not widely applied in
the power industry (Rees and Lu, 2002). The article addresses some of the reasons for this and
suggests how the situation might be improved.

A number of researchers have studied the application of advanced controls to the boiler-turbine
coordinated control problem. In two articles, the authors (Wen Tan, et. al. 2008 and Wen Tan, et.
al. 2004) present the use of linear control to a boiler-turbine unit. These authors discuss the
application of SISO controls with the addition of feedforward and decouplers. The successful
implementation of the control on an actual power plant is presented in the more recent paper.
Another interesting result is that a linear controller works well even though the process is very
nonlinear. This is due to fact that the process is typically operated along a sliding pressure curve
resulting in a process that is fairly linear in character.

An interesting approach to the boiler-turbine control problem is presented by (Li et. al. 2005).
These authors present the application of a supervisory algorithm that adaptively tunes the
classical SISO scheme. The SISO control strategy employs feedforward and decoupling. The
supervisory system uses identification and fuzzy logic auto tuning to determine the adaptive
tuning values that are sent to the regulatory control system. One obvious advantage of this
approach is the graceful degradation of the control should the supervisory level be temporally
unavailable.

A generalized predictive controller is applied to a boiler-turbine system (Rossiter, et. al. 1991).
In this earlier paper, the authors report that the predictive control provides better performance
than PID control; however, the improvement in control is reported to be modest.

In two papers, the authors report the application of DMC to power plants. The first authors,
(Aurora, et. al. 2004) present the application of this multivariable control technique to the
problem of steam temperature and pressure control. The second authors (Moon and Lee 2009)
apply DMC to the boiler-turbine process including drum level control. The authors design the
DMC controller using both step response data from a linear model and from actual process tests.
The process test data works significantly better. This is believed to be due to the high
nonlinearity of the drum level dynamics. The authors use the commonly studied Bell and
Alstrom (Bell and Alstrom, 1987) nonlinear model of a boiler-turbine system.

In two papers that we found particularly interesting, the authors discuss the application of
additional measurements for the multivariable boiler-turbine control problem. These additional
measurements offer the opportunity for improved control through feedforward and by providing
intermediate measurements that have less deadtime and less complex dynamics than the primary
measurements. Although MPC can do a better job of controlling processes with deadtimes and
loop interactions they cannot overcome the fundamental control limitation imposed by the

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 3
presence of deadtime. These additional measurements offer the possibility of achieving better
control by reducing this fundamental limitation. In the first paper, the authors (LU and Hogg,
1997) use the turbine first stage pressure. This measurement is a faster indicator of turbine power
than the megawatt output of the generator. The use of first stage turbine pressure is not new. It is
also presented in a popular book on boiler control (Dukelow 1991). The second improved
measurement is presented by the authors (Wang et. al. 2007) of the second publication. In this
article, the flame radiation intensity signal is used to improve the load following control of a coal
fired power plant. This measurement responds faster than steam pressure to changes in boiler
firing rate. We feel that future research work should investigate adding these measurements to
the MPC.

One of the economic justifications for improved boiler-turbine control is to improve the load
following capabilities of a coal fired power plant. A plant that has good load following capability
gives the power dispatcher more flexibility when power generation needs to be added or
removed from the power grid. Power that can be delivered with more flexibility is more valuable
than power sources that are based loaded. In one article, the authors (Stump and Williams 2009)
reports a 70% improvement in ramp rate using MPC on a coal fired power plant. This new
control application has significantly improved the plant revenue.

Simulations are useful in helping to develop and test advanced control strategies. In almost all
the papers we reviewed, the authors use simulations in their evaluations. In one interesting paper,
the author (Maffezzoni 1997) presents the differential equations for basic boiler dynamics. These
equations can be used to build first principle dynamic models of power plant boilers. Modern
methods for simulating power plants are reported in a fairly recent article by (Donne, et. al.
2001). In this article, the authors discuss power plant models implemented in ACSL and
Matlab/Simulink. In one paper already discussed, the authors (Wen Tan, et. al. 2008) use a
nonlinear power plant simulation implemented in Matlab/Simulink.

The application of MPC to boiler-turbine control has been fairly extensively studied over the last
25 years. Hopefully, the robust and proven MPC tools that we report in this paper will lead to
more application to commercial power units.

3. Background:

3.1 Process Description

The power plant that we feature in this paper is a coal fired power plant employing two
Circulating Fluidized Bed (CFB) boilers and a single three stage turbo-generator. The plant
configuration is illustrated in Figure 1. The dynamic simulation that we use in this investigation
is validated by dynamic tests that were done on this 500 megawatt power plant. The power plant
is highly heat integrated. The combustion air and boiler feed water are preheated in many stages
using waste heat streams.

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 4
Figure 1 – 500 MW Coal Fired Power Plant

3.1.1 Circulating Fluidized Bed Boilers

CFB boilers can cleanly combust a wide variety of fuels. The unit featured burns a low grade
coal called lignite. Lignite typically has a low heating value and contains a lot of moisture. The
combustion takes place in a fluidized bed where the fines that come off the top of the bed are fed
to cyclones which separate gas and solids. The solids are re-circulated back to the boiler and the
gases are exhausted to the atmosphere. One common property of these types of boilers is the
significant deadtime between an increase in fuel flow and the resulting increase in steam
pressure.

3.1.2 Turbine Generator

The turbine is a three stage reheat unit. The steam exhaust from the high pressure stage is fed
back to the boiler and reheated before injection into the intermediate pressure stage. The
discharge from the low pressure stage is fed to a condenser which is installed directly beneath
this stage. The turbine drives the electrical generator.

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 5
3.2 Process Control Objectives

The process control objectives for this process are:


 The megawatt output of the plant must be able to respond to the load change demands
from dispatch.
 The main steam pressure must be controlled within limits and, if needed, it should be able
to follow a sliding pressure curve (i.e., different setpoints as a function of plant
conditions).
 The above must be accomplished without severely upsetting other critical control loops
including the steam temperature, the boiler drum level and the boiler air/fuel ratio.

3.3 Classical Boiler-Turbine Controls (i.e., Unit Management)

Boiler-turbine controls are the top level controls for a utility power plant. These controls
determine the amount of power produced by the generating plant and match the boiler firing rate
to deliver that power level. For the purposes of this discussion, we will present this process as a
two by two multivariable system. There are two process variables and two manipulated
variables. The process variables are the Megawatt output of the plant and the main steam
pressure. The main steam pressure is the primary indicator of the energy balance between steam
production (i.e., the boiler) and consumption (i.e., the turbine). If the pressure is falling the boiler
is not producing enough steam and vice versa. The manipulative variables are the fuel feed rate
to the boiler and the turbine throttling valve position. The turbine throttling valve controls the
steam feed rate to the turbine. Figure 2 illustrates this system using a simple block diagram. It is
worth noting that the controls actually manipulate the boiler and turbine masters rather than
directly manipulating the fuel flow and the turbine throttling valve. In addition, there are multiple
turbine throttling valves.

Figure 2 – Two by Two Process

There are two possible SISO variable pairings. The two pairings are called turbine follow and
boiler follow. These terms are widely used in the power industry to describe unit management
controls. For example, the Emerson Ovation control system uses these terms to describe some of
the different operating modes for their boiler-turbine coordinated control package.

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 6
3.3.1 Turbine Follow Mode

The SISO pairing for turbine follow mode is where the steam pressure is controlled by
manipulating the turbine throttling valve and the megawatts is controlled by manipulating the
fuel feed rate to the boiler. This scheme is illustrated in Figure 3. The advantages of this scheme
are that it provides very good and responsive steam pressure control. This scheme will stabilize
the plant quickly from a severe upset. The disadvantage of this scheme is that it provides less
precise control of the plant megawatt output. Boiler upsets can cause long and large megawatt
deviations from setpoint.

Figure 3 - Turbine Follow Mode

3.3.2 Boiler Follow Mode

The SISO paring for boiler follow mode is where the steam pressure is controlled by
manipulating the fuel flow and the plant megawatt output is controlled by manipulating the
turbine valve. This pairing is illustrated in Figure 4. The main advantage of this scheme is that it
provides for precise control of megawatts. The main disadvantage is that this pairing is less
stable than the turbine follow pairing. It is not able to control through large disturbances and it
takes longer for the steam pressure to settle out after an upset.

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 7
Figure 4 - Boiler Follow Mode

3.3.3 Boiler-Turbine Coordinated Control

Boiler-turbine coordinated control combines the advantages of the two schemes discussed above.
This scheme is illustrated in Figure 5. This scheme is a multivariable approach using decoupling
and feed-forward actions to improve the overall control performance. It is reported, (Dukelow
1991) that load changes can be made twice as fast and more smoothly with this scheme. One of
our motivations for this investigation is to see if modern MPC technology can be employed to
achieve the benefits of coordinated control and take advantage of the superior tuning, ease of
implementation and long term maintenance properties of MPC. In addition, the dimensionality of
the MPC control structure can be easily increased to achieve better overall unit control.

Figure 5 – Power Plant Coordinated Control

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 8
4. Discussion

4.1 Dynamic Simulation (Control Model)

As we suggest in the above, a best practice for developing any advanced control approach is to
model the process in question and simulate different control strategies to evaluate which
approach can best achieve the control objective and realize the opportunity.

A schematic of the control model is illustrated in Figure 6. We model one boiler in detail, the
steam header and the turbine-generator. In addition, we include simple PID controls for drum
level, steam pressure and turbine-generator megawatts. The PID controls are in a boiler follow
configuration.

Figure 6 - Dynamic Simulation of Boiler-Turbine Process

We augment the model dynamics by using a deadtime transfer function in the steam pressure
control signal path. This deadtime was observed in dynamic test on the actual 500 megawatt
plant. This is a proven simulation technique that we use to validate a dynamic model so that it
more accurately predicts the behavior of the actual plant. Observed apparent deadtime can be due
transport delay as well as high order fast dynamic elements that normally are not built into a
control model. Accurate deadtime representation is critical to accurate control model predictions
because of the detrimental impact it has an on feedback control.

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 9
We can develop accurate control models quickly and easily with modern dynamic simulation
tools.

4.2 ARC vs. MPC

Possible Advanced Regulatory Control (ARC) control designs can be very complicated
depending on the control objectives. An ARC scheme can be designed to accomplish almost
everything a MPC controller will do using only the control elements found in the typical DCS.
This will result in a very complex control configuration.

In the following paragraphs, we describe the ARC elements that would be necessary to duplicate
the control functionality of a MPC applied to a boiler-turbine process.

In a hypothetical ARC scheme, the steam pressure would be controlled by adjusting the coal feed
rate. The signal would be compensated by a decoupling network to deal with the interactions
between movement in the boiler feed water. In addition, a Smith predictor would be used to
compensate for the deadtime introduced by the coal delivery system. The boiler drum level
would be controlled by the boiler feed water. The boiler drum level signal would be compensated
for fuel changes and shrink-swell effects by a three element circuit. Both pressure and level
controllers would be fed a dynamic feedforward signals to help them deal with rapid megawatt
demand changes. The megawatt load would be set by steam flow let down to the turbine.
Overrides would be required to protect the steam header pressure and drum level should they get
outside of bounds.

This hypothetical ARC structure is attempting to accomplish the control objective, but is this a
practical and maintainable control design? It is complicated and yet does not come close to
being as flexible and maintainable as the MPC we present in the next section. In addition, it has a
“locked in” philosophy. In other words, the one for one pairing assumed by this ARC design
limits what this controller can do. This strategy begs the questions: Could we keep this calibrated
and working? Would anyone really try to implement an ARC design this complicated? What is
minimum essential?

We did not attempt to simulate this design.

4.3 MPC Coordinated Control

MPC is a good fit for coordinated control because it is flexible and the control objective can be
elegantly mapped into the MPC structure. Multiple Manipulated Variables (MVs) are available
to deal with the interacting multivariable nature of the problem. The high, low and target
constraints are built in to the MPC control objective. An inherent part of the MPC structure is
that it automatically builds a controller that handles the difficult dynamics with implicit
decoupling, deadtime compensation and feedforward action.

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 10
4.3.1 MPC Model Development

We develop the MPC model in much the same way as one would develop it for a real process.
The dynamic simulation takes the place of the real process. The dynamic simulation is put
through a series of step tests. These step test results are shown in Figure 7.

Figure 7 - Step Tests

These tests are then used to build the linear transfer function matrix of models shown in Figure 8.

Figure 8 - Models for MPC

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 11
This testing and model development is, of course, easier and faster than doing the same thing on
an actual industrial process. The step tests are clean and free from signal noise. The step tests are
large and easily repeated if necessary.

The models resulting from step testing the simulation are very good because the step tests results
are clean and not influenced by other process disturbances. These models should result in good
control of the simulation. It also should be noted that these models make a great starting point for
development for a real world application. The real world step tests and models can be compared
to the simulation results. If there is good agreement this gives the developer more confidence that
the models are correct.

The development of the MPC controller is a matter of building a filter, translating the control
objective into the MPC objective and hooking it up to the process. This paper is not going to
discuss the details of the MPC development due to scope limitations.

The MPC control strategy can be inferred from Figure 9. Rather than the more traditional 2x2
boiler-turbine problem this controller structure also includes drum level and boiler feed water
flow to expand the system to a 3x3 system. This controller also includes 5 DV or feedforward
variables.

Figure 9 – MPC HMI

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 12
In this case, the steam pressure, drum level and turbine load controller are all set up with
adjustable external targets and high and low limits. The priorities of the Controlled Variables
(CVs) are set as wattage, level and pressure in that order. The limits are setup with much higher
priority than the targets. The MVs are configured with adjustable high and low limits and given
appropriate move concerns for their priorities.

This coordinated control is setup to match the given control objective of a load following system
with the best possible regulation to push the steam pressure as high as possible and maintain
drum level, all subject to the know disturbances.

4.4 MPC Benefits

4.4.1 Simulation Results

It should be noted that this comparison is not intended to be a rigorous comparison of different
control strategies but instead a quick and cost effective evaluation to determine if MPC offers the
potential for significant improvement.

Figure 10 shows simulation results that compare the boiler follow PI control strategy to the MPC
coordinated control strategy. The annotations on the figure help explain the comparison. For the
first 80 minutes the system is running under MPC control. As soon as the MPC is turned on
(annotation 1.), the controller pushes the pressure up to its upper bound as that is the desired
target for this production rate. At about 27 minutes (2.) the turbine demand is dropped to
400MW. The pressure barely moves and the level deviates, but not outside of limits due to good
regulation provided by feedforward and decoupling. Next, the turbine demand is returned to
600MW (3.) and the response is similar in this direction. There is an upset to the coal fuel supply
(4.) at about 60 minutes. Again the system responds well, increasing the fuel signal. At almost all
times the MPC is able to regulate the disturbances well and keep the pressure closer to the high
limit.

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 13
Figure 10 - Simulation Results Comparing Boiler Follow PI to MPC

Figure 10 also shows the system response for the boiler follow PI control strategy. At time ~ 80
minutes (5.) the model is restored to the original steady state and the PI control is turned on. The
target for the steam pressure is left to the operator to set. He or she typically sets it to keep the
pressure low enough so that the relief pressure is never exceeded which is a function of how well
the controller can maintain setpoint. At time ~110 minutes (6.), the turbine demand is dropped to
400MW. The pressure swing is such that the operator is glad that he or she put the setpoint at a
conservative low value (8.). The drum level swing is large and long due to the interaction of the
pressure and level loops. At ~ 170 minutes, the turbine load setpoint is set to 600MW (7.) and
the system responds in a fashion similar to the initial upset.

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 14
4.4.2 Proof of Concept

The simulation environment is a very powerful way to demonstrate proof of concept for an
advanced control strategy. It is sometimes difficult to demonstrate the benefits of advanced
control to operations and therefore get their buy in. They can be hesitant to changes things or
rock the boat. Being able to prove that the proposed control will work and result in better
performance can be extremely helpful.

4.4.3 Tighter Regulation

From Figure 10 it is clear that the MPC is achieving better or tighter regulation for the tested
upsets. MPC can achieve better regulation because of the internal model it uses for deadtime and
interaction compensation. The model based controls ability to look into the future and calculate
control moves based on where it is going versus where it is, is analogous to the driver of a car
trying to drive looking out the front window versus trying to drive by looking though a hole in
the floor. The better the regulation, the closer the variables can be pushed to constraints, the
more money can be made, not to mention the improvement in quality.

4.4.4 Staff Training

An additional benefit of having a dynamic simulation hooked up to a control system, like an


MPC, is that this system can be used for engineer, maintenance and operator training. A typical
MPC HMI is shown in Figure 9. This is a web based interface that is driving the dynamic
simulation. It can be run in real-time and can be made to look a lot like the real process by using
plant tag names, adding noise and drifting disturbances. Building realism into the simulation is a
great way to convince the people who have to use the system that it can do the job.

The engineering staff can use the system to try different control strategies for different control
objectives by changing what gets controlled to setpoint versus gaps, variable priority, variable
weighting and other tuning considerations. The operator can learn how to use the HMI and can
be put through different disturbance scenarios to test their response to disturbances that they do
not often see. The more comfortable the staff is and the more they see the system as something
that can help them, the more successful the application will be.

4.4.5 Savings Estimates Calculation

Another useful application of a simulation system is calculating the potential savings for an
advanced control application. Running the MPC controller in parallel with the simulation (or real
process) and having it calculate (and keep history) where it would potentially like to move the
process can be useful in calculating savings. To demonstrate this concept, Figures 11 and 12
shows the MPC run in open loop making predictions on where it would like to move the process.

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 15
Figure 11 - Running MPC Against Real Data to Calculate Savings

Figure 12 - Using Predictions to Calculate Savings

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 16
The process is being controlled using PI control. Figure 11 shows the MPC is “OFF” but making
predictions. It shows the suggested control moves and the would-be predicted results. Figure 12
shows the historical plot of the segment of time. The blown up chart shows the pressure response
and where the MPC would have controlled the pressure if it were making the moves. The
difference between the as controlled and the predicted result could be though of as the
opportunity for savings from improved control. If the area between the actual data and the
would-be predicted results were integrated and scaled to some economic value this would
represent the potential savings of this application.

5. Conclusions

In conclusion, there are significant benefits to be obtained by using MPC for coordinated boiler-
turbine control. The model based control provides better regulation to target because of the
natural deadtime compensation and model based feed forward action provided by MPC. Better
regulation allows the coordinated control to push and hold tighter to the upper bounds of target
limits which allows for more efficient operation. The simulation and MPC test environment
provides for an efficient means for evaluating and demonstrating the benefits of advanced
control. In addition, it provides a good platform for engineer, maintenance and operator training.

References:

(1) N. W. Rees and C. X. Lu 2002: Some Thoughts on the Advanced Control of Electric
Power Plants. Transactions of the Institute of Measurement and Control 24; 87

(2) Wen Tan, Jizhen Liu, Fang Fang, Yanqiao Chen 2004: Tuning of PID Controllers for
Boiler-Turbine Units. ISA Transactions 43, 2004 pgs. 571–583

(3) Wen Tan, Fang Fang, Liang Tian, Caifen Fu, Jizhen Liu 2008: Linear Control of a
Boiler–Turbine Unit: Analysis and design. ISA Transactions 47 (2008) 189–197

(4) Shaoyuan Li, Hongbo Liu, Wen-Jian Cai, Yeng-Chai Soh, and Li-Hua Xie 2005: A
New Coordinated Control Strategyfor Boiler-Turbine System of Coal-Fired Power Plant. IEEE
Transactions On Control Systems Technology, Vol. 13, No. 6, November 2005 943

(5) J.A. Rossiter, B. Kouvaritakis, R.M. Dunnett 1991: Application of Generalized


Predictive Control to a Boiler-Turbine Unit for Electricity Generation. IEE Proceedings-D, Vol.
138, No. I , January 1991

(6) C. Aurora1, L. Magni1, R. Scattolini1, P. Colombo, F. Pretolani and G. Villa


2004: Predictive Control of Thermal Power Plants. International Journal Of Robust And
Nonlinear Control Int. J. Robust Nonlinear Control 2004; 14:415–433

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 17
(7) Un-Chul Moon and Kwang. Y. Lee 2009: Step-Response Model Development for
Dynamic Matrix Control of a Drum-Type Boiler–Turbine System. IEEE Transactions On
Energy Conversion, Vol. 24, No. 2, June 2009 423

(8) R.D.Bell and K. J.Alstrom 1987: Dynamic Models for Boiler-Turbine-Alternator


Units: Data Logs and Parameter Estimation for a 160 MW Unit, Lund Inst.
Technol., Lund, Sweden, Rep. TFRT-3192, 1987.

(9) S. Lu and B. W. Hogg 1997: Predicitive Coordinated Control for Power-Plant Steam
Pressure and Power Output. Control Eng. Practice, Vol 5, No. 1. Pp 79-84, 1997

(10) S. G. Duckelow 1991: The Control of Boilers 2nd Edition. © 1991 by ISA

(11) Fei Wang,* Qunxing Huang, Dong Liu, Jianhua Yan, and Kefa Cen 2008:
Improvement of Load-Following Capacity Based on the Flame Radiation Intensity Signal in a
Power Plant, Energy & Fuels 2008, 22, 1731–1738

(12) Dana Stumpf and Jeff Williams 2009: Optimization Technology Contributes to Ramp
Rate Improvement. Power Engineering March, 2009

(13) C. Maffezzoni 1997: Boiler-Turbine Dynamics in Power-Plant Control. Control Eng.


Practice, Vol. 5 No. 3, pp 301-312, 1997

(14) M. S. Donne, A. W. Pike and R. Savvy 2001: Application of Modern Simulation and
Control. Computing & Control Engineering Journal April 2001

Distributed with permission of author(s) by ISA 2010


Presented at Mid Atlantic Section Conference; http://www.isa.org 18

Potrebbero piacerti anche