Sie sind auf Seite 1von 7

White Paper:

Advanced Multivariable Steam Temperature Control


On a 500 MW Coal-Fired Power Plant

Advanced Multivariable Steam Temperature Control on a 500 MW


Coal-Fired Power Plant

William L. Brown
Authors: Neal North Station
MidAmerican Energy Company

Cyrus W. Taft, P.E.


EPRI I&C Center

Jeffery J. Williams
Westinghouse Process Control
(Emerson Process Management Power & Water Solutions)

Introduction: As the electric power generation industry enters into a new deregulated era, many
plant owners are taking steps to prepare their plants for operating in the new
environment. Although no one can be certain what the future holds, the steps
being taken are usually directed at reducing operating and maintenance costs,
reducing fuel costs, or improving plant-operating flexibility. The George Neal
Station of MidAmerican Energy had this in mind when they initiated a project on
Unit 3 to improve the responsiveness of the unit to load dispatch signals.
Excessive steam temperature oscillations were limiting the unit response to about
1%/min. at low loads and 0.3%/min. at high loads. The unit manager felt that the
unit could do better with an improved control system. MidAmerican approached
EPRI and their control system vendor, Westinghouse Process Control, about
installing an advanced steam temperature control system to improve the
temperature response during load changes and thereby allow the unit to respond
faster on load regulation.

Plant The George Neal Plant is located along the Missouri River just south of Sioux City,
Iowa. The plant consists of two complexes about 1 mile apart. Units 1-3 are in the
Description: North complex and Unit 4 is in the South complex. Unit 3 is a 515 MW Foster-
Wheeler wall fired drum boiler and a GE tandem compound turbine. The unit has
six MPS-89 pulverizers supplying western coal to 24 burners on the front and rear
furnace walls. The superheat steam temperature is controlled by upper and lower
spray valves and the reheat steam temperature is controlled by spray valves and
by main and reheat pass dampers. The boiler control system is part distributed
digital control system (DCS) and part analog electronic control. The digital system
is a Westinghouse WDPF and is used on the feedwater, fuel and steam
temperature control. The analog part is a Bailey 820 system and is used on the
front end, air flow, and furnace pressure control.

Emerson Process Management


[0] Proprietary
Page 1

www.emersonprocess-powerwater.com
White Paper:
Advanced Multivariable Steam Temperature Control
On a 500 MW Coal-Fired Power Plant

Project Plan: The project had two objectives. The first was to improve the steam temperature
control and as a consequence improve the plant load following response. The
other objective was to document the amount of the improvement compared to
conventional PID-based control. An overall project plan was prepared which listed
the major tasks of the project and who was responsible for each task.

The major tasks were:

• Document the current system design and performance.


• Optimize the current system and test again.
• Design and install advanced steam temperature control system.
• Test and document advanced control system performance.
• Prepare report and paper.

Each of these tasks will be discussed in more detail below. One change was made
to the way the project was actually done. All the official testing was done at the
end of the project instead of with each task as originally planned.

Existing The existing system is a conventional PID based control system. The superheat
upper sprays were controlled using a cascade arrangement although only one
System outer controller was used with two inner controllers. The lower sprays were also
Assessment: controlled with a cascade arrangement with two outer and two inner controllers.
The pass dampers were controlled in a split range configuration with a single PID
controller on reheat outlet temperature. The reheat spray was a simple PID loop.
The only thing unusual about the existing system design is that no feedforward
signals were used on any of the steam temperature control loops. Figure 1 shows
the functional diagram for the upper spray control loop.

Figure 1:
Functional diagram
for the upper spray
control loop

Emerson Process Management


[0] Proprietary
Page 2

www.emersonprocess-powerwater.com
White Paper:
Advanced Multivariable Steam Temperature Control
On a 500 MW Coal-Fired Power Plant

The tuning settings of the existing PID controllers and other tunable blocks in the
loops were recorded before beginning the optimization process. Significant
changes were made to the superheat spray tuning parameters and the response
was improved considerably. Further investigation revealed that at one time the
tuning settings had been very close to the newly optimized settings, but they were
changed within the past year. Small changes were made to the pass damper
tuning settings and the reheat spray settings.

Model In order to design any advanced multivariable controller, a model of the process
must be developed first.
Identification There are several methods available for developing the model, but the most
Testing: practical one for an existing plant is usually model identification. In model
identification (or system identification), plant data for the controller inputs, outputs
and disturbance variables is collected while specific tests are run on the plant. The
tests are designed to excite all the plant characteristics that need to be modeled.
Typical tests include open loop step tests, pseudo-random binary sequence
(PRBS) tests, and frequency response tests. Closed loop setpoint step tests can
also be run if needed. If there are known disturbance effects that are important,
tests to introduce the disturbances should also be run.

For this project, open loop step tests and closed loop setpoint response tests were
conducted on all steam temperature control loops. In addition, load changes and
pulverizer startup and shutdown tests were run to gather disturbance effects. The
testing took about 5 days to complete. The DCS historian was used to record the
data from the tests. About 150 data points were collected for the tests. The
configuration of these test points in the historian was reviewed and in many cases
the deadbands for the collection were reduced to near zero for the tests. Model
identification software does not work well when excessive deadbands exist in the
data. High fidelity data is critical to identify accurate process models.

Once the model identification tests were completed, the data was extracted from
the historian and model identification was begun. Two different models were
required for this project because two different control techniques were being
utilized in the advanced system.

Controller The advanced control system designed for this project consisted of two separate
techniques. A model predictive control was used to adjust the outer controller
Design: setpoints and a dynamic feedforward system was used to dynamically
compensate the inner loop setpoints. Figure 1 shows a how the advanced control
system was interfaced with the conventional control system. The implementation
was similar for the other loops
.
The model predictive control system used the Aspen Target controller and the
dynamic feedforward controller used a system called TTFuzz. The Target
controller had 5 manipulated variables, 5 controlled variables and many

Emerson Process Management


[0] Proprietary
Page 3

www.emersonprocess-powerwater.com
White Paper:
Advanced Multivariable Steam Temperature Control
On a 500 MW Coal-Fired Power Plant

disturbance variables. The TTFuzz system had six feedforward outputs and about
40 disturbance inputs.

The Aspen Target controller uses a linear Partial Least Squares (PLS) model with
a non-linear correction factor (neural network). The resultant combination is
referred to as a hybrid model. This hybrid model overcomes extrapolation
problems with pure neural network based models. The neural network model has
two learning rules, Kalman and Variable Metric. The hidden units or layers of the
neural network model are also adjustable.

The TTFuzz feedforward control is characterized as a Takagi-Sugeno type fuzzy


model. These models have been designed in order to compensate for the
measured disturbances, which influence the process. Based on these disturbance
variables, a feedforward correction signal is developed to compensate the PID
loops for the disturbances. An accurate calculation of this correction requires an
accurate model of the process and disturbances effects. The feedforward concept
can be explained as follows: The model (transfer function) from the manipulated
variable to the controlled variable can be learned from testing as can the model
from the disturbance variable to the controlled variable. Since the current value of
the disturbance variable is measured, its effect on the controlled variable can be
determined. A correction to the manipulated variable can then be calculated which
will minimize the effect of the disturbance on the controlled variable.

Controller The two advanced controllers were implemented on a Sun Ultra workstation
(WEStation) directly connected to the WDPF data highway. Interface software
Implementation developed by Transition Technologies was used to exchange data between the
controllers and the conventional DCS logic. A new operator graphic was
developed to allow selection of the advanced control mode or the conventional
control mode. The graphic also allowed the operator to start and stop the
advanced control program if needed. The existing operator graphics for steam
temperature control were modified to show the current mode of control.

Since there are two manipulated variables for the reheat temperature control, the
TTFuzz controller must be designed to coordinate it outputs between the dampers
and the reheat spray valve. Since the use of reheat spray water has a negative
impact on plant heat rate, the system was designed to minimize spray flow when
possible. This has been accomplished by the following logic. If the spray
correction is negative and it does not fully close the spray, only the spray
correction is applied. If the spray correction is negative and it fully closes the
spray, it will be applied but the remaining correction will be applied to the damper
circuit. If the spray correction is positive (valve opening), than the damper
correction will be applied first (until the saturation limit), while the remaining effect
will be directed to the sprays.

Emerson Process Management


[0] Proprietary
Page 4

www.emersonprocess-powerwater.com
White Paper:
Advanced Multivariable Steam Temperature Control
On a 500 MW Coal-Fired Power Plant

Acceptance A detailed acceptance test plan was produced to make certain all parties' needs
were addressed by the tests and that the plant staff was aware of the testing
Testing: plans. Three types of tests were conducted with four different controller
configurations for a total of 12 tests. The types of tests were:

• Load ramp from full load (approximately 500 MW) to 400 MW and back to
full load.
• Load ramp from 380 MW to 280 MW and back to 380 MW.
• Pulverizer shutdown.

The controller configurations were:

• Old PID tuning settings and no advanced control in service.


• New PID tuning settings and no advanced control in service.
• New PID tuning with TTFuzz in service.
• New PID tuning with TTFuzz and Aspen Target in service.

On Neal 3 sootblowing is done continuously and is a major disturbance to steam


temperature control. In order to minimize the variability in the test results from
sootblower operation, the sootblower sequence was repeated for each type of
test. In other words, the same sootblowers were in service during each test so the
effect on the steam temperature should have been very similar.

Results: The WDPF Historical Storage and Retrieval (HSR) system was used to record the
results of the acceptance tests. Figure 2 shows the superheat outlet temperatures
from the high load response tests for 3 different control configurations. To quantify
the loop performance, the data for all the tests was analyzed for the average error,
the maximum error and the standard deviation of the error during the tests. There
was a considerable improvement from the old PID tuning to the new tuning which
was expected. There was also a considerable improvement from the new PID
tuning without TTFuzz to the new PID tuning with TTFuzz. And there was a
marginal improvement with the Aspen Target controller in service. Figure 3 shows
the loop error standard deviation results of the low load response tests for all 4
control configurations. The reheat temperature control data is somewhat
misleading because the unit can not achieve design reheat temperature under
most operating conditions. Figure 4 shows similar data for the Mill F shutdown
tests.

Emerson Process Management


[0] Proprietary
Page 5

www.emersonprocess-powerwater.com
White Paper:
Advanced Multivariable Steam Temperature Control
On a 500 MW Coal-Fired Power Plant

Figure 2:
Superheat outlet
temperatures from
high load response
tests for 3 different
control
configurations

Figure 3:
Loop error standard
deviation results of
the low load
response tests for
all 4 control
configurations

Emerson Process Management


[0] Proprietary
Page 6

www.emersonprocess-powerwater.com
White Paper:
Advanced Multivariable Steam Temperature Control
On a 500 MW Coal-Fired Power Plant

Figure 4:
Mill F shutdown
error standard
deviation

Conclusion: This project demonstrated that advanced control techniques can be used to
improve steam temperature control performance. One can see from the results
that there is some variability in the results but that in general the best overall
performance was achieved with the full advanced control system in service. One
should not draw too many conclusions about the relative merits of the two
advanced control techniques used on this project because the results are very
implementation specific. The important conclusion to be drawn from this project is
that this advanced control system was able to improve the steam temperature
response considerably over a conventional PID based control system.

Unfortunately, the improvement in the steam temperature control system did not
lead to better load response. The pressure control system in the old analog
electronic system is an older design and was not able to adequately control
pressure during faster load changes. However another benefit was realized from
the improved steam temperature control. The advanced control system was put in
service in late May and the unit did not experience any boiler tube leaks
throughout the summer. The unit manager is convinced that the improved steam
temperature control contributed greatly to this reduction in tube leaks.

Emerson Process Management


[0] Proprietary
Page 7

www.emersonprocess-powerwater.com

Das könnte Ihnen auch gefallen