COAL MILL AND COMBUSTION OPTIMIZATION ON A ONCE-THROUGH, SUPERCRITICAL BOILER WITH MULTIVARIABLE PREDICTIVE CONTROL

Size: px
Start display at page:

Download "COAL MILL AND COMBUSTION OPTIMIZATION ON A ONCE-THROUGH, SUPERCRITICAL BOILER WITH MULTIVARIABLE PREDICTIVE CONTROL"

Transcription

1 contents COAL MILL AND COMBUSTION OPTIMIZATION ON A ONCE-THROUGH, SUPERCRITICAL BOILER WITH MULTIVARIABLE PREDICTIVE CONTROL Steve Barnoski Project Manager Dayton Power & Light Donald Labbe Consulting Engineer Invensys Jim Graves Principal Engineer Invensys William Poe Business Development Manager Invensys KEYWORDS Model Predictive Control, Coal Mill Controls, Furnace Optimization, Advanced Process Control, Coal Grinding, Minimizing LOI, Efficient Power Generation, NO X Emissions, Multi-Variable Control, Pulverizer Control, Boiler Efficiency ABSTRACT This paper describes how multivariable predictive control techniques have been applied to optimize and improve the dynamic control of a 600 MW once-through, supercritical boiler at Dayton Power and Light s Stuart Station. Power plants have experienced significant heat rate improvement, NOx reduction and operational stability on a wide range of boilers including coal, gas and oil-fired through application of multivariable predictive control techniques. A variety of boiler sizes and configurations have shown high return on investments. This successful project will be discussed as well as lessons learned during the implementation of this technology. Some of the issues that this paper will address are: project development and justification, project implementation, auditing results and sustaining benefits. The variety of control and optimization issues that can be addressed and benefits derived from multivariable predictive control include NOx minimization, heat rate improvement, ramp rate improvement, pulverizer optimization, smart soot blowing, improved steam temperature control. Details of the Dayton Power and Light implementation will be given including some of the obstacles overcome in concluding a successful project. The role of controlling coal mills will be emphasized. The paper will include suggestions for maintaining the control and optimization solution to sustain maximum benefits.

2 Following the success of the initial installation on Unit #3, further projects are underway to apply the same advanced process control and optimization to Units #1, 2 and 4 (also 600MW each).

3 INTRODUCTION Dayton Power and Light (DP&L) operates boiler units at their J. M. Stuart Station near Aberdeen, Ohio. The units contain Babcock & Wilcox once through, supercritical boilers with six coal mills each capable of producing 600 gross MWs. DP&L embarked on a project to reduce their NOx emissions and improve heat rate in Unit #3 boiler was chosen as the first unit to evaluate for advanced process control. A Combustion Optimization System (COS) was installed on Unit #3 in mid 2003 and performance testing was completed in January A study was conducted as the first phase of a project. This evaluation identified potential NOx, heat rate and loss on ignition (LOI) benefits. The study also determined a baseline by which to measure the benefits of future improvements. Based on the positive conclusions of the study, a project proceeded. The project was commissioned and the multivariable control system is now a key element of DP&L s operation for maximum performance. Heat rate and reduced LOI benefits are obtained through coal pulverizer optimization, reduced air preheating, lower excess air, and balanced superheat and reheat steam temperatures. Coal pulverizer optimization is obtained by maximizing preheated air and minimizing cold air supply to the pulverizers. This optimization approach provides benefits by taking advantage of the boiler efficiency benefits of the air heater. Since air preheating is applied to improve electrostatic precipitator (ESP) performance, lowering opacity levels through improved furnace control allows reduced levels of air preheating, thus making more steam available to the turbine cycle. A side benefit of reduced opacity is the potential for lower unburned carbon, indicated by improved ESP performance. Lower excess air is achieved through the balancing of O 2 distribution in the furnace. Estimated Benefits from Study The study conducted to verify the justification for a project concluded that NOx and heat rate benefits of the boiler optimizer would be derived from a combination of effects associated with coal pulverizer optimization and O 2 optimization. The heat rate benefits are described first, followed by the NOx benefits. The estimation of benefits from the application of multivariable predictive control was determined by the comparison of baseline conditions during normal operation and test conditions simulating optimal operation. A significant portion of the NOx and heat rate benefits were based on a comparison of the initial settings of the control variables, such as primary air (PA) flow setpoint bias. If a different set of initial conditions had been applied, the benefits would have been different. The heat rate benefits were determined by calculations similar to Controllable Losses. In some cases, the benefits were extracted from the Controllable Losses display. For example, the O 2 and Reheat Spray Flow losses can be extracted from this display.

4 Heat Rate Benefits The study recognized that coal pulverizer optimization is achieved through a combination of primary air flow set point bias and pulverizer outlet temperature setpoint bias adjustments. The primary objective for coal pulverizer optimization is to minimize the flow of cold air (non-preheated) to each pulverizer governed by the following constraints: Maintain the pulverizer loading in a non-overloaded state Balance the furnace O 2 distribution Reduce the stack opacity levels By reducing the utilization of cold air at each Pulverizer, a higher percentage of the air to the combustion process is preheated, yielding a higher inlet air temperature to the furnace. An 18% reduction in the flow of primary air is expected through the reduction of cold air flow by the optimizer. This would raise the average air inlet temperature to the furnace by approximately 11 F resulting in an efficiency improvement of about 0.22% (STEAM, Babcock & Wilcox, Copyright 1978.). The primary air flow set point bias and pulverizer outlet temperature setpoint have a large impact on the measured O 2 distribution in the back-pass. By coordinating these control signals the O 2 distribution can be brought very close to a balanced condition. Since the production of CO is linked with low O 2 regions of the furnace, balancing the O 2 allows the average O 2 to be lowered without adversely impacting CO and unburned carbon. Stack opacity is directly affected by primary air flow set point bias and pulverizer outlet temperature setpoint bias. There is a significant heat rate benefit that can be achieved through reduced opacity. Due to high opacity the current operating approach was to increase the average cold end temperature through higher air preheating to increase the stack temperature and improve the electrostatic precipitator performance (ESP). With this higher stack temperature, the ESP could maintain opacity within constraint. Lowering the opacity through the optimizer decreases the need for air preheating and reduces the extraction steam consumption (16 th stage extraction) for the air preheater. Tests indicated that average cold end temperature setpoint could be lowered from 190 F to 170 F, resulting in more power production from the steam and a heat rate savings of 0.27%. An improvement in opacity is also an indication of a potential reduction of unburned carbon. Since unburned carbon degrades ESP performance, the lower opacity achieved during the test may be an indication of a significant reduction in unburned carbon. The feasibility test periods were too brief for reliable unburned carbon measurements; the potential improvement was not directly quantified. A reduction in unburned carbon of 1% was estimated for a heat rate benefit of 0.12%. Following commissioning the reduction in unburned carbon was confirmed. O 2 optimization and dynamic control coupled with O 2 balancing provides an opportunity to achieve a significant reduction in average O 2 without adversely impacting CO production and unburned carbon. The O 2 control approach was structured to address CO. Rather than controlling the average O 2, the lowest O 2 was regulated, since the lowest O 2 is the best precursor for CO. Also, the model predictive approach provides much tighter regulation of O 2 during both normal operation and upsets. This

5 approach reduces the average O 2 from 3.2% to 2.7% at full load for an average heat rate improvement of 0.16% based on dry gas loss calculations. NOx Benefits NOx benefits are derived from a combination of coal pulverizer optimization and O 2 optimization. Data from the coal pulverizer optimization test indicate a 6% reduction in NOx. The relationship between O 2 and NOx indicates a 5% reduction in NOx for the 0.5% lower O 2 levels. The combined effects result in 11% NOx reduction. NOx and Heat Rate Benefits Summary The table below demonstrates the benefits achievable through the implementation of the multivariable predictive NOx and Heat Rate Optimization system. These benefits represent improvements that can be made to the process using the study test period as a baseline. Table 1. Heat Rate and NOx Benefits Parameter Change Heat Rate Benefit NOx Reduction Pulverizer Optimization Primary Air Flow Reduced 18% 0.22% 6% O 2 Optimization Avg. O 2 Reduced 0.5% 0.16% 5% Pulverizer Optimization Unburned Carbon Reduced 1% 0.12% NOx and Heat Rate Optimization System Totals 0.50%* 11% * Reduced Opacity from Pulverizer Optimization achieved by reducing the average cold end temperature is expected to provide an opportunity for another 0.27% heat rate benefit. However, due to air preheater valve problems on this unit the cold end temperature is not reduced at this time. A snapshot of a portion of the screen built by DP&L to monitor the performance of the COS is shown below. This screen shows an improvement of 0.53% in heat rate, NOx reduction of 14.8% and opacity reduction of 20.5% yielding annualized savings of $370,060 per year. The data indicate that the performance objectives defined in the feasibility study were achieved by the implemented COS. Figure 1: Benefits Display

6 Multivariable Controller Design The Advanced Process Control (APC) strategies are designed to accomplish multiple objectives. Due to process interactions between mills, the combustion optimization system (COS) is comprised of a single multivariable predictive controller. The equipment covered by the controller is: Pulverizers A-F O 2 Control Air Preheaters The main objectives of the multivariable predictive controller are: Minimize NO x Emissions Maximize Thermal Efficiency Enforce specified operating and safety constraints Stabilize the unit operation in the presence of unmeasured disturbances Reduce operator workload In order to achieve the stated objectives the COS optimizes three separate areas of the unit simultaneously. These areas are the coal pulverizers (A-F), O 2 Control and Air Preheater control. Pulverizer Optimization (A-F) The six pulverizers are included in the pulverizer optimization. The pulverizers can be placed in or out of COS control as necessary. The objectives of the pulverizer optimization are the following: Minimize Cold Air to Pulverizer, subject to inlet and pulverizer temperature constraints Minimize NOx formation Decrease back end O 2 measurement variance (between different sensors) Reduce opacity The COS system manipulates the primary air bias to each mill along with the hot/cold air damper position. The DCS pulverizer temperature PID controller is disabled and pulverizer temperature control is taken over by the COS. The hot/cold air damper is allowed to move in the range of 0 100%. The primary air bias range is defined by operations between 10 to 25 %. The pulverizer temperature is controlled within high/low limits primarily by moving the hot/cold air damper. The Optimizer objective is to reduce the amount of cold air charged to the pulverizer, subject to mill temperature constraints and operational constraints. The primary air bias is minimized until operational constraints are reached. Pulverizer AMPS are included as a constraint.

7 Pulverizer Controller Variables (A-F) Variable Type Comments PA Bias MV Minimize to reduce NOx Emissions Hot/Cold Air Damper MV Maximize to decrease cold air to boiler Coal Flow FF Controller will adjust to changing coal flow Pulverizer Temperature CV Minimum/Maximum Constraint for safe mill operation. COS will typically maximize this temperature unless it is encountering the pulverizer inlet temperature constraint. Pulverizer Inlet Temperature CV Maximum constraint for safe mill operation. COS will maximize this value. Pulverizer Primary Air Flow CV Minimum constraint. This constraint represents the minimum flow at which the mill develops operational issues. Pulverizer Amps CV Maximum constraint. This constraint represents an overload point for the mill that may correlate to increasing rate of rejects. Pulverizer Coal Feed Pluggage Detection Pluggage detection calculations are included as part of the pulverizer optimization. This logic, based upon future predictions, is designed to give immediate notice that a pluggage event has occurred. A special controller is set up in the multivariable predictive control system to create these predictions. This controller generates future predictions of pulverizer amps and pulverizer temperature. It is known that the following conditions indicate a pluggage event; drop-off in pulverizer amps (because coal is no longer flowing to mill) increase in pulverizer temperature The multivariable predictive control continuously predicts the above two variables and the calculations monitor for departures from the predicted values. When pulverizer amps drop 3 AMPS below the prediction a mill pluggage is suspected and the first alarm is set. If in addition to this the pulverizer temperature increases 5 degrees Fahrenheit above the current prediction a pluggage event is very likely and a second alarm is set. Operational data is being gathered to ascertain the effectiveness of this system in providing early warning to the operators of coal feed pluggage. O 2 Optimization The objectives of O 2 optimization are: Minimize excess Oxygen Minimize NOx

8 Improve Boiler Thermal Performance Respond intelligently to unmeasured disturbances, such as mill pluggage events The O 2 optimization system takes over control of excess O 2 from the distributed control system. The resulting control and optimization is an improvement over the regulatory level control because it becomes more aggressive during upset conditions and will maintain tighter control of the boiler excess O 2. In addition the controller looks at not only the O 2 average but also the O 2 low select. The O 2 low select is the lowest of the O 2 measurements. By controlling against a low limit for the O 2 low select some margin is gained in further reducing the average O 2, without violating CO production concerns. This margin translates into NO x reduction and improved heat rate. The controller manipulates the O 2 controller output (Fuel/Air Ratio). The economic optimum is to lower the fuel/air ratio thus resulting in lower excess O 2, lower air flow and less heat loss from the stack. O 2 is minimized to either an O 2 Average low limit or O 2 Low Select Low limit. The low limits for each of these controlled variables adjusts automatically depending upon the current demand and number of pulverizer in operation. Furnace pressure, FD Fan A AMPS and FD Fan B AMPS are included as operational safety constraints. The furnace pressure high limit will temporarily override O 2 control, if necessary, during an upset condition to prevent a boiler trip on high furnace pressure. FD Fan AMPS (A&B) are included to prevent the COS system from adding too much air to the boiler. O 2 Optimization Variables Variable Type Comments O 2 Controller Output MV Minimized subject to low O 2 average of low O 2 low select. Reduced NO x emissions. O 2 Average CV Calculated Average Value. Minimized to low limit O 2 Low Select CV Lowest Single O 2 Measurement. Minimized to low limit. FD Fan A Amps CV Maximum Constraint to prevent overload of FD Fans. FD Fan B Amps CV Maximum Constraint to prevent overload of FD Fans. Furnace Pressure CV Maximum constraint to prevent O 2 control from nearing the trip point during an upset condition

9 Control System Interface The multivariable control system communicates with the basic control system through an interface with full read/write capability. Key information is read into the multivariable controller, control moves are calculated and in most cases setpoints are sent to the basic PID controllers. In some instances the multivariable controller writes directly to valve outputs. ACTUAL CONTROL SYSTEM BENEFITS The combined benefits result in NOx reductions greater than 10% and heat rate improvements of 0.5%. The following is a screen capture demonstrating the benefits. On the far left side of the trend the COS system is ON. During the center portion of the screen COS was OFF and the unit was returned to normal conditions. The right side of the trend is where multivariable predictive control is again ON. This trend demonstrates the following benefits; > 0.5 % Heat Rate Improvement, > 10% NOx Reduction. The data was observed on 10/21/2003 at 3 PM. A short term testing program was conducted over several weeks in order to ascertain the minimum O 2 level, based on the average of 6 oxygen probes that the system can achieve without adverse effects on the boiler. The O 2 setpoint was lowered each week and manual readings were taken 3 times a week for CO and NOx. The lab also analyzed the ash for LOI while the boiler was visually inspected for slagging. The results of this testing confirmed the operational constraints for the COS. SUSTAINING BENEFITS The long-term viability of a multivariable control system hinges on its ability to provide sustained benefits. Multivariable predictive control includes a variety of tools to assist in maintaining the robustness of the controller. Although the controller does not require much maintenance, it is important for DP&L personnel to be familiar with the system and perform routine tasks with the controller. In addition to informal training provided during the course of the project, a couple of DP&L engineers attended a formal training course to become more familiar with the functions of the system. The system is being maintained by DP&L at this time. CONCLUSION Heat rate benefits of greater than 0.5% and NOx reductions of greater than 10% were obtained through coal pulverizer optimization and lower excess air. The reduction of operating margins to achieve these benefits was made possible through the application of multivariable predictive control.

10 With the success of this initial installation on Unit #3, further projects are underway to apply the same advanced process control and optimization to Units #1, 2 and 4.

11 Figure 2: Performance Test Trend: COS System Post- Commissioning OPT ON OPT OFF OPT ON Future Improvements Future improvements to the system that could potentially improve the overall optimization of the system and bring significant benefits to DP&L. are: An automatic Smart Soot Blow System Additional heat rate benefits could be derived by a Smart Soot Blow system. A significant heat rate penalty is sometimes suffered by the imbalance between superheat and reheat steam temperatures or by high reheat spray flows. The heat rate penalty associated with spray flow can be substantial. The only means of adjusting the distribution between the superheat and reheat sections is through soot blow action. Such a system is configured to address this energy distribution task and could have a significant impact on heat rate performance. Dynamic Steam Temperature Control Dynamic steam temperature control can allow the unit to achieve faster ramp rates. It may also allow the unit to achieve maximum heat rate and NOx benefits at low loads when steam temperature may otherwise be a constraint on the boiler operation.

12 REFERENCES 1. Sandoz, Perspectives on the Industrial Exploitation of Model Predictive Control, 1988 Lecture, Measurement and Control, Vol. 31, May Bakul, C.E. (editor), The John Zink Combustion Handbook, CRC Press, Boca Raton, Fla., Babcock & Wilcox Company, Steam Its Generation and Use, 39 th Edition, Shinskey, F.G., Energy Conservation Through Control, Academic Press, New York, New York, Gas Processors Suppliers Association, Engineering Data Book, Volume 1, Tenth Edition, Section 15, VanDoren, V.J., Model Predictive Controller Solves Complex Problems, Control Engineering, March, Gottier, M., Runkle, D. and Flint, M., Performance-based Contract Maximizes IT Payback, Power, July/August, A.S.M.E., Steam Generating Units, Power Test Codes 4.1, Section 5.0, Reaffirmed Nix, A., Morrow, A., and Gordon, L., Using Multivariable Predictive Control to Optimize the ASARCO Mission Mill, 2000 SME Annual Meeting, February, Flint, M. and Gottier, M., Model Based Control for Process Performance Enhancement, Joint ISA POWID/EPRI I&C Conference, St. Petersburg, Florida, June 14-18, Gordon, L., Roberts, D., and Labbe, D., Optimizing Heat Rate With Model Predictive Control On Riley Turbo-Furnace Units, Joint ISA POWID/EPRI Controls Conference, San Diego, California, June Lange, G., Labbe, D., Mahajanam, R., and Litchko, T., NOx & Heat Rate Supervisory Control at NRG-Huntley Operations by, Presented at Joint ISA POWID/EPRI Controls Conference June 2002, San Diego, California. 13. Runkle, D., Chapa, R., Labbe, D., and Morrow, A. Sim Gideon Station: Multi Variable Control for Enhanced Dispatch and NOx Mitigation, 13th Annual Joint ISA POWID/EPRI Controls and Instrumentation Conference, Williamsburg, Virginia, June